YT YouTube2Text

YouTube2Text → Transcripts → Beyond DX: AI Transformation Forum for the Philippine Public Sector

Beyond DX: AI Transformation Forum for the Philippine Public Sector — Transcript

by Development Academy of the Philippines · 23,614 words · 4,060 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:04everyone
0:05to President Lucajio Sebastian of the
0:08Development Academy of the Philippines,
0:12Dr. Indra Pradana Shinga Vinata uh the
0:16Secretary General of the Asian
0:17Productivity Organization,
0:20Under Secretary Maria Francesca Del
0:22Rosario of the Department of uh Energy,
0:27Assistant Secretary Angelica Serento of
0:31the Department of uh Finance, Assistant
0:34Secretary Ron De Guzman of the
0:37Department of the Migrant uh workers,
0:41Dr. Jose uh Elvenia of the Asian
0:45Productivity Organization and Dr. Arani
0:49Buana of the APO, our delegates uh APO
0:53delegates from uh Fiji and I understand
0:58from Mongolia and uh maybe other
1:02countries who uh opted to join uh this
1:06forum to our officers of the DAP led by
1:10our acting senior vice president for
1:13programs uh Aya Luen, vice president M.
1:17Neves, director Tristan uh Seratus and
1:20all the officers of the DAP, our
1:23colleagues in government, partners from
1:25the private sector, distinguished uh
1:28speakers and exhibitors and fellow
1:31public servants. I warmly welcome you to
1:34all or rather all to the beyond DX AI
1:39transformation forum for the Philippine
1:41public sector.
1:43Every day, citizens experience
1:46government through very practical
1:48encounters.
1:50Applying for a permit, accessing health
1:53care, requesting assistance, or waiting
1:56for a decision. For them, the value of
1:59transformation is measured in clearer
2:02answers, shorter waiting times, and
2:05services that they can depend on. These
2:09are the outcomes that should guide our
2:11conversation on artificial uh
2:14intelligence. Beyond DX expresses an
2:17aspiration for government to build on
2:21the foundations of digital
2:22transformation
2:24and explore how AI can responsibly
2:27improve public services, workflows as
2:31well as decision making. Pursuing this
2:34vision requires capable people, sound
2:37institutions, and a clear understanding
2:40of the public value we seek to create.
2:43So when the opportunity came for the
2:46development academy of the Philippines
2:48to lead AI capability building for the
2:51Philippine public sector, we embraced
2:54it. We saw an opportunity to play a
2:57critical role in transforming how
3:00government works and how it serves our
3:03people. We are translating this
3:06commitment into action through the
3:09national capability building program on
3:12artificial intelligence.
3:15Aligned with the national AI strategy
3:18for the Philippines, this program equips
3:21government leaders to guide AI adoption,
3:25for technical specialists to develop and
3:28implement solutions, and for public
3:31servants to apply AI responsibly in
3:35their daily work. And we are already
3:38seeing progress. I'm pleased to report
3:41to you that as of October 9, the program
3:45had already recorded 1,215
3:49training completers
3:51and as I understand this already
3:55surpassed our target for the year. Our
3:58online foundational course that is the
4:00AI 101 had reached 3,228
4:05enrolles.
4:07One-third of the enrolles already
4:10successfully completed a course and we
4:13have trained over or around a hundred
4:16officials on AI course for local
4:19government units in part in partnership
4:23with the Asian productivity
4:25organization. We have also trained 35
4:28executives in AI governance,
4:31strengthening their capacity to guide
4:33responsible AI adoption within their own
4:37institutions.
4:38And within the academy, we are building
4:41our pool of AI trainers. And while on
4:44training, they develop AI capstone
4:47projects, three of which will be
4:50showcased on stage today. Through this
4:53investment, we are strengthening our own
4:55capability to teach, mentor, and support
4:59government institutions and putting into
5:02practice the capabilities we seek to
5:04build across the public sector.
5:08Our work also extends to understanding
5:12the institutional conditions that enable
5:15AI adoption.
5:17A few years back, we partnered with the
5:20Asian Asia-Pacific
5:23uh center on information technology in
5:26Korea to conduct short courses and
5:30executive orientation on digitalization
5:33for senior officials.
5:35Earlier this year, we partnered with the
5:38British Embassy in Manila to convene our
5:41senior government executives for a
5:43roundt discussion facilitated by experts
5:47from the London School of Economics and
5:50Political Science to discuss barriers to
5:53AI adoption and investment including
5:56challenges in leadership, data
5:58governance, procurement, and
6:01accountability.
6:03These accomplishments provide a
6:05foundation for the next stage. We are
6:08preparing further learning opportunities
6:10in planning, human resource management,
6:14financial management and advanced
6:16applied AI which we aim to launch
6:21towards the end of this quarter.
6:23Today's forum gives us an opportunity to
6:27see AI in action, understand what it
6:30takes to apply it effectively in
6:32government, and consider the safeguards
6:35needed to uphold public trust. I
6:39encourage everyone to ask what service
6:42problem does this application address?
6:46How will they know it if improves
6:48outcomes? What data and capabilities are
6:52required and how will we preserve
6:55accountability and public trust? To our
6:58speakers, to our exhibitors and
7:00partners, thank you for sharing your
7:03experience and expertise, and to our
7:06fellow public servants, thank you for
7:09your willingness to learn and help shape
7:11this transition. May today's exchanges
7:15lead to practical collaboration and
7:18thoughtful action within our
7:20institutions. Welcome to Beyond DX and
7:23let us advance an AI ready Philippine
7:27public sector that serves our people
7:30with greater competence, responsiveness,
7:33and care
7:35and good morning.
7:44Thank you, Executive Vice President
7:46Mendoza.
7:48Next, we have the honor of hearing a
7:51special message from the Secretary
7:54General of the Asian Productivity
7:56Organization. Please welcome Dr. Indra
7:59Pradana Singa Winata.
8:19Thank you, MC
8:22Under Secretary Maria Francesca del
8:24Rosario of the Department of Energy of
8:27the Philippine government.
8:30My dear friend DAP President and CEO Dr.
8:34Sebastian
8:35and the Magdalena and all my DAP
8:39friends, distinguished officials of the
8:42different government agencies in the
8:44Philippine government, fellow public
8:46sector leaders, distinguished guests,
8:48ladies and gentlemen, good morning,
8:50magumaga.
8:53Let me begin by congratulating the
8:56development academy of the Philippines,
8:58DIP and its center of strategic futures
9:01for convening this very important forum
9:04on moving beyond digital transformation
9:07toward AI enabled governance. I also
9:11commend the Philippines for its efforts
9:14to strengthen public sector AI
9:15capabilities, particularly through the
9:18AP's newly established national capaci
9:22capability building program on AI, which
9:25responds to the evolving needs of the
9:27public sector amid the rapid advancement
9:31and expanding applications of AI.
9:34The theme beyond the X AI transformation
9:38for the Philippine public sector does
9:41not mean abandoning
9:43digital transformation rather it means
9:46building on our digital foundations to
9:48rethink how government operates make
9:51decision and deliver services to the
9:54citizens.
9:56The real question is no longer simply
9:58how government can digitize existing
10:01processes.
10:03It is how we can use AI to reimagining
10:07those processes, improve decisions and
10:11create greater public value. Ladies and
10:14gentlemen, AI offers significant
10:16opportunities to improve public service
10:18responsiveness, operational efficiency,
10:22evidence-based policym and institutional
10:25performance. However, acquiring AI tools
10:29alone does not guarantee these benefits.
10:32As highlighted in the APO publication,
10:35AI readiness, capability gaps, and
10:38productivity pathway in selected APO
10:41members, which is released in June 2026
10:44this year. AI adoption across our member
10:47economies remains uneven. Differences in
10:50digital infrastructure, human capital,
10:54institutional readiness influence how
10:56effectively economies can translate AI
11:00investment into real productivity gains.
11:04This bring us to the fundamental
11:06principle. The measure of AI
11:08transformation is not how many AI tools
11:11and institutions deploy but whether
11:15those tools really help deliver better
11:18outcomes with the resources available
11:21for public sector institution. This
11:23means shorter processing times, more
11:26accessible services, better informed
11:29policy policy decisions and greater
11:32staff capacity to focus on higher value
11:34work. These outcomes must be
11:37demonstrated and measured not simply
11:40assumed particularly in frontline
11:42service delivery policy formulation and
11:46legislation where government decisions
11:48can have farreaching effects on the
11:50citizens lives to move from digital
11:53transformation to genuine AI
11:55transformation. I would like to empas
11:57emphasize three priorities. First build
12:01the foundations for AI ready government.
12:04This requires clear governance, reliable
12:07and interoperable data, secure digital
12:11infrastructure and appropriate
12:13accountability mechanisms.
12:16Responsible AI governance should not
12:18view as a barrier to innovation but a
12:22foundation for public trust and
12:25sustainable adoption.
12:27Second, move from experimentation to
12:30institutional adoption and skill.
12:33Government agencies should begin with
12:36clearly defined public sector challenges
12:38rather than looking for problems to fit
12:41available technologies.
12:44AI applications must be tested against
12:47measurable objective evaluated based on
12:50result and scaled when they demonstrate
12:52genuine value. The ultimate goal is to
12:56embed effective solutions into
12:58institutional practices and deliver
13:00lasting improvement. And third, build
13:04transformation from within. AI
13:07transformation is as much a human
13:10capability and organizational challenge
13:13as a technology technological one.
13:16Public officials need to skill to use AI
13:19effectively while leaders must be
13:21prepared to redesign workflow, manage
13:24organizational change, and maintain
13:26appropriate human oversight rather than
13:29simply adding AI to existing processes.
13:33We must be willing to rethink how work
13:36is organized and how public value is
13:39created. The Asian Productivity
13:41Organization, APO, stands ready to
13:44support this transformation through
13:46research, capacity building, and
13:48regional cooperation. The APO's 2026 AI
13:52readiness study examines the
13:54capabilities gaps, cooperation needs of
13:5711 APO member economies. It underscores
14:01the importance of differentiated
14:04demand-driven approach and recognizing
14:06that each economy has different starting
14:09points in institutional and capacity.
14:13During the 67th workshop meeting of
14:16heads of NPOS's hosted by the Philippine
14:19government through the AP over the past
14:22two days, we also shared updates on one
14:25of the APO's flagship initiative,
14:27genuine AI action or GAA APO. through
14:32GAA APO. The APO promotes knowledge
14:34sharing and capacity building on AI
14:36governance, transparency and responsible
14:39adoption within the framework of
14:41Hiroshima AI process or hype through
14:44communications, collaborative research
14:47and talent development activities. GAA
14:49APO connects stakeholders across
14:51government, industry, international
14:54organization and also academia
14:57university to help translate AI
14:59knowledge into practical applications.
15:02We also recognize the Philippines
15:04potential to contribute to regional
15:07learning on public sector AI. Under the
15:10Gaia initiative, the country has
15:12demonstrated
15:14strong participation in GCI world, a
15:17free AI and data science learning
15:19program associated with the Matsuo
15:21Iwasawa Laboratory, the University of
15:24Tokyo. Since the program was launched in
15:282014, the Philippines has the highest
15:32recorded 3,29
15:36uh 911 participants, once of the highest
15:39participants in this GCI program by Maso
15:43Lab, University of Tokyo. This
15:45experience underscores the importance of
15:48investing the people and creating
15:50opportunities to apply their knowledge
15:52to real world challenges.
15:54Through stronger cooperation with the AP
15:56and other institution both within the
15:58Philippines and across the region, we
16:00can further strengthen the public sector
16:02capabilities, exchange implementation
16:05experiences and identify practical
16:08opportunities for collaboration at both
16:11national and regional levels. No economy
16:14needs to navigate the AI transition
16:17alone. By sharing knowledge, developing
16:20talent and learning from practical
16:22implementation, APO members can
16:24accelerate responsible AI adoption and
16:27broaden its productivity benefits. Let
16:30me also briefly share the experience of
16:32my home country Indonesia where AI is
16:36being applied across various areas of
16:38public sector including social assistant
16:42target social assistant targeting
16:45immigration services cyber security and
16:48threat detection civil service
16:50productivity among other areas. The
16:54application demonstrate the potential of
16:56AI to move beyond digitalization toward
16:58measurable improvement in public sector
17:00productivity. The key question now is
17:03whether this initiative translate into
17:07faster service delivery, more accurate
17:10decisions, reduce administrative burdens
17:14and ultimately better outcomes for the
17:16citizens. As I con as I conclude, I
17:19encourage each participating agency to
17:22identify priority challenges where AI
17:26could deliver tangible public value.
17:29Define clear objective. Establish
17:31appropriate safeguards. Equip your
17:34people.
17:36Equip your people with the necessary
17:38skills and measure of the results.
17:41Moving beyond digital transformation
17:43means moving beyond technology
17:45deployment towards institutional
17:47transformation. Let us use AI not simply
17:51for automate what government does today
17:54but to build public institution that are
17:57more productive, responsive, accountable
18:00and prepared for the future. The APO
18:03stands ready to support your journey.
18:07Thank you.
18:17Thank you, Secretary General
18:19Singawinata, for reminding us that AI's
18:22value lies in measurable improvements in
18:25public service and for the support of
18:27the Asian Productivity Organization in
18:30this journey.
18:32Now for the keynote address. Our speaker
18:36previously served as under secretary and
18:39head of the information and
18:41communications technology group at the
18:43department of budget and management.
18:46Prior to entering government, she was
18:48vice president and head of artificial
18:51intelligence and data policy at Union
18:54Bank of the Philippines where she
18:56spearheaded the open digital trust
18:58framework for the Institute of
19:01International Finance. She also held key
19:05technology governance roles at Indra,
19:08Globe Telecom and TV5
19:11to deliver beyond DX reimagining
19:14government and public value in the age
19:16of AI. Let us warmly welcome the under
19:20secretary, chief information officer and
19:23head of information technology and
19:25management services at the department of
19:28energy, attorney Maria Francesca M. Del
19:32Rosario.
19:47Good morning everyone. Thank you to the
19:51development academy of the Philippines
19:52and um center for strategic futures for
19:57for APO for convening this forum under
20:01the AA AI capability building program.
20:06I come from the energy sector now. Um I
20:09just joined a few months ago. A place
20:11where systems are not abstract. So when
20:15they work the lights stay on, hospitals
20:19run and the economy moves. When they
20:22fail, everyone feels it. That is the
20:27lens I want to bring to beyond DX.
20:32My message this morning is simple.
20:37AI transformation is not a technology
20:41project. It is a decision about the
20:44public value we want to create and about
20:48whether people can trust the government
20:51that creates it.
20:55Most of us live through the first two
20:57waves
20:59here. We digitized. No paper became
21:04PDFs. Cues became portals. Then we
21:08pursued digital transformation. We
21:11redesigned processes, connected systems
21:14and built eervices end to end. When I
21:18joined DBM four years ago, that's what
21:20we did. Um, and um, but recently with AI
21:24transformation, we had to restrategize
21:28because AI transformation is different
21:30in kind.
21:33It does not only move a process online.
21:36It can shape the decision itself from
21:40who gets served first to which feeder
21:43gets restored first and how a permit is
21:47reviewed. So the question changes no
21:51with digitization we ask is it online?
21:54With DX we ask is it seamless?
21:58With AI transformation, the question is
22:01now here is it better for the people and
22:06do they trust it?
22:08With every agency will begin from a
22:12different starting point that is fine.
22:16Beyond DX is not a race to adopt tools.
22:20It is a deliberate choice about the
22:22government we want to become.
22:28If you take one idea from this forum,
22:32let it be this. Public value is the real
22:36measure.
22:38It is tempting to count deployments. How
22:41many chat bots, how many models, how
22:44many dashboards we've created. Citizens
22:48do not experience deployments. They
22:51experience a faster permit, a clearer
22:54answer, a shorter queue or an outage
22:58restored sooner. For energy, that's very
23:01important. I see four kinds of value.
23:05Better services for the people we serve,
23:08better decisions for leaders.
23:11We need to act earlier with evidence,
23:14better public work so our people spend
23:17less time on repetitive tasks and more
23:20on judgment and empathy and underneath
23:24of it greater public trust.
23:28Trust is the one we can lose fastest. A
23:32single failure in a critical service can
23:35set adoption back for years.
23:40So let me bring this home to energy
23:44because our sector sees AI from both
23:48sides.
23:49First, AI runs on power.
23:53The Philippines AI++ infrastructure
23:57master plan envisions AI data center
24:00capacity growing from about 50 megawws
24:03to 1.5 gawatt by 2033.
24:08with a target of 40% of that power
24:11coming from renewables.
24:14The commitment is that this demand will
24:16be met without drawing power away from
24:19Filipino households.
24:21Every AI ambition in this room depends
24:25on a reliable, affordable, and
24:28increasingly clean grid.
24:31Second, power can run smarter with AI.
24:36better forecasting of demand and of
24:39variable renewables, faster outage
24:42response, simpler permitting and
24:45application processing, clearer
24:48information for consumers. These are the
24:50opportunities, not finished projects.
24:54Energy is also where the stakes are
24:56highest, which makes it the right place
24:59to learn how to use AI safely.
25:05Now the other side my original field is
25:08cyber resilience
25:10and the lesson there applies directly to
25:13AI. When systems are trusted they become
25:18tr targets.
25:20An AI system in public service is
25:23critical infrastructure for trust. Put
25:26simply, a bad data means bad decisions
25:29out.
25:31Technical teams call this data
25:33poisoning. Every cloud service provider
25:36uh vendor and connection is another door
25:39to guard. If a decision cannot be
25:42explained, it cannot be defended or
25:45corrected. And when AI fails feasibly,
25:49the public's trust fails with it.
25:53So picture the scenario. This is
25:55illustrative and not a reported
25:57incident. A forecasting tool is fed
26:00manipulated sensor data. It produces
26:04confident plausible wrong projections
26:06dispatch decision drift until
26:08experienced operators notice the
26:11mismatch and fall back to manual
26:13procedure. The question I want to ask
26:15every agency that are um that is here is
26:19is our AI system compromised? Can our
26:23people still keep the service safe and
26:26running? If the answer is no, we are not
26:30ready to rely on it.
26:34The good news is that we are not
26:36starting from zero.
26:40Executive order 58 made the national
26:43cyber security plan our whole of nation
26:45roadmap in energy. The department or
26:48circular DC2025-01-01
26:54set up the energy sector cyber security
26:56and cyber resilience framework built on
26:58governance, compliance, collaboration
27:02and capacity building. In June this
27:04year, the DICCT and civil service
27:06commission issued JMC3,
27:09our shared principle for ethical and
27:12responsible AI in government. It applies
27:15to national agencies, GOCC's, LGUs, and
27:18state universities alike. And the AI
27:21infrastructure master plan sets out the
27:24compute power and skills behind it all.
27:28So look at the JMC principles, people at
27:32the center, human accountability,
27:35transparency, security, reliability, and
27:38these are just the principles that keep
27:41a grid safe. So we are not short of
27:44policy. The test now is consistent
27:48execution
27:49across agencies and LGU.
27:53So what do we do? Um what you actually
27:56do on Monday um for foremost. So
27:59whatever your starting point is one with
28:02a public problem.
28:04So another tool find the pain point
28:07citizen feel such as long quue the slow
28:11approval of unanswered query and define
28:14what better
28:16uh looks like before you choose any
28:18technology. Two get the foundations
28:22right. AI is only good as a data under
28:25it and only as safe as the systems
28:28around it. Govern your data and sec and
28:33secure identity cloud and vendor access
28:36from day one not after the pilot.
28:40Three, keep humans accountable.
28:43High impact decisions need human review,
28:47explainable outputs, and a safe
28:50fallback. The same way our control rooms
28:53can always revert to manual operations.
28:56And four, build people and measure
28:59value. Programs like um the AI capacity
29:03building exists exactly for this. Then
29:06measure what citizens feel, time saved,
29:10access widened and trust earned. Uh I'm
29:14also a product of the um AI capacity
29:18building of the DAP. So
29:20ongoing
29:22um uh lectures and discussions no since
29:27March.
29:29Um so to make this concrete
29:33here is an agenda for any agency or LG
29:36you can adopt. So these are just
29:38suggestions no um not mandates. So in
29:41the first 30 days you name who is
29:43accountable have one leader and
29:45operational counterpart. Take stock of
29:48where AI is already in use, often
29:50quietly through the providers. Then
29:53choose one public service problem worth
29:56solving. Then the next 90 days, run a
29:59bounded pilot that follows the JMC3's
30:02principles. Before you rely on it, hold
30:05a tabletop exercise where you can assume
30:08the AI fails or is manipulated and
30:11practice what your people do. Enrolled
30:13your team in AI capacity building. And
30:16by 12 months, measure your outcomes,
30:18citizens feel, scale what works, retire
30:21what doesn't, and share the lessons. The
30:24fastest way to transform government is
30:27for agencies to learn from each other.
30:34Let me close where I began. The question
30:37in front of us is no longer do we have
30:39AI. Every agency will have AI soon
30:43enough through providers, platforms, and
30:47our own people. The real question is
30:50whether AI makes government more useful,
30:54more fair, and more worthy of public
30:56trust. I invite every leader here to
31:00make three commitments.
31:02Lead with purpose, name the public
31:06problem, and the person accountable.
31:10Build trust in by design. Secure
31:13explainable systems with a safe human
31:17fallback.
31:18Prove the value report outcomes citizens
31:22can feel.
31:24In uh when I was still in the private
31:27sector, one of my major projects was the
31:30uh we built an AI
31:32uh it's a it's a fraud detection model
31:35that runs on AI. So AI learns from past
31:40data. No. So um so we feed it the the
31:43the data there. But what we also
31:46discovered in in that particular
31:48activity is that there is um uned
31:53um discrimination for some some fraud
31:56detection.
31:58Uh let's say uh this particular person
32:00is prone to fraud. But if we um delete
32:05some of the data fields, it might be
32:08more fair. Example um for of course for
32:10credit scoring. So most of the time um
32:16this particular group of people which we
32:18call um in in in AI governance that's
32:21what we call um indiscriminate
32:23disadvantagement. Uh it might sound um
32:27very uh very technical or philosophical
32:31at most. Uh but it it means that there
32:35are times that the AI reads the model
32:38based on the
32:42perceived human processing or the value
32:45judgment of the humans that process the
32:47information because that's what the date
32:49that what's the AI understands how the
32:53data was processed in the first place.
32:56Example,
32:58if if the bank is has a tendency to to
33:02favor men over women because the the one
33:07checking the application forms are
33:09mostly men. You understand where I'm
33:12going at? So they tend to favor men for
33:16credits for loan applications.
33:20So what we did was we we we created um a
33:24governance AI governance mechanism which
33:26we call um fairness by elimination. So
33:30when we did the fairness by elimination,
33:32let's say we take out the gender data
33:34from the from the model, the AI
33:40um output and outcomes, they are
33:43different. They're more fair.
33:46Um so we we tend to um uh have more
33:52clients. That's um across the board. And
33:56um and as I mentioned
33:59we there's no more indiscriminate
34:02disadvantagement.
34:04Okay. So this is where you are very
34:07important as as um government where know
34:12um officials no or employees.
34:16So if you're going to run models let's
34:18say for for certain people to get
34:21benefits or for certain um sector to get
34:26IUDA or you know financial assistance
34:29and then you run it in the AI model the
34:32tendency is um
34:35the the valid judgment of the of the
34:37people doing that prior to you that's
34:40the same valid judgment that the AI will
34:42adopt. So your guard rails right now as
34:46government um employees for AI
34:49governance is that you have to make sure
34:52that what the AI produces as output and
34:56outcome is fair
34:59so that um you get to if you're in in
35:02the SWD or you get to have uh you the AI
35:06can un can feed you a more fair
35:10um list of people who are supposed to
35:13receive receive benefits, right? So, so
35:16um that's my challenge to to everyone.
35:19Um it's based on experience and I hope
35:21that this is something that you can also
35:24use in your um in your respective um
35:28agency. I every time I'm asked to talk
35:32about AI, I always take this opportunity
35:36[snorts] to highlight no
35:39um we are in the Asaian region that's
35:44stricken by calamities you know name a
35:46calamity we have it right now we have El
35:49Nino we have flood we have volcanic
35:51eruption we have earthquake
35:53but if we can leverage on that context
35:58in in artificial artificial
36:00intelligence. It AI gets AI models get
36:05more intelligent or smarter
36:08when they're being fed into a context
36:13that will make them better.
36:16So if you can invite like AI companies
36:19elsewhere in the world to come train
36:21their models in the Philippine context,
36:25their AI models will become smarter and
36:29more intelligent. Like um
36:32if you invite self-driving cars
36:36to train their models in Manila traffic,
36:39mind you, it's going to be the best car
36:42in the world. Um that's and and the and
36:45the Philippine government can benefit
36:46out of that like um because they put
36:49their um they put their models within
36:52the context of the Philippine situation.
36:56No. Um we can get royalties out of it.
37:01uh I don't know what kind of mechanism
37:02that you know can can pursue it but this
37:06is something that um that AI can can
37:11benefit our country as a whole. So those
37:14are just some thoughts in your head like
37:16you know you can think later on and um
37:19with that um
37:22so in energy we judge a system whether
37:25the lights stay on and when it matters
37:28most. So let us hold AI in government to
37:32the same standard. So beyond the X from
37:35digital government to public use and
37:38with that maming salamat po.
37:45Thank you under secretary del Rosario
37:48for imparting insights on beyond DX and
37:50how the government can advance toward AI
37:53transformation. May we request [music]
37:56that you return on stage as we present
37:58our certificate of appreciation and
38:01token of gratitude. [music] To present
38:03these, we also invite our officers,
38:06President Sebastian, Executive Vice
38:09President Mendoza, Vice President Kuan,
38:12and Director Soratos to join us on
38:14stage.
38:16Allow me to read the citation. The
38:19Development Academy of the Philippines
38:21presents this certificate of
38:23appreciation to the under secretary of
38:26the Department of Energy, Attorney Maria
38:28Franchesca M. de Rosario for delivering
38:31the keynote address during the Beyond DX
38:34AI Transformation Forum for the
38:36Philippine public sector.
38:51given this 22nd day of October 20 26 at
38:54[music] Cornra Manila Pasai City signed
38:57by the president and CEO Dr. [music]
38:59Leoko Sebastian.
39:11Once again a warm round of applause for
39:13Under Secretary Del Rosario.
39:22And another one of course we'll also be
39:25recognizing our secretary general Singa
39:28Winata.
39:30May we request our DAP officials to
39:32return back on stage.
39:36We'll also be providing a certificate of
39:38appreciation and token of appreciation
39:42for our secretary general for the APO.
39:45Allow me to read the citation.
39:51May we also call on our VP Maidan to
39:55join [music] us on stage.
39:59Allow me to read the citation. [music]
40:01The Development Academy of the
40:03Philippines presents this certificate of
40:05appreciation to the Secretary General of
40:07the Asian Productivity Organization, Dr.
40:10Dr. Indra Pradana Singawinata for
40:13delivering the special message from the
40:14APO during the [music] beyond DX AI
40:17transformation forum for the Philippine
40:19public sector conducted by [music] the
40:21academyy's center for strategic futures
40:25signed Dr. Leucio Sebastian.
40:28A warm round of applause [music] for our
40:30secretary general.
40:47[music]
40:51May we now invite our officials on stage
40:54to join [music] everyone on the floor
40:56for our group photo.
41:01I [music] also request everyone to stand
41:03and move forward toward the center of
41:06the room. Gather in the middle, please.
41:09And please face [music] our
41:10photographers on stage and give your
41:13most charming smiles.
41:17Hey, hey, hey.
41:56Three,
41:58two,
42:00one. Oip. Olet. Three, two, one. Smile.
42:08Say cheese.
42:16for our video. Kindly wave.
42:23Thank you. Thank you everyone. You may
42:25now return to your seats.
42:42Now to help us explore what beyond DX
42:45could mean in practice, we asked
42:47government leaders and our friends at
42:49the APO to share their perspectives on
42:52AI adoption in public sector and lessons
42:55from their own experience. We begin set
42:59one building the foundations for AI
43:01ready government with perspectives on
43:04governing the shift from the executive
43:07director of the National Innovation
43:08Council Secretariat Director Dian Gail
43:12Maharajan. Let us watch.
43:18>> Government, private sector, academ,
43:21civil society, development partners and
43:23all stakeholders committed to advancing
43:26responsible digital transformation in
43:29the Philippines.
43:31Artificial intelligence is rapidly
43:33transforming how economies operate, how
43:35industries innovate, and how governments
43:38serve their people. Globally, AI is
43:41expected to contribute around $15.7
43:44trillion to the world economy by 2030.
43:48In the Philippines, the AI market is
43:50expected to grow by 27.75%
43:54annually from 2025 to 2030, reaching
43:58$3.487
43:59billion by 2030. Although the Philippine
44:03market remains modest compared with the
44:05global market, this projection showed
44:07that AI adoption is accelerating locally
44:10and creating more opportunities for
44:13investment, innovation, and economic
44:15growth. The potential benefits of AI are
44:18evident across sectors. In health, for
44:21instance, AI can assist doctors in
44:24diagnosing diseases and planning
44:26treatments. In the workplace, it can
44:28automate written task, allowing workers
44:31to focus on higher value activities. In
44:34agriculture, it can help farmers monitor
44:36crops, predict harvests, and manage
44:39resources more efficiently. For
44:41businesses and government, it can
44:43improve decisions and deliver better
44:45services at a lower cost. However, these
44:48opportunities also come with known and
44:50unknown risk including the threats to
44:53the data privacy and security,
44:55algorithmic bias, lack of transparency
44:58and potential misuse and even technical
45:01capabilities, fragmented institutional
45:04arrangements and limitations in data
45:07quality and interoperability could
45:10further constrain responsible AI
45:12adoption. The challenge before us is
45:14clear. How do we harness AI's potential
45:18while safeguarding people's right and
45:20maintaining public trust? This is
45:23precisely what the Department of Economy
45:25Planning and Development seeks to
45:27address through the country's AI
45:29governance framework. The framework
45:31brings together ethical principles,
45:34technical standards and assurance
45:36systems, riskbased policy instruments
45:38and enabling resources to support
45:40responsible AI development, deployment
45:43and use. Developed through a whole of
45:46government and whole of society
45:48approach, the framework seeks to
45:50establish a trusted, inclusive, and
45:52ethically governed AI ecosystem that
45:55accelerates innovation and drives a
45:57globally competitive Philippines. It
46:00brings together ethical principles,
46:03standards, and assurance mechanism, a
46:05multi-tered riskbased policy toolkit,
46:08and enabling resources to support
46:10responsible AI development, deployment,
46:12and use. Together, these mechanisms seek
46:16to ensure two complimentary outcomes.
46:19Human- ccentric AI development,
46:21deployment, and use that uphold people's
46:23rights, dignity, safety, and well-being.
46:27And a riskbased policy instruments that
46:30applies proportionate safeguards based
46:32on the nature, use, and potential impact
46:35of AI systems while allowing innovation
46:38to drive. Importantly, the framework
46:41recognizes that responsible AI adoption
46:44requires an enabling environment that
46:47supports innovation and builds the
46:49country's readiness for emerging
46:51technologies. To do this, we must foster
46:54a strong enabling ecosystem, which is
46:57why the framework identifies five
46:59mutually reinforcing enabling resources.
47:03Institutions,
47:04data and R&D, capabilities,
47:07infrastructure, and demand generation.
47:10For institutions, we must strengthen the
47:13government's capacity to assess,
47:15regulate, and oversee AI systems while
47:19integrating an agile governance into
47:21institutional practices so policies
47:24remain responsive to emerging risk and
47:27technologies. For data and R&D, we must
47:30invest in a stronger research
47:32institution, highquality data sets,
47:35secure and interoperable data systems
47:38that enable secure information sharing
47:41and innovation across sectors. For
47:44capabilities, we must develop a future
47:47ready AI workforce through national
47:50competency standards, AI literacy,
47:53upskilling and reskilling, scholarships
47:56and workforce transition mechanism. We
47:59must also ensure that this opportunities
48:02reach the underserved and vulnerable
48:04groups so that no Filipino is left
48:06behind in this transformation.
48:09For infrastructure, the framework, we
48:11must prioritize expanding access to
48:14affordable and secure connectivity,
48:17scalable computing capacity, cyber
48:20security, and invest in resilient
48:22digital systems while ensuring that AI
48:25development remains environmentally
48:28sustainable. Finally, for demand
48:30generation, we must create an
48:33environment where AI innovation can
48:35thrive. This means integrating AI
48:38investments into government planning and
48:40budgeting, encouraging private sector
48:43participation, and mobilizing financing,
48:46procurement and incentive mechanism that
48:49will accelerate the development and
48:51adoption of AI solution across sectors
48:54and industry. To make this possible, we
48:57must invest in the foundation of AI
49:00adoption, particularly interested data
49:03systems, reliable digital
49:05infrastructure, research capabilities,
49:08and a workforce equipped to use AI
49:10responsibly. Stronger coordination
49:13across agencies will also allow
49:16successful innovation to be shared and
49:19scaled across government. This is how
49:22governance becomes an enabler of
49:24innovation by giving the agency the
49:27confidence to adopt AI while assuring
49:30citizens that its use remains
49:32transparent, secure and accountable.
49:35Through agile governance, capable
49:38institutions and sustained collaboration
49:40across sectors, we can build an AI ready
49:44government that is more effective, more
49:46responsive, and better equipped to serve
49:49every Filipino. Thank you and I wish
49:52everyone a productive and
49:55>> thank you DeepDev. We now turn to
49:58international perspectives on AI
50:00transformation [music]
50:01beginning with lessons from Thailand.
50:04Let us hear from director of strategy
50:06and organizational development division
50:09Thailand productivity institute
50:11Nantaporn Aantichart.
50:23>> [music]
50:33>> We develop like uh using AI to
50:35categoricalize
50:36and loot the leot from the people. So
50:40what kind of the report is like uh
50:42potholes, street light, garbage or some
50:47uh unsafety issue. So they they they
50:50categoricalize and directly to the
50:52municipal responsible office. So you
50:54know what uh they get they got like uh
50:571.4 million report and what what is the
51:01restoration rate is 77%. It's very high
51:05right. [music] So it's like a datically
51:08respond reducing response uh times and
51:12we can cutting emitative [music] head
51:15cost and also uh the more important
51:19things is they boosting like uh civic
51:22trust. We use AI for uh public health
51:26care and tele medicine optimization and
51:31we use AI to like assist radiological
51:34uh diagn diagnosis. So this is can
51:38reduce the wasting time would like to
51:40develop AI in your countries. You have
51:44to think about what the ultimate outcome
51:47if the AI can enhancing your quality of
51:51life uh make like a better well-being
51:55and also equalities
52:06uh for the public sector. I mean the
52:09data from the public sector is very
52:11sensitive [music] and very confidential.
52:13So for the next future I think for my
52:17country we try to develop large language
52:20model in [music] Thai version. So this
52:23is for you know collect all confidential
52:25data for the Thai people in inside my
52:28country.
52:36>> Thank you Thailand. Now we look to Korea
52:39for insights [music] from its own AI
52:41transformation journey. Let's hear from
52:43the APO director for the Republic of
52:45Korea and the chairman and chief
52:47executive officer of the Korea
52:49Productivity Center, Dr. Park Sunjun.
52:56[music]
53:08Okay.
53:35Yeah,ch.
53:38[music]
53:47[music]
54:20>> [music]
54:23>> Can you
54:59Thank you, Republic of Korea. Now, let's
55:02see AI in action, [music] shall we?
55:05Across government, teams are turning
55:07ideas into practical solutions. [music]
55:09And today, we get to see their work
55:12firsthand. Our AI in action government
55:15solutions segment showcases initiatives
55:18from three government agencies and one
55:20local government unit. [music] To kick
55:22things off with Tala or transparency,
55:26accountability and law in action, let us
55:29welcome the project manager of [music]
55:31ECMS and Tala from the anti- red tape
55:35authority, [music] engineer Jericho A.
55:37Labonete.
55:41[music]
55:42[bell]
55:45Blessed morning to everyone who among us
55:48wants a clear and very defined service
55:51delivery of government.
55:54>> Yes, we all want clear requirements,
55:56transparencies and someone to talk.
56:00Our Ta connects citizens concerned with
56:03complaints tracking and agency action.
56:05With our limited budget and complaints
56:08officer, we we maximize the use of a AI
56:12wherein Tala was made and born means
56:16transparency, accountability, law and
56:19action. In Filipino, Tala means star
56:22reflecting its role as a guide. It also
56:26means itala or to record. Every
56:30citizen's concern deserves documentation
56:32and follow through.
56:34Tala is yes sir is a star. Also Tala use
56:39conversational AI. We also tested it for
56:42the 19 Philippine languages or dialects.
56:46Next slide. It provides 247 digital
56:49assistant guides complaint filing and
56:51supports tracking through the complaint
56:53tracking number NC and CTN
56:56for
56:58we also has
57:01integrated it to EGO super app tal
57:04explain service requirement fees
57:07timelines using approved information it
57:10connects assistance summarize and
57:12internal support
57:14and for inquiry
57:16file track and guidance especially
57:19filing complaint
57:21with this uh citizens can reach Tala
57:25through EGOB's PH ECMS kiosk and also we
57:28integrated it to EGOB kiosk of DICCT for
57:32and also to all the participating agency
57:35widgets. The kiosk serves participating
57:38agency and service hub. the widget
57:41focusing in hosting agency or LGU. And
57:44this morning, allow me to share with you
57:47uh our live demonstration for Tala.
57:56>> The citizen can file complaint and also
57:58inquire for the citizen's charter with
58:01our ECMS tala.
58:15Hello, it's great to connect with you.
58:18How can I assist you today?
58:20>> Hi Tala. Uh, I'm filing for anonymous.
58:24My name is Paulo Raanit, age 32, gender
58:29male, and my contact number is 0912 3456
58:32789.
58:34Contact
58:47number 09123456789
58:58agency.
59:03I can't understand you. Can you speak in
59:05English?
59:08>> Sure, I can switch to English. Let's
59:10continue. Could you tell me the name of
59:12the person in the test agency that
59:14you're complaining about? If you don't
59:16know, it's fine to say unknown.
59:18>> Uh, I remember his name. His name is
59:21Jessie.
59:31Department
59:38unknown.
59:40>> BLO or business permit licensing office
59:55or others. Complaint tala.
1:00:02complaint.
1:00:13No tala
1:00:29actually
1:00:31last month
1:00:33and then about one month until now
1:00:38business permit and I'll try follow up
1:00:41since November October. number five. But
1:00:43until today,
1:00:56Edad 32. Gender male. Contact number
1:01:0009123456789.
1:01:04Agency: Test agency. Person complaint
1:01:07division unit BPLO Business Permit
1:01:10Licensing Office. Complaint type
1:01:12complaint reporting for someone else. No
1:01:16narrative business permit last month
1:01:22follow
1:01:25nature of complaint failure to render
1:01:28government services within the
1:01:29prescribed processing time.
1:01:50Ed 32 gender naki contact number
1:01:5409123456789
1:01:57agency test agency
1:02:01Jesse department bloicing
1:02:05office
1:02:07complaint
1:02:12payment last month
1:02:17follow
1:02:22their government services within the
1:02:24prescribed processing time.
1:02:44Record Nagan Pao Raanit Edu
1:02:48gender laki contact number 09123456789
1:02:54agency test agency department BPLO
1:02:58Business Payment Licensing Office
1:03:07narrative
1:03:17follow.
1:03:52Ombre numero 09123456789
1:03:59agency
1:04:04business permit licensing office
1:04:24Failure to render government services
1:04:26within the prescribed processing time.
1:04:33Correction.
1:04:42corruption. Once submit uh the
1:04:45complainant can scan the the QR code and
1:04:49basically uh the the complaints in the
1:04:52CPN will email to the complainant and
1:04:55also uh the complainant which al which
1:04:59also wants to inquire for the cit
1:05:01citizens charter can use also our one of
1:05:03our feature the inquiry for the service
1:05:06fee all the uh transactions
1:05:09of the government especially of that
1:05:11particular agency is also So welcome
1:05:13with our system. So again this is Tala
1:05:16from Anti- Red Tape Authority. God bless
1:05:18us all.
1:05:21>> Wow.
1:05:24Thank you engineer Labonet for
1:05:26introducing us to Tala and thank you to
1:05:28[music] the anti- red tape authority. I
1:05:31think that's incredible because how
1:05:32often do you see AI models large
1:05:35language models trained in Filipino or
1:05:38Philippine languages and dialects,
1:05:40right?
1:05:41and
1:05:45okay now we proceed to project Bella
1:05:50which stands for budget expert and
1:05:52learning assistant which is DBM's
1:05:55localized [music]
1:05:56version of chat GPT to show us how it
1:05:59works please welcome the director of the
1:06:02information and communications
1:06:03technology system service at the
1:06:06department of budget and ma management
1:06:09attorney Maria Anna Es Espinosa. [music]
1:06:26Good morning everyone and I'd like to
1:06:28thank the development academy of the
1:06:30Philippines for the kind invitation. I
1:06:32was just with DAP last week as a student
1:06:35for the AI501 uh govern uh AI governance
1:06:38for executives and now I'm back and I'm
1:06:40very happy to be here po. Good morning
1:06:44also to our partners both from the
1:06:46government and the private sector today.
1:06:49We are proud to introduce a significant
1:06:51leap forward in how the department of
1:06:54budget and management manages
1:06:57institutional knowledge, promotes
1:06:59learning and delivers better public
1:07:02service. Allow us to introduce Bella or
1:07:05the budget expert and learning
1:07:07assistant, an AI powered initiative
1:07:10designed to make budget related
1:07:12knowledge more accessible, convenient,
1:07:16and efficient for everyone. Shout out to
1:07:18my former bossa who actually launched
1:07:21this project before she transferred to
1:07:23the department of energy.
1:07:26Yes po.
1:07:29And so here's a road map for today. Yes.
1:07:33We'll start by defining the problem and
1:07:35why we built the Bella. Then we will
1:07:38define what Bella is and then we'll look
1:07:41at its architecture and how it solves
1:07:43fragmentation challenges. explore its
1:07:46active knowledge base with a quick live
1:07:48look and wrap up with our future
1:07:51direction.
1:07:56Okay.
1:07:58Today's discussion focuses on a
1:08:00significant challenge that is probably
1:08:03faced by a lot of government
1:08:05organizations. The fragmentation of
1:08:08information. Imagine a landscape where
1:08:12vital knowledge for example with DBM a
1:08:15lot of budget information across
1:08:18numerous systems share drives email and
1:08:22archives. This fragmentation leads to a
1:08:26lot of problems. Scattered knowledge
1:08:29outdated references and slows search
1:08:32processes which affects the efficiency
1:08:35of our services to the people. The real
1:08:38struggle is not the lack of information
1:08:41but the inability to preserve, connect
1:08:45and use in appropriate context
1:08:48efficiently across offices and personnel
1:08:50all our existing knowledge in those
1:08:53databases. When information is
1:08:56fragmented, trusted answers become
1:08:58elusive and valuable knowledge can be
1:09:01lost. This is one of the challenges
1:09:03we're addressing with Project Bella.
1:09:08So let's just dive into the core
1:09:10features of project Bella.
1:09:12First and foremost, we have the
1:09:14authoritative repository. So we don't
1:09:16just add any information in Bella. This
1:09:20is where all budget, legal,
1:09:22administrative, and official documents
1:09:25are consolidated. By centralizing these
1:09:28essential resources, Bella ensures that
1:09:31users have reliable and accurate
1:09:33information at their fingertips. Next,
1:09:37Bella harnesses a localized AI
1:09:40specially tailored to integrate DBM's
1:09:42unique knowledge assets. This means that
1:09:45the AI understands the context and
1:09:47intricacies specific to DBM, providing
1:09:51more relevant and effective assistance.
1:09:55Third, the platform offers validated
1:09:58access with secure and controlled
1:10:01access. Bella not only protects
1:10:03sensitive information of DBM but also
1:10:07makes it more accessible only to DBM
1:10:09employees when they need it. And lastly,
1:10:13of course, Bella is a powerful public
1:10:16financial management reform enabler. By
1:10:19facilitating digital transformation, it
1:10:22aligns with DBM's reform goals, turning
1:10:25scattered assets and knowledge into
1:10:27unified institutional intelligence.
1:10:33At its core, Bella is DBM's enterprise
1:10:37AI and knowledge management platform.
1:10:40Think of it as a centralized
1:10:41intelligence hub built specifically for
1:10:44the government, modern government
1:10:45workforce. It combines intelligent
1:10:48search and conversational AI to
1:10:50instantly surface policies, circulars,
1:10:53and budget issuances that used to take
1:10:55hours to track down.
1:11:00So that is why we created Bella. For the
1:11:03longest time, government data was
1:11:05siloed. Bella changed that with a single
1:11:09secure platform to store, access and
1:11:12archive documents. More importantly, it
1:11:15empowers our staff to query historical
1:11:17and scan documents using natural
1:11:19language directly advancing our digital
1:11:22transformation and go governance
1:11:25technology roadmap.
1:11:27So just a quick look uh on into the
1:11:30Bella architecture. Um first we have
1:11:32document pre-processing. This is where
1:11:34the system begins by organizing and
1:11:37preparing data for further analysis
1:11:40ensuring as I said that the information
1:11:42is uh correct, accurate and updated.
1:11:47Next we move into knowledge building.
1:11:50Bella constructs a comprehensive
1:11:51structured understanding of the data
1:11:54turning raw information into valuable
1:11:56insights. And of course there's the
1:11:58embedding and the large language model
1:12:00stage or LLM. This is the core where the
1:12:03data is transformed into a format that
1:12:06can easily be interpreted and used for
1:12:08decision making. Finally, of course, we
1:12:11have question qualification plus chat.
1:12:14So this is where users can engage with
1:12:16Bella in a conversational manner asking
1:12:19questions naturally and receiving
1:12:21accurate context aware responses. And so
1:12:24you also see the different tools that we
1:12:26use as uh as uh we try to uh implement
1:12:31Bella
1:12:38just uh this slide this slide just
1:12:40highlights the tailor design of the
1:12:42knowledge platform. So as I already
1:12:45mentioned this is authorized knowledge.
1:12:48This is very much curated with only what
1:12:50DBM needs. we don't ask information from
1:12:53outside or outside those that uh
1:12:56documents that we allow to be updated.
1:12:58So for example when we are actually user
1:13:00testing it one of the um user uh was
1:13:03asking how do you cook adobo and then of
1:13:06course Bella did not answer that because
1:13:08it's not part of the knowledge that was
1:13:10updated to her and the user was like why
1:13:13can we answer my question and we had to
1:13:15explain really that is specific
1:13:18specifically tailored for only what DBM
1:13:20needs. So there and next of course Bella
1:13:23contain budget data most of which could
1:13:27be uh uh personal and very confidential
1:13:31and so we have to take care of that.
1:13:33Secure governance ensures that only
1:13:36authorized personnel can access
1:13:38information appropriate for their role.
1:13:40The system not only protects data but
1:13:43ensures it is governed responsibly
1:13:46aligning with DBM's operational
1:13:47standards and requirements. And of
1:13:50course, n natural language retrieval.
1:13:53Through a simple question and answer
1:13:54interface, the users can ask the
1:13:57questions and retrieve answers
1:13:59immediately. And finally, the AI is
1:14:02specifically designed to align with
1:14:04DBM's ethics and operational goals,
1:14:07ensuring that the technology supports
1:14:09rather than complicates the work.
1:14:14So just we just want to feature here
1:14:17that the uh that uh a security feature
1:14:20of project Bella is that it is fully
1:14:22airgapped after it went live. This means
1:14:26that Bella services, knowledge
1:14:28repository, AI components and user
1:14:31interactions will only operate within
1:14:34controlled DBM environment. External
1:14:37connectivity will only be restored when
1:14:40necessary for an approved maintenance
1:14:42activity such as downloading patches,
1:14:45software upgrades or other required
1:14:47effects. These activities will take
1:14:50place with a defined maintenance window.
1:14:53Once the maintenance window is complete,
1:14:55internet connectivity will again be
1:14:57disconnected. Basically
1:15:00connect
1:15:02you can access Bella. The important
1:15:05point is that connectivity is not the
1:15:07normal operating condition for Bella.
1:15:09The normal condition for Bella is
1:15:12secure, isolated and fully airgapped
1:15:14environment after it went live.
1:15:20So there uh just showing you that we
1:15:22have two chat bots. So FAQ's chatbot for
1:15:25the uh frequently asked questions about
1:15:27the uh budget data and we have the
1:15:30budget chatbot, the math expert as I
1:15:32call it. So all the history in terms of
1:15:35budget data is there. You can ask how
1:15:37much was the NEP this year to this year
1:15:39and make a comparison as well as the guy
1:15:41and SA
1:15:46and I just want to highlight the people
1:15:49behind Bella. Um on this slide we can
1:15:51see key moments when the official Bella
1:15:54went live. The launch brought together
1:15:56DBM officials, technology partners, and
1:15:59project team members to formally
1:16:01introduce Bella as an AI powered
1:16:03knowledge management and decision
1:16:05support platform to support budget and
1:16:08public financial management. So really,
1:16:10thank you, Sakasa. And we continue that
1:16:12project right now with the DBM and all
1:16:15its uh public financial management
1:16:17reform. And so the next will be a short
1:16:19video so you can see better or have a
1:16:22better idea on how we actually use
1:16:23Bella.
1:16:26>> Introducing DBM Bella, the Department of
1:16:29Budget and Management's AI powered
1:16:31enterprise knowledge assistant. Bella
1:16:34automatically retrieves documents
1:16:35[music] from SharePoint, processes them,
1:16:37and transforms them into a centralized
1:16:39searchable knowledge base. Built on the
1:16:42Diifi platform, Bella [music] lets DBM
1:16:45users ask natural language questions
1:16:47English or Tagalog and get instant
1:16:49accurate responses with retrieval
1:16:50augmented generation. Each response
1:16:53includes document citations,
1:16:59ensuring accuracy, traceability and
1:17:02transparency.
1:17:04Secure access is powered by Microsoft
1:17:06Entra ID through single sign on and
1:17:09rolebased access. Every user sees only
1:17:12what they're authorized to view behind
1:17:13the scenes. Developers use [music] DIY's
1:17:16visual chat flow designer to create
1:17:18workflows, manage knowledge bases, and
1:17:20continuously refine the chatbot through
1:17:22feedback.
1:17:29All AI processing is done on premises
1:17:31through GPU stack DBM secure GPU cluster
1:17:35delivering [music] fast high performance
1:17:36AI without cloud dependency. DBM Bella
1:17:40transforming institutional knowledge
1:17:42into instant reliable insights.
1:17:48And this is just the beginning. [music]
1:17:50Bella will continue learning, improving,
1:17:53and evolving alongside DBM, [music]
1:17:56helping us build a smarter, faster, and
1:17:59more empowered way of working [music]
1:18:01together. Meet Bella, your AI assistant
1:18:04with a touch of personality.
1:18:17And with that, introduce to you Bella.
1:18:19Thank you so much.
1:18:23Thank you.
1:18:26Thank [music] you, Director Espinosa for
1:18:28introducing us to project Bella. I hope
1:18:30she also meets [music] Tala.
1:18:33Next is who want to know an AI chatbot
1:18:37for [music] science innovation and
1:18:39government support under project Propel.
1:18:41To demonstrate it, let us welcome
1:18:43assistant scientist [music] from DOSD's
1:18:46advanced science and technology
1:18:48institute, Mr. Elmer [music] Pamo.
1:18:56>> Good morning everyone. Uh first of all,
1:18:58I'd like to thank DAP or a de uh a
1:19:02development academy of the Philippines
1:19:05uh for the invitation and
1:19:08congratulations on this event uh
1:19:10especially for uh to the uh center for
1:19:14strategic futures. Um I I I also would
1:19:18like to thank DAP for the trust because
1:19:21as you may know uh DAP has been our
1:19:24partners or they're partnering with us
1:19:26in AI capacity building. So thank you
1:19:29for trusting us in uh uh in partnering
1:19:33with you doing this initiative.
1:19:36Now there's a question
1:19:38uh every agency in this room hears in
1:19:41some form every day.
1:19:44a collalapit who do I approach most of
1:19:48the time the answer already exists on
1:19:51government websites like SEC or DTI's
1:19:54website or do OSD's website now the
1:19:58problem isn't that we don't know it is
1:20:01that the knowing is scattered the
1:20:04problem uh I'm Elmer Pamo by the way uh
1:20:08of the department of science and
1:20:09technology and for today I'll be uh
1:20:14demonstrating ating want to know.
1:20:20This time I'm going to introduce you to
1:20:22Hana.
1:20:24Four kinds of people bring questions to
1:20:27government every day. uh a student
1:20:30asking our R&D priorities, an
1:20:34entrepreneur asking how to register a
1:20:38corporation or an LGU asking what
1:20:41technology exists for its farmers.
1:20:46The answers usually exist but they sit
1:20:48across SEC or in DTI, DOS and many other
1:20:52agency websites. So each person pieces
1:20:55the answer together.
1:20:57If you want AI ready government, the
1:21:00first foundation is not the model. It is
1:21:03knowledge people can actually find. So
1:21:07we built one front door. I'd like to
1:21:10introduce you to WANA.
1:21:13This one to know is a dod propel project
1:21:16developed by AI or the advanced science
1:21:19and technology institute
1:21:25for the main seat behind a single
1:21:27question box. Markets for student
1:21:31tracking which industries are growing.
1:21:33Innovation for researcher looking for
1:21:36technologies ready to commercialize.
1:21:40government programs for the entrepreneur
1:21:42who needs permits, funding for the
1:21:46startup looking for grants.
1:21:49Every question here is one uh
1:21:53one the portal suggests. It was built
1:21:56for all four of the people we just met,
1:21:59including the agencies in this room. You
1:22:02can use it and your programs are part of
1:22:04what it answers. Let me show you what
1:22:08for example an agricultural uh
1:22:10agriculturist would see.
1:22:17No account, no log in just ask
1:22:27LGU
1:22:29post harvest. So that's an example
1:22:32question like uh agriculturist would ask
1:22:36Notice the search query.
1:22:39She's asking a question in Taglish and
1:22:41the way she'd ask a colleague. And there
1:22:44it is. The answer is in Filipino, too.
1:22:48And this is what part uh this is the
1:22:50part that matters most.
1:22:53It doesn't just hand her a list. It
1:22:55tells her where to go next.
1:23:06Now let's uh try for example an
1:23:09entrepreneur. Okay. Now in plain English
1:23:13with a question that usually means a
1:23:15website then a phone call then a visit
1:23:18starts with a route
1:23:20file through SEC's own company
1:23:23registration system. And look there it
1:23:26ends. The SEC website the nearest SEC
1:23:30extension office. It doesn't to be uh it
1:23:34doesn't pretend to be the agency. It
1:23:36sends you to the agency. Two people, two
1:23:40languages, two agencies worth of rules,
1:23:44one question box.
1:23:49Now, here's the thing. What you just saw
1:23:53looks like a chatbot.
1:23:55And chat bots are easy to build these
1:23:57days. As you can see, uh, a lot of our
1:24:00examples here today are mostly, uh, chat
1:24:03bots. Okay. The model is often the easy
1:24:06part.
1:24:07The environment determine survival.
1:24:11What makes this work is what's
1:24:12underneath a knowledge base, okay, of
1:24:17SEC, DTI and DOS information
1:24:21in four domains with starter questions
1:24:23for each and it runs on DOS as this
1:24:27owned infrastructure. So that's the
1:24:30difference like instead of subscribing
1:24:33to for example cloud compute foreign
1:24:35cloud compute where your data will
1:24:38reside outside the country what we did
1:24:41is
1:24:43uh this knowledge bases doesn't leave
1:24:45the country okay so talk about AI
1:24:49sovereignity
1:24:50that's part most pilots never get to
1:24:52build here it already exist which means
1:24:57your clients already have a front
1:25:00What it needs now is you.
1:25:06So, let me leave you with three small
1:25:08asks. If you if I may try it, it's
1:25:13public and there's no log in. It's no
1:25:15longer just a P. You can visit it. I'm
1:25:18going to show you the QR code later and
1:25:21tell us where it falls short. We have
1:25:24already tested it in in terms of AI
1:25:26security
1:25:27and uh and tell us what you want people
1:25:32to find. We'll follow up on how to
1:25:35include it. We see partner agencies as
1:25:38enablers of national capability, not
1:25:41just funders of pilots.
1:25:44Remember that question.
1:25:48Let's make sure the answer is one
1:25:51question away.
1:25:55So let me show you the QR code.
1:26:05So this is the QR code or if you want
1:26:08you can just uh type on your browser or
1:26:11on your uh on your browser want to
1:26:13know.gov.ph.
1:26:17You can try it. Uh for in the interest
1:26:20of time, I will not demo it, but I'm
1:26:23just going to show you what it looks
1:26:25like. It looks something like this.
1:26:28Uh you can click this for the new chat
1:26:31and you can also have change the mode
1:26:33like there's a dark mode. As you can
1:26:36see,
1:26:39uh there are knowledge domains here and
1:26:43uh this is a project by DOSD.
1:26:48uh an example like for example the one
1:26:50that I asked a while ago uh this morning
1:26:53what are my options in investing skin
1:26:55care in the Philippines so this is uh
1:26:59what one uh uh suggested okay there's a
1:27:03market overview
1:27:05there are investment options key steps
1:27:09uh franchise partnership and so on let
1:27:12me just check another this one I I ask
1:27:16ano technologies ies now related to
1:27:18bamboo silipinas
1:27:20and then it will show you the different
1:27:24uh things uh it has because uh we have
1:27:29actually fine-tuned it on dod
1:27:32project information what what's good
1:27:35thing about want to know is this deep
1:27:38research thing okay for deep research
1:27:40it's going to take a while uh for you to
1:27:43to see the answer but the answer is very
1:27:46comprehensive received like this. I'm
1:27:48going to show you sample of deep
1:27:49research like something like this.
1:27:52Something like 40 pages long of uh
1:27:56answer. So if you're going to use deep
1:27:59research like this, yeah, you just ask a
1:28:01question about a certain technology and
1:28:04then click deep research and then it's
1:28:06going to do a research for you. So it's
1:28:08really agentic in nature. It's going to
1:28:11do the work for you.
1:28:14I think that's the end of my
1:28:16presentation. Thank you for listening
1:28:18and yeah po I hope we can you can
1:28:22partner with us on this initiative.
1:28:24Thank you.
1:28:27Thank you Mr. Pamo for introducing us to
1:28:29want to know. That's a very witty name.
1:28:31No. Do you want to know? I want to know.
1:28:35Now completing our government showcase
1:28:37[music] is Lifesaver PH, an AI assisted
1:28:40platform connecting community reports
1:28:42with [music] disaster response efforts.
1:28:45Presenting this initiative with the
1:28:46Quzon city government, let [music] us
1:28:48welcome Lifesaver PH co-convener, Mr.
1:28:52Voltater Tupas.
1:28:59Tupas Fight Media
1:29:03Coconer Lifesaver
1:29:06Carlos
1:29:08Manager Operation Center
1:29:11QCDR remote team shout out
1:29:21intelligence at AI disaster response
1:29:25city at hopefully in the future.
1:29:35So platform
1:29:38lifesaver
1:29:42community
1:29:44assisted at humanled
1:29:54climate action at disaster response.
1:29:58This is what lifesaver gives our
1:29:59communities a platform.
1:30:05The goal of the lifesaver app is to put
1:30:08these three things together. you
1:30:10information [music]
1:30:12um you government response and you
1:30:14community participation
1:30:17information
1:30:23at the end of the dayc
1:30:39Because after all, governance is really
1:30:42community and the government working
1:30:43together.
1:30:51I'm a journalist by profession.
1:31:07landing front page or prime time news
1:31:17Facebook
1:31:31partnership with ESPN
1:31:34who has been volunteering as a
1:31:36humanitarian volunteers
1:31:42build community of digital
1:31:44humanitarians.
1:31:52So
1:31:54publican
1:32:05updated oran gaps oran fake information
1:32:14systematiz
1:32:15um lenin
1:32:18etc.
1:32:21gaps response
1:32:30platform that first uh aggregates
1:32:35and verifies information
1:32:54platform
1:32:56realtime data or forecast
1:33:01or picture
1:33:04AI generated. So
1:33:10verify
1:33:20programs, DRRM programs like I report
1:33:24122 paral
1:33:33platform to organize information to
1:33:35support verification.
1:33:42City
1:33:43response effort
1:33:48department
1:33:52sign up report for now text and uh
1:33:56photos
1:33:57risk or hazards
1:34:00map view at feed viewer
1:34:04at amplify calls for help initiatives
1:34:08relieves relief operations
1:34:132025
1:34:17system respond in 15 minutes city
1:34:21government
1:34:30affected
1:34:35elderly
1:34:40action using the platform
1:34:44preposition goods relief goods
1:34:48augmentation
1:34:50platform to to call for help
1:34:57armed forces distribute relief goods
1:35:03features AI
1:35:06scraper system
1:35:11automatically scrape information
1:35:15we assume that it's verified
1:35:18but at the end of the day
1:35:26responder verified information
1:35:31accountability responsibility ility
1:35:35responders.
1:35:37More often than [music] not,
1:35:41fake information on the platform. You
1:35:44can also sign up as volunteers
1:35:48as a volunteer. You can undergo training
1:35:51uh for disaster preparedness, training
1:35:53for response, relief and recovery
1:35:56through a process designed to maximize
1:35:58your interest and skills
1:36:02caravan in the coming months wherein we
1:36:05will be visiting communities to empower
1:36:08um of course especially the young people
1:36:11respond um in times of disaster
1:36:16disaster resilience programs not just
1:36:19for the their families, their own
1:36:20families, but maybe also for their for
1:36:23their street, their barangi and their
1:36:25communities.
1:36:28The Filipino spirit is resilient,
1:36:31but too often it's our systems and
1:36:34infrastructure that crumble, but
1:36:37rebuilding them will take time, but we
1:36:39can start now. Let's turn our collective
1:36:41action into an infrastructure of trust
1:36:44and life-saving information. That's what
1:36:46we're building with lifesavers starting
1:36:48with K Queson City, but we hope to
1:36:51invite others to join us in this journey
1:36:54and build this together with other
1:36:56communities. Demoita
1:36:59exhibit
1:37:00QR code.
1:37:04Thank you, Mr. Tupaz, for introducing us
1:37:08to Lifesaver PH. I do hope you get to
1:37:10connect [music] with our LGU
1:37:12participants today and I also hope that
1:37:14you get volunteers for lifesaver PH.
1:37:18[music] Now it's your turn to ask the
1:37:21impletor any questions that you have for
1:37:24them will be entertaining one um
1:37:30u uh experience nila how how this could
1:37:34be implemented in your agency. Could
1:37:37there be opportunities to partner? Ask
1:37:39away.
1:37:41Do we have any questions from the crowd?
1:37:49Yes.
1:38:01Well, if there is none, uh, three of the
1:38:05four presenters have exhibit booths. So,
1:38:08you can always, uh, visit them, ask for
1:38:11flyers, ask for their websites, and get
1:38:14to know more about their solution. Now,
1:38:17we proceed to set two of our program.
1:38:26It's
1:38:39[music]
1:38:45[music]
1:38:51so I'm France de Leon, director of the
1:38:53DOSD advanced science and technology
1:38:56institute. So at DOSDS our approach is
1:38:59to help government agencies not just
1:39:02adopt artificial intelligence but
1:39:04develop the capabilities to understand,
1:39:06[music]
1:39:07evaluate and use it effectively and
1:39:09responsibly. We are doing this through
1:39:12several complimentary initiatives. Our
1:39:15computing and archiving research
1:39:16environment or core provides shared
1:39:19higherformance computing resources for
1:39:22research and development. Through our
1:39:25nexus for artificial intelligence
1:39:27research and applications or Naira, we
1:39:30are working [music] to bring together AI
1:39:32research development tools, reusable
1:39:35models and [music] expertise to support
1:39:37practical applications across
1:39:39governments and other sectors. We are
1:39:42also developing platforms such as
1:39:44dimemer [music]
1:39:45which supports the sharing and
1:39:47evaluation of AI models and itanong
1:39:52which explores how natural language
1:39:54interfaces can make government data and
1:39:57documents more accessible. So these
1:39:59initiatives [music]
1:40:00provide opportunities to develop, test
1:40:03and refine AI solutions with perspective
1:40:05[music] of the users. Now one important
1:40:09lesson is that developing a working AI
1:40:12prototype is very different from
1:40:15developing a reliable government
1:40:17service. [music] The challenge is not
1:40:19merely achieving good model performance.
1:40:22It is ensuring data quality, security,
1:40:27interoperability, accuracy,
1:40:29accountability and ultimately [music]
1:40:31the readiness of institutions and their
1:40:34people. We have also learned that AI
1:40:37solutions must be [music] designed
1:40:39around actual operational problems with
1:40:42domain experts and end [music] users
1:40:45involved from the beginning. Ultimately,
1:40:48successful AI is not measured by how
1:40:51many prototypes we develop, but by how
1:40:55effectively these technologies improve
1:40:57public services and deliver measurable
1:41:00benefits to [music] our citizens.
1:41:13I believe the next important step for
1:41:15government is to move from isolated AI
1:41:17experimentation toward a more
1:41:19coordinated, collaborative and
1:41:21sustainable AI innovation ecosystem.
1:41:24Today, many agencies are exploring AI
1:41:27independently.
1:41:28While these experiments are valuable, we
1:41:31risk duplicating efforts, fragmenting
1:41:34resources, and developing solutions that
1:41:37cannot easily be reused or scaled. What
1:41:40we need is a shared ecosystem where
1:41:43government agencies can experiment
1:41:45safely across common computing
1:41:47infrastructure,
1:41:49reuse data [music] sets and AI models
1:41:51where appropriate and evaluate
1:41:53technologies against consistent
1:41:55technical and responsible AI standards.
1:41:58So this is the direction we are pursuing
1:42:00through initi in initiatives such as
1:42:02Naira [music] and the broader
1:42:04development of the National AI Center
1:42:05for Research and Innovation or Nik as
1:42:09part of strengthening the country's AI
1:42:11research and innovation capabilities.
1:42:14Now equally important however is
1:42:16establishing a clear pathway from
1:42:19research and prototyping to validation,
1:42:23operational deployment and continuous
1:42:25improvement. This requires collaboration
1:42:28among technology developers, government
1:42:30agencies, policy makers, academic
1:42:33institutions, and the industry. AI
1:42:36systems must also be evaluated according
1:42:38to their intended applications and
1:42:40associated risks with appropriate
1:42:43safeguards for [music] privacy,
1:42:44security, fairness, transparency, and
1:42:47human oversight.
1:42:49Over the next three to five years, I
1:42:51hope to see the government agencies
1:42:53increasingly [music] become not just
1:42:56users or procurers of AI technologies
1:43:00but informed partners in their
1:43:02development, evaluation and responsible
1:43:06deployment. Our goal is to simply make
1:43:09sure that the government be more
1:43:12digital. [music]
1:43:13It should be to make government more
1:43:15capable, responsive, trustworthy, and
1:43:18genuinely better at serving the Filipino
1:43:21people.
1:43:26>> And that was director France of DOSDI.
1:43:30So, not only is there want to know,
1:43:32[music] there's Ethanong and there's
1:43:34Dimemer and there's Naira. So many
1:43:36initiatives that we hope would we we
1:43:38would see scaled soon.
1:43:44Now we turn to Japan for lessons [music]
1:43:46from its experience with AI adoption.
1:43:49Let us hear from the executive director
1:43:51of the Japan productivity center at
1:43:53Sushi Miyasaka.
1:43:58[music]
1:44:09Many of the big company implemented
1:44:10[music]
1:44:11AI as a organization wide but most of
1:44:15the case still [music] this AI is
1:44:19utilized for to improve the productivity
1:44:21at a personal [music]
1:44:23level.
1:44:24So uh the best practice company
1:44:29had successfully implemented [music]
1:44:32AI is very you know um AI to utilize the
1:44:37organization productivity improvement
1:44:41um uses two approaches one is like top
1:44:44down approach the other one is a bottom
1:44:47up [music] approach.
1:45:02But uh actually uh important thing is AI
1:45:05[music] is only a tool.
1:45:08So that the important thing is [music]
1:45:11how the company transform.
1:45:16So that top and bottom
1:45:19these are the ones we need a human
1:45:22because human can decide the strategy
1:45:26which direction the company goes or what
1:45:29business they have to develop. So this
1:45:31is a human and usually the middle
1:45:34portion of the task is many people
1:45:37assemble many of the information and
1:45:40make a proposal right. So this portion
1:45:43can be replaced by AI.
1:45:46So the com the [music] integrated human
1:45:49skills which is we call it teamwork to
1:45:53optimize a human skill to optimize AI
1:45:56usage [music] that's going to be the
1:45:58differentiator for the future company.
1:46:00Continuing
1:46:05our international perspectives, we look
1:46:07to Taiwan. Sharing their experience is
1:46:10the director of the management
1:46:12consulting [music] business group at the
1:46:14China Productivity Center, Ying Fu
1:46:17Chang.
1:46:23[music]
1:47:12Repeat.
1:47:22[music]
1:47:27[music]
1:47:30Speech. Speech.
1:47:41>> [music]
1:48:45>> Thank you, DOSDS, Japan and Taiwan.
1:48:49[music] Now, let's see more AI in
1:48:52action. This time, we turn to private
1:48:54sector innovations whose solutions offer
1:48:57possibilities for public service, from
1:49:00education to urban mobility and document
1:49:02management. [music]
1:49:04First up is Haya, an AI science tutor
1:49:07that [music] runs entirely offline. To
1:49:10show us how it works, let us welcome
1:49:13Haya's creator, Mr. Louise
1:49:16Buenavventura.
1:49:23>> Good morning. Good morning. Hopefully
1:49:25this microphone is working. Uh my name
1:49:27is Luis uh Buenovventura. I am uh not
1:49:30really from the public sector as as the
1:49:32host mentioned. No, I am technologist.
1:49:35I've been working here in the
1:49:36Philippines building uh innovative
1:49:38software solutions for about 25 years
1:49:40now. Uh and very recently I came upon
1:49:43this very interesting problem. Now this
1:49:46idea that uh the Philippine education
1:49:48has been really kind of falling behind
1:49:50the rest of the world uh in a way that
1:49:52is actually pretty dramatic. If you take
1:49:55the average 15-year-old public school
1:49:58Filipino child, you his math and science
1:50:01capabilities are roughly the same as a
1:50:0410year-old from Singapore. So, we're
1:50:07about five academic years behind in
1:50:09terms of the abilities of our children.
1:50:12And you know, I was listening to some of
1:50:13these other uh AI presentations earlier.
1:50:15I I really believe that uh with the
1:50:18power of AI, we are becoming almost
1:50:20super human. No, in the sense that you
1:50:23know my productivity, my my abilities
1:50:26feel like I am one uh 10 times uh more
1:50:29productive, 10 times uh a uh more
1:50:32capable uh with AI. But that creates a
1:50:36different problem because people who can
1:50:38afford AI, people who have access to AI,
1:50:41they're continuing to trend upwards
1:50:44faster and faster with their abilities.
1:50:46But the people who cannot afford AI, the
1:50:48people who don't have access get further
1:50:50and further behind. And with this Heriah
1:50:53project, my goal was to try to work from
1:50:55the ground up, no? Figure out a way to
1:50:58give AI even to our public school kids,
1:51:01those same kids that are almost five
1:51:04academic years behind our neighbors in
1:51:06Singapore. figure out how to start
1:51:09empowering them now, this early on with
1:51:13AI that does not require an online
1:51:16connection or an internet connection. I
1:51:18think that is kind of the biggest uh
1:51:20unlock here. No, I think that all of the
1:51:22solutions that you will see uh in
1:51:25general are going to be uh you know kind
1:51:28of solutions that require some kind of
1:51:31internet connectivity. And the the truth
1:51:33of the matter is here in the Philippines
1:51:35only about 30% of our households
1:51:37actually have fixed internet. So what do
1:51:40we do with the 60 plus% that can't
1:51:43actually get online uh to you know kind
1:51:46of help their kids take advantage of
1:51:48this stuff? So Haya is the response to
1:51:50that. What I've done here is I have
1:51:53taken the uh Department of Education's
1:51:55revised K to2 curriculum. I've uh
1:51:58trained a what we call a small model.
1:52:00No, it's kind of the opposite of a large
1:52:02language model. It's a small model. It's
1:52:04only about 1.7 billion parameters, which
1:52:06is tiny by uh AI standards. Your average
1:52:10anthropic model is about 100 times
1:52:13larger than that. Um, and I've
1:52:15compressed it in a way that fits on an
1:52:17entry-level Android phone. So, these are
1:52:19screenshots here, no, of uh what the
1:52:22what the UI looks like. So, I've made it
1:52:25uh very
1:52:26uh you know, kind of easy to understand.
1:52:29uh not not overwhelming. Um I've turned
1:52:32it into an interface that kids will
1:52:34probably uh recognize very easily
1:52:36because you're just scrolling uh new
1:52:38cards uh uh over and over again. You're
1:52:41seeing new information over and over
1:52:42again. Uh there's also uh testing within
1:52:46the app where you know after you've read
1:52:48a few of these science cards, you will
1:52:51get a quiz asking you just
1:52:53double-checking if you remember the
1:52:54things that you have read. And uh very
1:52:58importantly all of that information is
1:53:00collected by the AI again locally on
1:53:03device. None of this stuff requires
1:53:05internet and it will actually make
1:53:07decisions on what uh level of uh
1:53:11difficulty to continue to serve to the
1:53:14reader. So for example, if the kid is
1:53:17constantly making mistakes with the
1:53:19quizzes, the AI will silently regress
1:53:22the quality of content or the difficulty
1:53:24of content. I should say the difficulty
1:53:26of content so that the maybe it'll be a
1:53:29little bit easier, right? Uh and as the
1:53:33quiz scores improve over time, then the
1:53:36AI will kind of return that kid back to
1:53:40uh their their prescribed grade level.
1:53:43The the other thing I'm very proud of
1:53:45with uh with the Haya project is that it
1:53:48also uses uh uh homegrown homedeveloped
1:53:53uh English and Filipino voice models. Uh
1:53:56again, this is happening entirely on
1:53:58device. There's no need for an internet
1:54:00connection. Um if you're familiar with
1:54:02any of this stuff, uh voice models are
1:54:06actually quite large in reality. No. And
1:54:09uh I'm very proud to say that I was able
1:54:11to get it to under 100 megabytes for
1:54:13English and 100 megabytes for Filipino.
1:54:16Um the reason why I felt like that was a
1:54:19very critical uh requirement was because
1:54:22of course the you know your average
1:54:24Android phone does not have a lot of
1:54:26space and secondly it seems like our
1:54:29children are not just suffering from uh
1:54:31you know kind of limitations in math and
1:54:34science they also have a problem with
1:54:36literacy. Uh so in order for the science
1:54:40cards in Hera to actually be effective
1:54:43sometimes the only way to uh share that
1:54:45information with the child is to read it
1:54:47out loud, right? So it needs to be both
1:54:49in Filipino and English. I'm also
1:54:51working on a Sibuano uh voice model as
1:54:53well. Uh turns out that's a little bit
1:54:55more technically uh challenging to do
1:54:57because there's not enough training
1:54:59material uh that is public source and
1:55:01open domain and all of that stuff. But
1:55:03uh in general these are all possible
1:55:05with enough time and uh attention. Um so
1:55:09again just to kind of give you an idea
1:55:10of how small this application is the
1:55:12entire thing fits in 280 megabytes which
1:55:16again uh by most standards if you're
1:55:18talking about like Facebook or Tik Tok
1:55:21that's that's quite large no but
1:55:22remember those apps all of their content
1:55:25is coming from an an always on internet
1:55:27connection. Whereas here that 285
1:55:30megabytes represents the entire depth ed
1:55:33curriculum from grade 3 to grade 10 for
1:55:36science. Right? So you can you download
1:55:39it once and you are basically good to go
1:55:41forever. In fact if you downloaded it
1:55:43when you were grade three theoretically
1:55:45you could continue using that same file
1:55:48for the next seven years of your
1:55:49education. So I feel like it's a very
1:55:51meaningful uh improvement. Uh, one of
1:55:54the things I'm also quite proud of is
1:55:55that we are available in Android, Chrome
1:55:58OS, and Windows. Although those are the
1:56:01three most common places. Uh, those are
1:56:03the most commonly donated devices to our
1:56:05public schools. Uh, that was kind of a
1:56:07request from some of the public schools
1:56:09I've already uh piloted in. And I've got
1:56:11three active pilots as of uh last week.
1:56:14Uh, Kalapakuan in Zambales, Better World
1:56:18in Tondo, and then Danghari in Navatas.
1:56:21and I'm looking for more pilots uh to
1:56:24kind of roll these out. What we're doing
1:56:25is we're doing uh 35 to 40 uh students
1:56:30in a class and we actually donate the
1:56:32mobile devices that they can use because
1:56:34turns out not not a lot of these kids
1:56:37have access to their own mobile phones.
1:56:39So, we had to donate the devices
1:56:40themselves. Um yeah, and so what am I
1:56:44doing right now? I'm basically waiting
1:56:46for uh data from these different pilots
1:56:49to see what else we can improve. Uh
1:56:51certainly the most important thing for
1:56:53us is to make sure that the the content
1:56:55and the approach and the what the
1:56:58technical term is pedagogy is actually
1:57:01uh effective. Meaning uh is it actually
1:57:03helping kids? Are they scoring higher in
1:57:06uh official examinations? Those are the
1:57:08data points that we're waiting for right
1:57:10now. Um if you want to get involved
1:57:13somehow or you want to learn more, uh
1:57:16the website is heriah.org. That's h i a
1:57:20i a.org. Uh I'm the oneman team building
1:57:24the whole thing. Um you can help right
1:57:28now in either uh helping me equip my
1:57:31pilot classrooms and that usually
1:57:33involves procuring somewhere between 30
1:57:36and 40 mobile phones. We can work with
1:57:39very cheap mobile phones. The whole
1:57:41point was it needs to be accessible to
1:57:42everyone. So, the budget that I was uh
1:57:45targeting for every phone was less than
1:57:475,000 pesos in most cases. Um, you could
1:57:50also help by introducing me to pilot
1:57:53schools that are interested in
1:57:55participating. I think the schools that
1:57:57would have the most to gain from things
1:57:59like this are last mile schools, meaning
1:58:01schools that are not reachable by fiber
1:58:05internet, don't have uh, you know, uh,
1:58:08strong internet connection, all of those
1:58:09things. No, because we can ship the
1:58:11phones to them uh with everything
1:58:13already installed and they're good to go
1:58:15from that point. Um you could also
1:58:17support the work directly by helping out
1:58:20with uh training materials uh helping us
1:58:23improve the models and all of that
1:58:24stuff. Um and that's it. Uh hopefully
1:58:27you were found something illuminating in
1:58:29my talk. Uh thank you very much for the
1:58:31time.
1:58:32[applause]
1:58:35>> Thank you Mr. Brenaventura. Continuing
1:58:38with education, we have Front Learners
1:58:41featuring [music] Mastery Learning with
1:58:43AI feedback tutoring. Please welcome the
1:58:46chief executive officer of Front
1:58:48Learners Incorporated, Miss Maria Elaine
1:58:51P. Develis.
1:59:13Um, most education technology solutions
1:59:18start where the internet connectivity is
1:59:20the strongest. Teach aid started where
1:59:24the need is greatest. Good morning
1:59:26everyone. I'm Elaine Develis, the
1:59:28co-founder and CEO of Front Learners.
1:59:32We are an e-learning content and
1:59:34solutions provider for schools,
1:59:37universities, and institutions.
1:59:40Today I'm here to talk about Teach Aid.
1:59:44Teach Aid is a mastery learning and AI
1:59:48feedback tutoring system designed for
1:59:51last mile schools and geographically
1:59:54isolated
1:59:56disadvantaged area schools even without
1:59:59internet.
2:00:0111 years ago, a corporate social
2:00:04responsibility program took us deep into
2:00:07the mountains of Kapa Starlac.
2:00:10After two hours of riding uh 4x4
2:00:16crossing 14 rivers and rugged mountain
2:00:20trail um four teachers and 63 students
2:00:25in an ITO community welcomed us and that
2:00:28became teach aids first deployment. So
2:00:33this is the actual picture and these are
2:00:35the children.
2:00:37Um so we gave them um some tablets and
2:00:43e-learning content which runs even
2:00:45without internet. We told them oh this
2:00:48is how you use it and then these are the
2:00:51same materials that the children the
2:00:54schools in the cities are able to
2:00:56access. So,
2:01:00e-learning resources,
2:01:02you know, months after the teacher
2:01:05called us and they said,
2:01:18S Andrea an IT student
2:01:22math contest sa baba
2:01:26I'm sharing with you this story
2:01:30and this is Andrea
2:01:33you know this year the same teacher
2:01:36called me mommy
2:01:41120
2:01:44teachers and you know Andrea she is now
2:01:47a secondyear BS elementary education
2:01:50student
2:01:52at Tarlac State University.
2:01:59I'm sharing with you this story because
2:02:04about 2.5 million learners in around
2:02:079,000 giga schools
2:02:10face a profound learning challenge.
2:02:14The national achievement test of depth
2:02:16ed showed that among the grade 10
2:02:19assessed students in these schools
2:02:22almost none reached proficiency
2:02:25and among grade 12 students among
2:02:28indigenous people's schools none reached
2:02:31proficiency
2:02:33and story matters why because it changes
2:02:37the way we frame the problem these
2:02:40students are not behind because they
2:02:42lack ability
2:02:43They are behind because they lack access
2:02:47to quality learning materials. They lack
2:02:49individualized support and their
2:02:52learning progress is not visible to
2:02:54school leaders in the city and the
2:02:56central offices because there no real
2:02:58time monitoring
2:03:01and that is what teach aid provides.
2:03:05Teach aid is a portable offline online
2:03:09mastery learning system with built-in AI
2:03:13shooter purposebuilt for GIA schools
2:03:16even without internet
2:03:19through its built-in Wi-Fi. Teachers and
2:03:21students can access depth ed curriculum
2:03:25aligned lessons, exercises and quizzes,
2:03:29virtual laboratory and AI feedback
2:03:32tutoring.
2:03:34It is not just a device. It's a complete
2:03:37mastery learning system that helps close
2:03:40the learning gap.
2:03:43You know with teach aid mastery learning
2:03:47is possible. Why? Because students can
2:03:50go back to the lessons and repeat and
2:03:52repeat until they understand the lesson.
2:04:04And that is the problem most students
2:04:06face today. We proceed to the next
2:04:09lesson without full understanding of the
2:04:12previous lesson. and it compounds until
2:04:15it's too late to go back and correct
2:04:18their foundational skills.
2:04:21And then may AI feedback tutoring, but
2:04:25the way we use AI feedback tutoring is
2:04:28very researchbased.
2:04:30Um the feedback is only used for
2:04:35assessment. Only when a student answers
2:04:37a a test item will the AI provide
2:04:41feedback. So if the answer is correct,
2:04:45the the AI feedback ttor stills provide
2:04:47a feedback because
2:04:53so with the AI feedback, the the kid
2:04:57will understand better why
2:05:00and if the sagot is wrong,
2:05:04the AI will also feedback because
2:05:06perhaps the kid understand the lesson
2:05:10but it lacks the deepness of
2:05:13understanding to be able to apply the
2:05:16concept and so the feedback helps the
2:05:18kid have a deeper understanding of the
2:05:21lesson and the teacher said
2:05:25AI feedback and not only that the
2:05:29insight to the students answer guides
2:05:32the teacher on how to approach the topic
2:05:35so that they can be more effective in
2:05:37explaining to the student the concept
2:05:40step
2:05:42and the other beauty is if there is no
2:05:46answer the candidate answer there will
2:05:49be no AI feedback because what we're
2:05:53trying to avoid is over reliance or
2:05:55overdependence on AI which is very risky
2:05:59and it can even impede the learning of
2:06:02the student.
2:06:04So the AI also feedback is very guarded
2:06:09meaning only it is only fed the lesson
2:06:14the question the multiplechoice answers
2:06:18or the choices the students answer and
2:06:22the correct answer. So there's no way
2:06:24the AI will hallucinate. It is a very
2:06:28guarded feedback system. So that's how
2:06:31we use AI effectively and safely in
2:06:36education.
2:06:38Now the other gem of teach aid is the
2:06:41visible learning progress.
2:06:44Usually
2:06:45people from the central office or the
2:06:47supervisors, the school leaders have
2:06:50really no idea what's happening in the
2:06:53classroom with every student. But teach
2:06:56aid is able to record
2:06:59every activity in the classroom of the
2:07:03students. So even offline and then once
2:07:07an internet becomes available for
2:07:10example the teacher on the weekend goes
2:07:13home sa Bayan even for a 10 minute time
2:07:17interval with internet connectivity.
2:07:20Teach aid automatically synchronizes
2:07:24and feeds back the data to the central
2:07:26office up until the depth at central
2:07:29office in Manila. And so the leaders
2:07:33will have near realtime data. They will
2:07:37know how to intervene because there is
2:07:41data that they can base their
2:07:43intervention on.
2:07:47So teach aid is not just a device. It
2:07:50integrates the complete mastery learning
2:07:53cycle and this is why it helps close the
2:07:58learning gap.
2:08:00So at this point I'll provide a very
2:08:02short demo of a just a couple of
2:08:05features of teach aid. There are many
2:08:08features but for the interest of time I
2:08:10will demonstrate to you just some of
2:08:13them.
2:08:15So for example this is math. I know many
2:08:19of us either love or hate math and it's
2:08:24because normally the way math is taught
2:08:26is formulas, numbers without really
2:08:30allowing the students to imagine or to
2:08:33conceptualize what the topic is about.
2:08:36So this is about calculating the surface
2:08:38area of a cuboid or the surface area of
2:08:42a box. Imagine a carton box.
2:08:46If we unfold the box, you will notice me
2:08:49six faces.
2:08:52Okay. So, what you can do? Yeah, you can
2:08:56just tell the student, oh, this is just
2:08:57a box. The surface area is the sum of
2:09:00all these faces. There are there are
2:09:03four faces. So, you know, you can just
2:09:07simply imagine that is the summation of
2:09:09the six faces. That's the surface area
2:09:13of a cube or a cuboid.
2:09:17If you allow the kid to imagine it, then
2:09:21because of our technology, you can
2:09:23simulate. Oh, what if you increase the
2:09:26length of the box, the width of the box,
2:09:29the height of the box, and then this is
2:09:33just how it will look phenormous.
2:09:37It's so interesting.
2:09:39Okay. So you can have a situation we're
2:09:42in for example. Oh what if the length is
2:09:44four the width is three the height is
2:09:47wid or the six feet are this then 52 is
2:09:51the surface area
2:09:54that easy instead of going straight into
2:09:57the formula which is what typically how
2:10:00math is being taught to us. Okay. So
2:10:03that's set A. Again, you demonstrate
2:10:07again. Let the kid play and imagine and
2:10:11expand his imagination
2:10:13together with numbers. And then magiz.
2:10:20So that's how you uh use
2:10:25um technology in the classroom.
2:10:29Okay. and then you can have you know a
2:10:32quiz
2:10:34later on. So that's that's just a
2:10:38demonstration of what we do and there's
2:10:39AI feedback quiz t. Okay. So again just
2:10:45to go back to uh my presentation
2:10:51um
2:10:53actually
2:10:56teach aid has been proven by evidence in
2:10:59terms of of effectiveness. The
2:11:01department of science and technology ran
2:11:04a pre-post test uh among 38 schools
2:11:08across the seven provinces of region
2:11:10three and they really showed with teach
2:11:13aid the you know performance of kids
2:11:17really improve and this is among GIA
2:11:20schools.
2:11:22Another private school study together
2:11:24with the professors of the Asian
2:11:25Institute of Management again showed
2:11:29with evidence that statistically
2:11:32um an AI with feedback AI feedback
2:11:35tutoring significantly improves students
2:11:38performance versus traditional right or
2:11:41wrong only.
2:11:43And through our partnership with the
2:11:46department of education,
2:11:48they included teach aid as part of the
2:11:51pizza readiness program wherein we
2:11:54participated along with 91 countries
2:11:59um which was conducted by the OECD.
2:12:03So teach aid was used across um all the
2:12:06participating Philippine schools
2:12:08nationwide.
2:12:10If you were um aware of the news in
2:12:13September, OECD congratulated the
2:12:15Philippines because of all the 91
2:12:18countries, we got the highest
2:12:20improvement in terms of reading and even
2:12:24in our science and math, we were one of
2:12:27the most improved countries among the 81
2:12:31part 91 participating countries, which
2:12:34goes to show really our Philippine
2:12:38education sector can benefit benefit a
2:12:40lot with teach aid
2:12:43and actually we have deployed teach aid
2:12:46in more than 200 schools across region
2:12:49three through a partnership with the
2:12:51OSD. We also have private schools who
2:12:54are already using teach aid who are our
2:12:57partners for the past seven years and we
2:12:59have three um partnerships with the
2:13:02department of education as I mentioned
2:13:05for PISA. We also supported DEP ed on
2:13:08the first ever computer-based national
2:13:10achievement test and today we are
2:13:12piloting in 510 schools across 17
2:13:15regions nationwide with 1 million
2:13:18students and recently because of the
2:13:20super El Nino uh we are working with
2:13:23DEPAD also to make sure learning
2:13:26continuity happens even with the heat
2:13:29wave that will be happening in many
2:13:31parts of the country.
2:13:34So teach aid was built by Filipino
2:13:37scholars. I have one of my co-founders
2:13:40here, Mr. Leo Develis, whose uh
2:13:42dissertation is AI for education. One of
2:13:46our partners is an world expert on
2:13:49educational measurement. And we are here
2:13:53to help uplift the quality of B
2:13:57education in the Philippines. And I'm
2:14:00here today to ask everyone to join us in
2:14:04our education advocacy of teach aid
2:14:07parasida.
2:14:09Partner with us because this is truly a
2:14:12scalable, sustainable and inclusive
2:14:15investment for the long underserved
2:14:18learners of the Philippines. And if
2:14:20you'll ask me, is this only for GIA? No,
2:14:24we just began with Gida because Geda is
2:14:26the most difficult place to be. And if
2:14:29it works in GIA, we guarantee it will
2:14:32work anywhere, especially in the cities
2:14:35where connectivity is really the in in
2:14:38their best uh position right now.
2:14:42So go back to Andrea. Andrea has shown
2:14:46that when ability is matched with access
2:14:49and support, everything is possible. And
2:14:52with teach aid, every GIA learner can
2:14:54have access and support. And when every
2:14:57Gita learner rises, the whole
2:15:00Philippines rises.
2:15:05[applause]
2:15:12Develis and front learners and for all
2:15:14that you do for the children's
2:15:15education. Now from learning to urban
2:15:18mobility our next solution is Sermo. To
2:15:21demonstrate it let let us welcome the
2:15:24CEO and founder of mobility vision plus
2:15:27[music] engineer Cyrus Borha Gumari.
2:15:39All right.
2:15:42All right. So, we're ser uh I'm Cyrus.
2:15:45Um in summo is AI empowered transport
2:15:49and urban planning. Basically, uh we use
2:15:51AI where it's relevant AI driven AI
2:15:55automated completely. Uh because if you
2:15:58if anyone here knows anything about
2:16:00cities and how they're actually planned
2:16:02actually
2:16:09and in in our case since our target
2:16:12really is uh
2:16:15right
2:16:17urban planners
2:16:19planners but how how come none of or
2:16:21none or even very few of those plans
2:16:24actually get implemented? And that's
2:16:26what we're trying to solve. Uh big
2:16:28picture problem. The what we're trying
2:16:31to avoid is left side this is
2:16:34Commonwealth. Uh and unfortunately
2:16:39from the different provinces a lot of
2:16:42the cities are turning into this mess,
2:16:46right? And of course uh this is a
2:16:48consequence of many uh you know decades
2:16:52of bad practices and guidelines all put
2:16:55together without understanding
2:16:57objectives infrastructure development
2:17:00and on the right side um that's the
2:17:03richest neighborhood basically in the
2:17:06entire Philippines right yeah that's
2:17:08Makati CBD next to the Ayala triangle
2:17:13triangle area
2:17:16line which is uh people oriented uh
2:17:19people is first the left side it's
2:17:21always about car accessibility assuming
2:17:24everyone has a car but only less than
2:17:2610% of Filipinos have a car so why are
2:17:29we planning around that it makes
2:17:31absolutely no sense uh we spend billions
2:17:34yearly on infrastructure that just leads
2:17:37to chaos and we don't really re uh rip
2:17:40the benefits from uh the funds we
2:17:42actually have and you sassa from DBM
2:17:44with [laughter] would
2:17:45previously from there would know how
2:17:48these budgets were used in the past.
2:17:50Anyway, so you when talking about ano um
2:17:55end result recommendations, just want to
2:17:58show that first before highlighting what
2:18:00we actually do. Uh one example to s Naga
2:18:03city when we start uh start started
2:18:05working with Mayor Lenny first last
2:18:07year. One of our early recommendations
2:18:09was street
2:18:13abandon building because foot traffic
2:18:16was being cut off plaza going down uh
2:18:20because they were allowing cars although
2:18:22very few
2:18:25it was still open. Um
2:18:28we through our system we were able to
2:18:30count that about 20,000 pedestrians walk
2:18:33into this plaza every day.
2:18:36my part real estate which is right next
2:18:39to the plaza which is basically you know
2:18:43uh useless right and this is very useful
2:18:46uh real estate that could be used to uh
2:18:49not just energize the community but uh
2:18:51reinvigorate the local economy within
2:18:54this area and it doesn't really cost
2:18:56that much and we're super happy now two
2:18:58three weeks ago two weeks ago when I was
2:18:59there
2:19:02fully pedestrianized and with without uh
2:19:05any any much traffic impacts. It another
2:19:07example
2:19:09pass uh we're propos hopefully next year
2:19:12construct we're proposing chaotic
2:19:16intersection
2:19:18traffic when they go during peak hour to
2:19:21convert it into something on the right
2:19:23side which is more uh it's good for
2:19:26traffic but but more importantly it's
2:19:28bet it's good for the businesses it's
2:19:30good for the pedestrians uh good for the
2:19:32customers going around
2:19:35traffic
2:19:37choke point which causes even more
2:19:40congestion uh down the road. The
2:19:44not any they're not u these designs are
2:19:48nothing innovative. No. Uh what's
2:19:50innovative here is are we able to
2:19:52justify this?
2:19:55We have a lot of great architects. I'm
2:19:57not an architect myself. I'm I'm an
2:19:59engineer. uh of course with the help of
2:20:02a AI I can render the things I want to
2:20:04to see now much better
2:20:08is not uh the how nice it looks like but
2:20:10our problem is how do we justify it and
2:20:13with the help of AI and all of the
2:20:15expertise I have in urban planning I can
2:20:18we can uh speed this entire thing up
2:20:21with the objective that we focus on
2:20:23what's implementable not another plan
2:20:32mayor within the next week, within the
2:20:34next two months, within the next year,
2:20:36right? Uh, put it short.
2:20:39The reason we can speed it up is um it
2:20:42just falls down into three categories.
2:20:44How fast we were able to capture that
2:20:46data on the ground.
2:20:51tricycle
2:20:53intersection.
2:20:54How fast can we analyze that
2:20:56scientifically?
2:21:03Which is a common phrase. This is not a
2:21:06joke. This even
2:21:12experience blah blah blah.
2:21:15But again, if it's not based on
2:21:17evidence, based on data, what's the
2:21:20point?
2:21:22Right. And the more most importantly for
2:21:25uh our decision makers in the LGUs
2:21:27especially even in the national
2:21:28government the end goal is decision
2:21:31recommendation. Right.
2:21:33So the with that we can get to decisions
2:21:35much faster at le up to even 90 times
2:21:38faster depending on
2:21:42uh the system is quite straightforward.
2:21:45Um our basically our mobile app is used
2:21:48to capture anything on the ground. Uh
2:21:50the CCTV cameras are used to uh capture
2:21:53the traffic uh conditions and the
2:21:56different modes even tricycles, ebikes
2:21:58and all of those inhouse
2:22:02video AI. So for every LGU we work with
2:22:07AI context within those. But how do we
2:22:10make sure it actually works for any LGU?
2:22:14It's just understanding the local
2:22:15context. Uh it doesn't matter if we have
2:22:18the nicest technology out there.
2:22:34no matter if our technology is nice or
2:22:36not, right? Um, and my and here I'll
2:22:39just quickly show uh it's a one minute
2:22:42video basically showing what uh our
2:22:46system look like looks like right now,
2:22:47but you can go to our booth later to
2:22:50actually see it. So, I'm going to play
2:22:51this for you guys.
2:22:58Video
2:23:03play. Can you press it
2:23:14or can we play it or I'll skip it?
2:23:19H okay. Anyway, you can look look at it
2:23:22at the booth. Uh
2:23:26anyway, uh the all of this uh when we
2:23:29started it um it was I came I came back
2:23:32from Germany uh four years ago now uh as
2:23:35a DOSC balik scientist now with Pasig
2:23:38and Tagig and with Pasig. Pasig has been
2:23:41Pasig was actually our first client and
2:23:44they're we're already working with them
2:23:45for three years now.
2:23:49Mayor Vgo, which is very much a
2:23:52validation stamp for us. Mayor Vgo is
2:23:55one of the hardest mayors to actually
2:23:57get approval for from. Uh I think if
2:24:01other companies here have tried they
2:24:03know the the hurdle but uh
2:24:08it was just a a very basic mobile app uh
2:24:11in 2022
2:24:12uh and just uh the uh typical video AI
2:24:16we could create. We actually started it
2:24:18all there before uh growing into uh what
2:24:22we are now. Um we've we're we already
2:24:25have about 10 different edge use we're
2:24:28working with all over the Philippines
2:24:29ibatang context even Isabella city
2:24:32basilan uh it doesn't really matter
2:24:35context city our system is able to
2:24:37capture it because it's not just the
2:24:39technology it's also us as planners
2:24:42understanding new needs decision makers
2:24:47but more importantly understanding
2:24:50having having empathy about um
2:24:54on the ground, right? And of course,
2:24:56none of this would be pos would be
2:24:58possible without my great team. Um,
2:25:03majority are young professionals. And
2:25:05that's one of the other sides of AI
2:25:09people don't see. Um there are news
2:25:17but if you if you're um if you have so
2:25:20much uh knowledge in a certain field uh
2:25:22you can also pass it down much faster to
2:25:24the young professionals who are all also
2:25:27more AI savvy to and uh I think this
2:25:30this approach so far is working with the
2:25:32LGUs that we're working uh with. So if
2:25:35there are any edges or other um entities
2:25:38who want to work with us, uh this is my
2:25:40contact and we're just right here. Thank
2:25:42you very much.
2:25:47>> Thank you, Mr. Gumari. We'll play their
2:25:49videos on their uh monitor here in their
2:25:52booths. We're so sorry for that.
2:25:55Completing this set is Tiwala E notary,
2:25:58document management for the AI era.
2:26:01Please welcome Tiwala's program manager,
2:26:03Mr. Aljun Kinetic
2:26:10test. I just hope that you guys can hear
2:26:13me. And um yeah, good morning to the
2:26:16members of the board and to our partners
2:26:18from uh government and private sectors
2:26:22and to the organizers of this event. Uh
2:26:25DAP, thank you for inviting us. Um,
2:26:29TWAL started its humble beginnings back
2:26:32in 2019. So, driven by a single vision
2:26:35to create value by digitizing document
2:26:38workflows which includes legal documents
2:26:41like notoriization.
2:26:43So, over the years uh we build our
2:26:45reputation on trust on delivering real
2:26:48value
2:26:51to our users through a platform that
2:26:53handles their end-to-end document needs.
2:26:57So, Tuala worked um side by side with
2:27:00the Supreme Court uh authoring the 2005
2:27:042025 rules on electronic notoriization,
2:27:07which is a milestone that brings our
2:27:09country closer to easier and more equal
2:27:12access to document notoriization for
2:27:14thousands, if not millions of Filipinos.
2:27:18Over the years, Tala has been recognized
2:27:20by multiple organizations including ARTA
2:27:23which is also uh participant at this
2:27:25event. So from our friends uh from ARTA
2:27:29uh actually uh when we launched the e
2:27:31notary uh last October 19 um uh we are
2:27:35lucky enough for uh to be graced by the
2:27:39presence of sec Ernesto Perez.
2:27:42And uh we are also uh recognized by edp
2:27:45space kubo uh la French tech and square
2:27:48startup trailblazers.
2:27:51Tala is also trusted by leaders and
2:27:53institutions across the country. We are
2:27:55accredited by the Supreme Court and just
2:27:57recently officially launched the
2:27:59legalization of legalization of
2:28:01authorization last October 19. So it's
2:28:04actually a very historical event. uh
2:28:07we've notoriized the very first document
2:28:09in in the in the country. So it's uh
2:28:12it's actually very historical and I
2:28:14probably shed a tear or two during that
2:28:15time cuz it's really um uh uh it's it's
2:28:21really driving the Philippines one step
2:28:23uh towards digit digit digitiz
2:28:26digitizing sorry the the the the
2:28:29document workflows.
2:28:34We are an official partner of the
2:28:35integrated bar of the Philippines and
2:28:37the department of science and technology
2:28:38as well. Uh we are also an official
2:28:40member of the cloud signature consortium
2:28:42or the uh and the Philippine chambers uh
2:28:46of commerce and industry or the PCCI.
2:28:49So the problem we're actually solving
2:28:51solving with our platform is the
2:28:53paperbased processes. So, it's
2:28:55inefficient, costly, and highly prone to
2:28:58risk because anything that is printed on
2:29:01a paper can be easily tampered or or
2:29:03altered.
2:29:05And TWA simplifies this by unifying the
2:29:08entire document journey into a single uh
2:29:11intelligent workflow. So, our platform
2:29:13is an end-to-end uh digital signature
2:29:15workflow. Uh then later on, if you want
2:29:18that document to get notorized, you can
2:29:20also do that within our platform. So, in
2:29:22in the interest of time, uh we're going
2:29:24to play a very quick demo on how our
2:29:26platform works. And if you have any
2:29:28questions or any follow-up concerns or
2:29:31if you just want to explore Tala, our
2:29:33boat is located uh there on the the left
2:29:35uh right side of the uh venue and you
2:29:38can approach our staff. So,
2:29:45I just hope the the video is ready. This
2:29:47is TW, where paperwork finally stops
2:29:50being paper. Draft, sign, notoriize,
2:29:54seal, all in one place. And this is Lex,
2:29:59TWA's AI.
2:30:03Lex reads the fine print so you don't
2:30:06have to. Let's ask it. How much notice
2:30:09does a tenant need to Okay, it's already
2:30:12answering. 60 days with receipts.
2:30:18Not in your files. Lex also searches
2:30:20Philippine law. Which law makes e
2:30:23signatures legal?
2:30:29Republic Act 8792.
2:30:33Lex, stop showing off.
2:30:37Need a contract? Pick a template. Tell
2:30:39Lex what to fill in. Then review the
2:30:42edits before you. I'll let Lex skip
2:30:45ahead. Fine.
2:30:48In the viewer, you can highlight,
2:30:51strike, scribble, and stamp. Lex, those
2:30:56are my notes. Good thing Twila keeps
2:30:59every version side by side in sync.
2:31:03Stack of leases. TWA pulls the fields
2:31:06into clean data, CSV or JSON, with a
2:31:10confidence score on each.
2:31:13Yes. Lex, spreadsheets are exciting.
2:31:17Send it for signature, set the order,
2:31:20and TwoA chases the signers for you.
2:31:23Lex, you don't have to celebrate.
2:31:25Everyone
2:31:28need it notorized. First, Twila checks
2:31:32every participant against the national
2:31:34ID registry through Phils E-Verify.
2:31:39Then, notoriize online. Tala is a
2:31:42Supreme Court accredited e- notary
2:31:44facility. Even Lex behaves in here.
2:31:50Then Hash Origin seals the document.
2:31:53Thank you. Lex hashes it with SHA256
2:31:58and anchors it on the blockchain.
2:32:02So anyone can verify a Twila document
2:32:05with or without TWA.
2:32:09Verified. Lex, no victory lap.
2:32:15Plus the everyday stuff. Find anything
2:32:18in two keystrokes
2:32:20and keep work moving on boards.
2:32:25Two, document management for the AI era.
2:32:29Lex, say bye.
2:32:32>> All right. So, ladies and gentlemen,
2:32:34this is Tala, your end-to-end document
2:32:37workflow platform. Um so the the vision
2:32:40is already realized right so the
2:32:42solution and the implementation is
2:32:44already live. Uh if you are someone who
2:32:46practices uh legal you can try out try
2:32:49out our platform and apply as an ENP or
2:32:51if you are someone who just want to have
2:32:53your document notorized you can also uh
2:32:55register to our platform and set an
2:32:57appointment to any commission ENPs under
2:32:59our platform. Um so the challenge now
2:33:02the the solution is here the
2:33:04implementation has been realized the
2:33:05challenge now is to spread awareness for
2:33:07the Filipino people to start using this
2:33:09solution. So um yeah uh if you are
2:33:14interested go to our booth uh ask
2:33:16questions to our stuff and kindly follow
2:33:18our social pages as well and to to to
2:33:21like keep yourself posted on any news
2:33:23and any updates regarding uh the
2:33:25movement of euthorization within our
2:33:27country. So, thank you so much and have
2:33:29a great day.
2:33:33>> Thank you, Touala. Thank you, sir Mr.
2:33:36[music] Kenit. And thank you to Lex.
2:33:39Now, it's time to
2:33:42um ask the implementers if you have any
2:33:45questions
2:33:47um for them.
2:33:51Um
2:34:00L
2:34:03do we have questions [music] for our
2:34:05four uh implementers for uh companies.
2:34:16We have a [music] question from our EVP
2:34:19Mendoza.
2:34:20>> Okay. Thank you very much. I think we
2:34:22have excellent presentations. I have
2:34:25just a simple question uh particularly
2:34:29for the Haya presentation and the front
2:34:32learners. I think um these two
2:34:36applications are really very useful to
2:34:39the education sector. So maybe you can
2:34:43give your uh perspective or insights.
2:34:47How do we scale up to ensure that the
2:34:50entire
2:34:52public education sector at least at the
2:34:54basic education level could benefit from
2:34:58the the systems that are already there?
2:35:02No. Cuz if you just uh par pilot or
2:35:06select certain areas, we are unable to
2:35:10know maximize the benefits from these
2:35:12two systems.
2:35:13>> Yeah. So what what do you think uh
2:35:17should the government or maybe the
2:35:20community do in order to roll this out?
2:35:23>> Thank you so much for the question. Uh I
2:35:25think that's the kind of the perfect uh
2:35:27sorry I'll just stand here so more
2:35:29people can see me. Uh I think that's
2:35:30kind of the perfect uh table setting for
2:35:33uh what we want to do next. I think the
2:35:35pilots are important because you have to
2:35:37establish effectivity first. uh but
2:35:39certainly once we have established that
2:35:42uh there's there's another school of
2:35:44thought here now I think I've kind of
2:35:45changed my mind on very recently I I'm a
2:35:48software developer by trade so what I
2:35:51mean by that is it's literally my job
2:35:53and my vocation to build software that
2:35:55people pay for but I think that what's
2:35:57happened with the AI revolution is that
2:35:59I don't think people should be paying
2:36:01for software anymore. I think that we
2:36:03are at the point now where software
2:36:04should effectively be free. Uh because
2:36:08you know you can literally build
2:36:10anything that you want now because AI
2:36:13makes it possible. I mean of course you
2:36:15do need to have some engineering knowhow
2:36:16but the idea that you're still paying
2:36:18software licenses for things I think
2:36:21that's going to be an outdated concept
2:36:22within the next few years. So
2:36:27is completely free. If you, in fact, if
2:36:29you go to the website right now, you can
2:36:31just download it. If you want, you can
2:36:33re reverse engineer it and make your own
2:36:35copy of it. It's open source. Uh, you
2:36:38can dismantle it and study how I did it.
2:36:40In fact, there's also a white paper
2:36:42there that explains exactly the recipe
2:36:44for how I put it together. So, if you
2:36:46want to make a version of it, you're
2:36:48welcome to. If you want to use my
2:36:50version of it, the official version,
2:36:51you're also welcome to do that. So in
2:36:53terms of answering your question about
2:36:55scaling, uh the software itself is no
2:36:58longer going to be the bottleneck
2:37:00because the software itself is already
2:37:02free. So the bottleneck now is how do
2:37:04you get it into the households of the
2:37:07people that are that need it the most?
2:37:08No. So I'm talking about the 65% of the
2:37:11households that don't have internet
2:37:13connections, which is I think the still
2:37:14the largest problem. uh if we can wire
2:37:17up our schools, give them at least basic
2:37:19Wi-Fi, the the students can download
2:37:22Haya when they're in school uh and then
2:37:26then they'll have that access. Um
2:37:28however, there is still that problem of
2:37:29what if they don't have a mobile phone?
2:37:31No. So, I guess that's the next
2:37:32question. Um you long-term vision for
2:37:36Haya is still to have a custom device
2:37:40because I don't think we should be using
2:37:42uh off-the-shelf Android phone for this.
2:37:45it's slightly inefficient to do that
2:37:46because there's a lot of stuff in
2:37:48Android phones you don't need for for
2:37:50child uh for the basic education it's
2:37:53not necessary um that's the long-term
2:37:55vision I think we need to actually
2:37:57manufacture a device that's specifically
2:37:59for this but in terms of scaling it in
2:38:02the short to medium term again the
2:38:05software is free I can't make it any
2:38:07cheaper than that bottom uh what I'm
2:38:11looking for now is I'm looking for
2:38:12donors that can help uh finance hands
2:38:15these this first generation of mobile
2:38:18phones. It's not going to be specific to
2:38:20Haya. It can be any mobile phone. Any
2:38:23cheap phone will work. Um and you I
2:38:27guess the the advantage there is uh
2:38:30because it is so uh flexible. You can
2:38:33run it on any hardware. uh scaling is a
2:38:36little bit easier
2:38:42household will generally work.
2:38:45The only thing I don't support is Apple
2:38:47devices
2:38:51Android, Windows, Chromebook.
2:38:54So not the perfect answer to the scaling
2:38:57question, but I can at least address the
2:38:59fact that there's no cost to using the
2:39:02software itself. Uh po. Thank you so
2:39:04much. Thank you.
2:39:07Good afternoon. Good afternoon po
2:39:10develers
2:39:14learning materials which is aligned with
2:39:16the deped curriculum from grade 1 to
2:39:19senior high math science English or
2:39:21Olymp.
2:39:27We have used the older version during
2:39:29the pandemic which was very useful. Um
2:39:33now we have the new curriculum and the
2:39:36new trimester. So it's already loaded
2:39:38there. The only uh remaining um activity
2:39:42is to have a proof that they are really
2:39:45effective. So we have early result
2:39:48effective but uh we are now doing the um
2:39:52pilot test on 500,000 students in 510
2:39:56school but it will be modified to 1,500
2:40:00schools to include uh those that will be
2:40:02affected by El Nino. So because we
2:40:05cannot uh deped is a very big
2:40:08organization. We cannot just do
2:40:10something that is new. We cannot use
2:40:13something that is new. We have to prove
2:40:14it uh properly so that it's uh auditable
2:40:19and once it's know proven talaga early
2:40:22results is okay but we need a solid uh
2:40:26uh solid proof then it can be know it
2:40:28can be uh used by all schools because
2:40:32it's ready. That's first step. The
2:40:34second step is the the the tools and the
2:40:38devices that will be used to access
2:40:40those materials because it's already
2:40:43available. So we made it available in
2:40:45four uh actually five modalities. One is
2:40:48for face toface, one is for online, the
2:40:52other uh the third one is online, the
2:40:55fourth one is downloadable and the other
2:40:58one is printable. So we cover all the
2:41:01modalities because not
2:41:03um the students and the schools don't
2:41:06have all the needed devices
2:41:09as of this time but hopefully in the
2:41:11future we will all have tablets we will
2:41:14uh all have laptop so we're ready we're
2:41:17just making sure that it's being done
2:41:19properly so that it's it will be um uh
2:41:24justifiable to to funding by by the
2:41:28government so that's all thank you po
2:41:30>> yeah ma'am if I may uh the pilot that we
2:41:34are that's ongoing now across the 17
2:41:37regions and hopefully the NI also um
2:41:41it's not just a technology pilot. It's
2:41:44not about it's not just about
2:41:46integrating digitalization or AI into
2:41:50the public school system. Our pilot is a
2:41:53whole of system approach meaning
2:41:56pedagogy
2:41:58like as I mentioned from lessons a uh
2:42:02the feedback tutoring the assessment up
2:42:04until monitoring because that's how we
2:42:08can really make it a structured and
2:42:11sustainable solution that regardless who
2:42:14the personalities are in the system the
2:42:19process stays. So that is how we're
2:42:22making the pilot really something that
2:42:26is a true solution
2:42:28and that's why we started with the last
2:42:30mile because we don't want to be you
2:42:33know I would say yeah it works but what
2:42:36if there's no internet what if there's
2:42:38no solar what if the teachers change so
2:42:42we're making it really a whole of system
2:42:44pilot so that uh whatever the outcome is
2:42:48something we can really implement met
2:42:50and will be the lasting solution to the
2:42:53learning crisis that we are faced now.
2:42:56So I hope that is a uh an indirect but a
2:43:01satisfying answer to your question on
2:43:05scalability. Thank you so much ma'am for
2:43:07that really insightful question.
2:43:10[laughter]
2:43:11>> Thank you. Thank
2:43:12>> thank you Mr. and Miss uh Develis and
2:43:14thank you Mr. Luis Benaventura. I guess
2:43:17that's all the questions that we can
2:43:18accommodate for uh the set two. Now we
2:43:21proceed to our final set. [music] Set
2:43:23set three building transformation from
2:43:26within. We turn to India for lessons on
2:43:30building capabilities for AI
2:43:31transformation. Let us hear from the
2:43:34director general of the India National
2:43:36Productivity Council, [music] Miss Nidra
2:43:38Sakar. Let us watch.
2:43:43[music]
2:43:47>> [music]
2:43:56>> So which means we have been uh uh we
2:44:00have developed uh um translation models
2:44:04in a big way through uh AI and we are
2:44:07able to uh translate into a number of
2:44:10languages. We have more than 20 official
2:44:13languages. So that makes very easy for
2:44:15the government to immediate immediately
2:44:18disseminate any information uh to the
2:44:20people of the country through various
2:44:21languages [music] and we are able to
2:44:24translate
2:44:25um uh our educational content across
2:44:29languages. And then for uh education
2:44:32creating content smart [music]
2:44:34agriculture is being adopted uh uh
2:44:37weather predictions or uh uh pest
2:44:41management. So we've been able to merge
2:44:45the data and talk about how we can
2:44:47[music]
2:44:47uh fasten the pace of our logistics and
2:44:50clear the bottlenecks wherever they are.
2:45:01Yes, prepare preparing people for
2:45:05technologies of age is as important as
2:45:07adopting those technologies. So uh at
2:45:10the same time we are also talking about
2:45:13uh [music] ethics, privacy and uh uh
2:45:18respect to individual data and where the
2:45:20data fuition the people who collect the
2:45:22data, the people who use the data or
2:45:25people who aggregate the data have been
2:45:27given immense responsibility
2:45:30uh under the law any individualized
2:45:32data. So uh these are some of the
2:45:35frameworks that have been laid down by
2:45:38the government because in coming years
2:45:40technology is going to change a lot of
2:45:42things, improve a lot of things and also
2:45:46create a lot of jobs uh for the people.
2:45:49So education and skills and have to move
2:45:52in such a way that [music]
2:45:54which job whichever set of jobs are
2:45:57being taken away away by technology uh
2:46:00will also at the same time end up
2:46:02creating another set of jobs for the
2:46:05people. So that so our people should be
2:46:07ready for those jobs uh when the time
2:46:10comes.
2:46:17Thank you India. And that's the last of
2:46:19the international perspectives videos
2:46:21that we have for you. Now it's time to
2:46:24see what our own colleagues have been
2:46:26building. [music] The three projects
2:46:28grew out of DAP's own AI training of
2:46:31trainers turning learning into practical
2:46:34responses to everyday workplace needs.
2:46:36Of course,
2:46:38DAP. First is [music] to class an AI
2:46:42assisted research discovery tool for
2:46:44DAP's knowledge resources. Please
2:46:47welcome the project lead from the center
2:46:49of excellence on public sector
2:46:51productivity, Mr. Gerald Matthew E.
2:46:54Yustakio.
2:47:07Hello. Good morning everyone. I'm Matt
2:47:09and um together with my teammates Sam
2:47:12also from the center uh center for
2:47:14governance an Erica from the center of
2:47:17CS development and we're here to share
2:47:21with you something we've been working on
2:47:23at DAP and here's a short story of how
2:47:28we developed this idea. So we first
2:47:30thought of a learner in the academy and
2:47:34this learner has just come up with a
2:47:36brilliant idea on what to do for a
2:47:39capstone project and of course uh uh
2:47:44this person is excited and ready to
2:47:46change the world. Such a big idea. So
2:47:48baba uh probably just or at least
2:47:51improve one process or service in
2:47:54government. Then a question suddenly
2:47:58pops up.
2:47:59Wait a minute. What if uh or has
2:48:03somebody already done this idea or is
2:48:05there any existing project or innovation
2:48:08that I can take inspiration from? So
2:48:12that's where this uh project or this
2:48:14story begins. So our learner then would
2:48:18start searching for resources. And
2:48:22here's the challenge. Library search
2:48:25works best if you already know the
2:48:28title, author or exact keywords. But
2:48:32let's say uh your learner doesn't have
2:48:34yet a very clear idea of uh what to look
2:48:37for. So they would try one keyword and
2:48:41then another and go through titles, open
2:48:44papers and check whether they're
2:48:46relevant. Sometimes the paper might
2:48:49already be existing but it uses a
2:48:52completely different set of words. So
2:48:55what we want uh what our team wants here
2:48:57is a situation wherein or just like
2:49:00asking someone for directions and being
2:49:03told sure just tell me exactly where
2:49:07you're going and I'll tell you how to
2:49:09get there.
2:49:11So you see the problem isn't that our
2:49:13libraries lack good resources or lack uh
2:49:16lack good research. The problem is that
2:49:19good research can be difficult to find.
2:49:24So what if our learner could simply ask
2:49:26a question?
2:49:28So ladies and gentlemen, meet our
2:49:30project um or our idea to class. It is
2:49:33an AI powered research discovery tool
2:49:36designed to help our learners in the
2:49:37academy find research that's relevant to
2:49:40their ideas. Think of to class as a
2:49:43research body.
2:49:46A person who understands what you're
2:49:48looking for and even when you don't know
2:49:51the exact title or keywords and it does
2:49:53three simple things discover, explain
2:49:57and verify. So with discover it helps
2:50:00you find research related to your topic.
2:50:02Explain it tells you why those set of
2:50:05papers or capstone projects might be
2:50:08useful to yours and verify is uh helping
2:50:11you show the original sources so you can
2:50:15check for yourself.
2:50:17Behind the scenes AI helps the class
2:50:19understand your question and find
2:50:21related ideas not just matching words
2:50:26and it searches only our approved
2:50:28research collection.
2:50:32So why does this matter for our learners
2:50:34at the academy? It means spending less
2:50:37time looking for research and more time
2:50:39for actual reading or reading actual
2:50:42helpful materials, studying them and
2:50:44developing their or developing great
2:50:47ideas. For DAP, it means valuable
2:50:51knowledge and research personnel.
2:50:55valuable knowledge from previous batches
2:50:57that doesn't simply sit in a library
2:50:59waiting to be discovered. So for our
2:51:02librarians,
2:51:04it means more time for guidance and
2:51:06mentoring instead of manually searching
2:51:08through records. We're not just making
2:51:10documents easier to find. We're giving
2:51:12old knowledge a chance to inspire new
2:51:15ideas.
2:51:18Now, um this is important. Um, doesn't
2:51:22AI sometimes make things up? And the
2:51:25answer is yes. And we certainly don't
2:51:28want our learners citing research papers
2:51:31that do not exist.
2:51:34So that's why we're designing to class
2:51:36with safeguards. It searches authorized
2:51:40documents, shows where information comes
2:51:42from, and should tell you when there's
2:51:44no good match. We're also considering
2:51:47privacy, security, and possible gaps in
2:51:51our research collection. Most
2:51:53importantly, the learner still decides
2:51:56what the research actually means because
2:51:58AI can help us discover, but people must
2:52:02still do the thinking.
2:52:05And here's what we're aiming for in this
2:52:07project. 50% less time spent in finding
2:52:11useful research. at least 80% relevance
2:52:15or user satisfaction. At least 90% of AI
2:52:18explanations traceable to actual sources
2:52:21and 100% accuracy in paper details with
2:52:25no invented citations when there's no
2:52:27match. Now, these are our targets, not
2:52:30results we're claiming to have achieved.
2:52:32But imagine the possibilities if we
2:52:35could reach them. less time searching,
2:52:38better informed research ideas and more
2:52:41opportunities to build on knowledge DAP
2:52:44has already produced. So now I've talked
2:52:48about what the class is, what it can do
2:52:50and why we believe it matters. So and we
2:52:54know that uh what people say seeing is
2:52:57believing. So enough of the
2:52:59explanations, let's see the project in
2:53:02action. So to walk us through our
2:53:04functional prototype, let me turn you
2:53:06over to my teammate An.
2:53:23Okay, thank you so much Matt. So for the
2:53:26live demonstration of our uh tool, this
2:53:29is to class.
2:53:32can show. Okay, so this is the user
2:53:36interface. Um, all you need to do is to
2:53:38enter your proposed research inquiry,
2:53:41policy, challenge or uh development
2:53:44management topic and the tool will uh
2:53:47search from the official data set and
2:53:49provides AI synthesized relevance uh
2:53:52relevance rationalize. Okay, so for this
2:53:55one we already inputed the question. So
2:53:58our question is what previous capstones
2:54:01explored AI for learner feedback. So
2:54:06once we click define papers it will list
2:54:09down or or it will generate a list of um
2:54:13capstone for this data set. We only have
2:54:16uh two batches from the middle manager
2:54:18class. So it will give you lists of um
2:54:23approved capstone projects. And as you
2:54:26can see here, there's a percentage which
2:54:29uh is the relevance uh percentage.
2:54:34So once um you've seen the top list, you
2:54:38can check the full details and it will
2:54:41show you the abstract of that capstone.
2:54:45And if you want to dig uh deep on its uh
2:54:48academic framework and governance
2:54:50analysis, you can also download this
2:54:53academic this deep academic framework
2:54:55governant an governance analysis. So
2:54:58later I will show you the PDF file.
2:55:03Okay, you can also copy the APA citation
2:55:05or if you want a Chicago or Harvard um
2:55:10uh citation.
2:55:12So that's for a single paper. How about
2:55:15um if a learner wants to know if he or
2:55:19she can compare two or more capstone
2:55:23papers that that is relevant to the
2:55:26question. So on to class we can also do
2:55:29that. So you need you just need to
2:55:31select the caps uh the capstones and
2:55:35then it will uh generate through the
2:55:40discovery matrix.
2:55:43So it will compare those um topics and
2:55:47it also it will also give you um
2:55:49comparative discovery synthesis.
2:55:52So uh the PDF file is also downloadable.
2:55:55So for the results,
2:56:01so our learners can also use this as
2:56:03reference to their um uh capstone plan
2:56:08together with the abstract. And for the
2:56:11comparative
2:56:16here uh it will show you the executive
2:56:18synthesis summary of uh the selected
2:56:20capstone projects, thematic overlaps and
2:56:24divergence, method methodological
2:56:26insights and research frameworks,
2:56:29governance and public policy impact,
2:56:31identified research gaps and scholar
2:56:33opportunities.
2:56:38So yeah that's um to class a tool for
2:56:42unified knowledge location and archive
2:56:44sources. Thank you.
2:56:49Thank you team to class. Thank you an
2:56:52Matt. Next we have an AI [music]
2:56:55supported foresight intelligence
2:56:58assistant that we call trendsatch 2.0.
2:57:02Let us welcome the project lead from the
2:57:04Center for Strategic Futures, Mr.
2:57:06[music] Ray Daniel Peralta.
2:57:12Hello. Hi.
2:57:19Um, good afternoon everyone. I'd like to
2:57:22uh a ask uh wait sorry the there's no
2:57:26screen here.
2:57:29Dis can you just hand me my laptop
2:57:31please?
2:57:34An when we talk about digital
2:57:36transformation we often we often ask our
2:57:40uh agencies whether the kind of
2:57:41technology that we've adjusted or the
2:57:43kind of technology we've adopted is
2:57:45something that's correct but I'd like to
2:57:46ask and pose a different kind of
2:57:48question.
2:57:50How do we prepare government for
2:57:52challenges that have not yet emerged?
2:57:55How do we make decisions today when the
2:57:57future itself remains uncertain?
2:58:00The United Nations 2.0 transformation
2:58:02areas defines the force identifies
2:58:06foresight as one of the essential
2:58:07capabilities that institutions must
2:58:10develop and foresight as we understand
2:58:13it is not necessarily us being able to
2:58:16predict the future. In fact, it is about
2:58:19anticipating emerging developments,
2:58:21exploring possible disruptions and maybe
2:58:24even to prepare ourselves so that we
2:58:26could make better decisions today. And
2:58:28that is precisely the kind of ambition
2:58:30that we want to share with you today.
2:58:33At the Center for Strategic Futures at
2:58:35Development Academy of the Philippines,
2:58:37we have been exploring how AI could help
2:58:40make strategic foresight much more
2:58:42accessible, much more systematic, and
2:58:44ultimately become more actionable. So
2:58:47ladies and gentlemen, this afternoon
2:58:48we're going to share with you trendsatch
2:58:502.0. At its core, trendsatch 2 trends
2:58:54begins with one simple question. What
2:58:57kind of futures are you exploring?
2:59:00Imagine being a government planner, a
2:59:02policy maker or even a researcher who is
2:59:05attempting to understand the possible
2:59:07implications of AI or in the recent news
2:59:10climate change super elino. Or maybe you
2:59:13want to know about demographic shifts or
2:59:14emerging technologies or super that we
2:59:18that is actually
2:59:21mainstream.
2:59:23So traditionally this would require us
2:59:25to go through countless reports,
2:59:27articles, database and even other
2:59:30sources of information. And while
2:59:32information access has actually become
2:59:34so much more accessible, making sense of
2:59:36that information is quite a significant
2:59:40challenge. So this is where trends watch
2:59:42comes in. The platform helps the user to
2:59:45discover and explore emerging trends,
2:59:48signals and developments that may have
2:59:50implications for the future. And
2:59:53importantly, we wanted to go that beyond
2:59:56just simply presenting a collection of
2:59:58uh summaries. We wanted to provide
3:00:01learners a more structured approach and
3:00:03the way so that they could examine the
3:00:05information that was presented to them.
3:00:07As you can see, the application uh
3:00:09organizes the developments into
3:00:11different types of signals. It provides
3:00:14uh an appreciation of their relevance in
3:00:17terms of the context on the Philippines.
3:00:19And it even considers the the kind of
3:00:21context that it relates to in the
3:00:23Philippine context.
3:00:25And when you click a result, the
3:00:28particular development uh becomes what
3:00:31does this mean for the Philippines? What
3:00:34should our government and institutions
3:00:36begin thinking about this kind of change
3:00:38or this kind of trend? So users can
3:00:41imagine the individual signals, look
3:00:44into the sources which is very important
3:00:46and explore why this particular
3:00:49development might deserve greater
3:00:51attention.
3:00:52Not every piece of information is
3:00:54essentially uh necessarily a meaningful
3:00:57signal and not every signal should
3:00:58automatically become a government
3:01:00priority. So the real value actually
3:01:03lies in our ability as government
3:01:05practitioners to distinguish between
3:01:07what is really uh interesting for our
3:01:10agency and what may actually be
3:01:12consequential. And of course you met uh
3:01:15you met a lot of different personas
3:01:18earlier through uh Tala and we also have
3:01:21Wana. So to make this experience a
3:01:23little more intuitive, we've also
3:01:24introduced our intelligent guide Sephi
3:01:26who takes after our cyerspace tar who is
3:01:29also a
3:01:32It's a it pays homage to the kind of
3:01:34vision that we have when we when we are
3:01:37scoping and scanning the future. So
3:01:39beyond giving the platform a literal
3:01:40personality, it uh we wanted users to be
3:01:43able to navigate the platform much more
3:01:46um naturally and with more information.
3:01:49So users can refine their research and
3:01:53explore different categories of emerging
3:01:55developments. They can examine the
3:01:57signals through political, economical,
3:02:00social, technological, legal, and even
3:02:03environmental perspectives when we're
3:02:05scanning the signals. But identifying
3:02:07signals is just the beginning. What
3:02:09excites us even more is its ability to
3:02:12turn those signals into meaningful
3:02:14strategic conversations.
3:02:16uh the application allows the users to
3:02:20maybe examine an emerging trend or a
3:02:22development through different
3:02:23institutional lenses like for example
3:02:25the implications of that signal on a
3:02:27particular technology may be different
3:02:29from for a for an NGA it may be
3:02:31different for an LGU or it may even be
3:02:33different for a researcher. So we're
3:02:35dawning different lenses as we're trying
3:02:36to understand the insights generated by
3:02:38the tool and through the strategy map
3:02:42that we have built, we want users to
3:02:44begin exploring these trends and
3:02:46developments in relation to their
3:02:49potential implications. So we've got
3:02:51here uh act now if you want to monitor
3:02:54or if you just want to you might want to
3:02:56look out into this trend or maybe uh
3:02:59just
3:03:01yeah nice to know or monitor. So we want
3:03:04to support a more anticipatory way of
3:03:06thinking about policy planning and
3:03:08public service delivery.
3:03:11Now one important consideration in
3:03:12developing trends 2.0 is understanding
3:03:15that AI has a tendency to be syncopatic
3:03:17and to of course hallucinate. So uh when
3:03:20AI produces information that sounds
3:03:22credible or it wants to uh tend it has a
3:03:25tendency to agree with you we don't want
3:03:27that to happen and this is particularly
3:03:29important for a tool that is being used
3:03:31in strategic foresight. So we don't want
3:03:33AI to just simply tell us what we
3:03:35already believe. So we've de
3:03:36deliberately built guard rails so that
3:03:39uh into trends search 2.0 to address
3:03:42this risk. Our approach emphasizes that
3:03:45AI generated insights should be should
3:03:48be could be retrieved in identifiable
3:03:50sources.
3:03:52So ultimately trends 2.0 O is our
3:03:55contribution to making strategic
3:03:57foresight much more accessible and much
3:03:59more uh appreciated by uh across the
3:04:02different government agencies. And as we
3:04:04continue exploring what artificial
3:04:05intelligence could do, uh we believe
3:04:07that its value is not really in being
3:04:10able to generate answers but to help us
3:04:12as government agencies ask better
3:04:14questions so that we could all uh thrive
3:04:16and be able to make technology that is
3:04:20only as useful as the decisions that it
3:04:22helps us to make. Thank you very much.
3:04:23And that's watch.
3:04:27>> Thank you, Mr. Peralta. Mr. Peralta is a
3:04:30futurist and therefore [music]
3:04:32uh that was his inspiration behind
3:04:34transwatch.
3:04:36Uh thank you Transwatch team. Completing
3:04:39our showcase today [music] is Project
3:04:41Oracle, an AI powered policy assistant
3:04:44for DAP. Please welcome our last
3:04:47presenter, our project lead from the
3:04:49finance department, Mr. Dean Hans
3:04:52Anetta.
3:04:59>> [music]
3:05:08>> All right. Good morning everyone. My
3:05:10name is Dean Hansenetta from the finance
3:05:13department of the DAP. So as uh many of
3:05:16us here uh right now uh
3:05:20product uh national capability building
3:05:23of uh in AI
3:05:26development academy of the Philippines
3:05:28specifically the training of uh
3:05:30trainers. So this one is one of the uh
3:05:33capsules projects that has that that has
3:05:35been produced and to formally introduce
3:05:38it to you let join me to in watching
3:05:41this video.
3:05:49The development academy of the
3:05:51Philippines runs on written policy. More
3:05:54than a thousand circulars, special
3:05:56orders and office orders, all of them
3:05:58PDFs. Somewhere in that pile is the
3:06:01answer to the question you have right
3:06:03now. The hard part is finding it one PDF
3:06:06at a time. [music] This is project
3:06:10Oracle, a policy assistant for DAP built
3:06:12on retrieval augmented generation. It's
3:06:15still a prototype. So let's look at why
3:06:18it's worth building, how it [music]
3:06:19works, and what it can do. To tell this,
3:06:22we'll borrow Simon Synynx golden circle.
3:06:24Start in the middle with why it should
3:06:26exist, then how it works, and only then
3:06:30what it is.
3:06:31So, let's start at the center.
3:06:34Picture someone at DAP with a simple
3:06:37question. Can I monetize my leave? Who
3:06:40qualifies for a communication allowance?
3:06:43What does a visitor need to enter the
3:06:44building? Today, finding the answer
3:06:47starts with a search. You open a
3:06:49circular, scroll through it, and realize
3:06:52it's the wrong one. So, you close it and
3:06:54open the next. 20 minutes later, you're
3:06:56on your 12th PDF and the question still
3:06:59isn't answered. And even when you find
3:07:02it, it might be outdated. The February
3:07:04circular on communication allowance was
3:07:07amended in July, and nothing on the
3:07:09February version tells you that.
3:07:12So, here's our why. Everyone at DAP
3:07:15should get the right rule in its current
3:07:17version with the page it came from. The
3:07:19answers are already there across a
3:07:21thousand issuances waiting to be
3:07:23harvested. [music] The how is retrieve
3:07:26augmented generation or RAD. Before the
3:07:29AI says anything, it pulls up DAP's own
3:07:31documents, then answers only from what
3:07:34it found. Step one, indexing. Every
3:07:37circular, the quality manual, and the
3:07:39citizens charter are cut into short
3:07:41passages. [music] Step two, retrieval.
3:07:44Your question is compared against every
3:07:46passage and the eight strongest matches
3:07:49come back circulars first. Step three,
3:07:52augmentation.
3:07:53Those passages and any amendment history
3:07:56go to the model with strict
3:07:57instructions.
3:07:59And step four, generation. [music]
3:08:01Claude writes the answer in plain
3:08:03language, and every sentence points back
3:08:05to the page it came from. On top of
3:08:07that, Oracle follows a few rules. If the
3:08:10documents don't cover your question, it
3:08:12tells you so and points you to the right
3:08:14office. When [music] a newer circular
3:08:16amends an older one, the newer one wins.
3:08:19Every claim can be traced to its page.
3:08:22And if you ask about a citizen's charter
3:08:24service, you [music] get the complete
3:08:25list of requirements. So, what does it
3:08:28look like? A web app. You ask and it
3:08:30answers with its sources attached.
3:08:32Answers come back in plain language and
3:08:36every statement carries a numbered
3:08:38citation so you know exactly where it
3:08:40[music] came from. Click a citation and
3:08:42the circular opens right on that page.
3:08:45The lines the answer relied on are
3:08:48highlighted in yellow. If a newer
3:08:50circular has amended an older one, the
3:08:52answer flags it at the very top before
3:08:55you read anything else.
3:08:57Ask about a citizen's charter service
3:08:59and you get every requirement where to
3:09:02secure [music] it, how long it takes and
3:09:04what it costs. Any answer can also be
3:09:07rewritten in simpler English or in
3:09:09Filipino for staff who would rather read
3:09:12it that way. And before you go ahead
3:09:15with a plan, Oracle can check it against
3:09:17the rules one requirement at a time and
3:09:20tell you what's missing. [music] So
3:09:22where does Oracle stand? It's a working
3:09:24prototype, not yet deployed, tested on
3:09:2626 circulars, the quality manual, and
3:09:29the citizens chart. To scale it, we load
3:09:32DAP's full library, host it securely in
3:09:34house, pilot it with a few offices, and
3:09:36index each new issuance the day it comes
3:09:39out. That's Project Oracle, built so
3:09:42that anyone at DAP can find the right
3:09:44rule in its current version with the
3:09:46page it came from.
3:09:50All right. So that's project Oracle and
3:09:52we'll now proceed with the demo. So what
3:09:55you are seeing right now in this screen
3:09:57is the live web app. So let's say I am a
3:10:01new employee in the DAP and I am uh
3:10:04about to ask um is there [snorts]
3:10:08a uh dress code in DAP?
3:10:17So it reads all the passages from the uh
3:10:21vector embeddings from its database
3:10:25and generates the answer using claudet
3:10:295.5.
3:10:31All right. So it answered yes. D has an
3:10:35official dress code under MC 2025016.
3:10:38So business casual Tuesday to Thursday
3:10:41smart casual during Fridays. So if you
3:10:44click this one, you can see the exact uh
3:10:49scanned PDF and the the highlighted
3:10:52portion where uh this answer is based
3:10:56from
3:10:57and so monitoring compliance your center
3:11:01uh uh endorses documented violations of
3:11:04the standing committee on the
3:11:06discipline. So you can see another
3:11:08policy that is cited. This is another uh
3:11:12policy 2026007
3:11:16uh where it documents all the
3:11:18violations. Next one uh let's say
3:11:23I am a CPA practicing
3:11:27my profession.
3:11:29Is this allowed in the DAP?
3:11:33So let's ask it
3:11:37again. It reads again all the passages
3:11:40in its database uh 2025 and 2026
3:11:44memorandum circulars in the DAP
3:11:48and it generates the answer.
3:11:51Yes, it is allowed but only with prior
3:11:54written permission from the DAP
3:11:55president and CEO and subject to
3:11:57specific conditions. So a request memo
3:12:00supporting document from the institution
3:12:02engaging your services and again you if
3:12:05you click the uh button right here you
3:12:09can see the exact uh citation in which
3:12:13the answer is based. So you can see
3:12:16actually if the large language model is
3:12:19hallucinating or not. So last uh let's
3:12:23say an employee is asking
3:12:29uh I am uh resigning next month. What
3:12:36documents should I prepare for the
3:12:41processing of my last pay?
3:12:47So this one includes a process that is
3:12:51uh indicated in the citizens charter of
3:12:54the DAP. So it will also site that one.
3:12:59So this one so based on the DAP citizens
3:13:02charter uh the documents you need to
3:13:05prepare is letter of resignation duly
3:13:08approved by the head of agency
3:13:10accomplished clearance form and all of
3:13:13that. And when you click it, you can see
3:13:16uh where it is exactly cited with the
3:13:20highlighted portion of it. That's the
3:13:22whole document. And it gives you the uh
3:13:27tab for the documents and requirements.
3:13:32Yeah. And that will be all for the
3:13:35project oracle of DAP. Thank you very
3:13:37much.
3:13:41>> Thank you Mr. Anetta. And thank you to
3:13:43everyone all of our presenters for
3:13:45today. We now come to our closing
3:13:48plenary where we bring together insights
3:13:50and look at the steps we can take beyond
3:13:53this forum. To help us connect the
3:13:55lessons from our three sets, please
3:13:58welcome our director at the center for
3:14:00strategic futures, Mr. Armand Tristan
3:14:03Soratus for connecting the dots from
3:14:06insight to action.
3:14:20Um hello everyone.
3:14:23So how do we begin to digest and make
3:14:25sense of such rich and thoughtprovoking
3:14:28discussions, presentations and
3:14:30perspectives?
3:14:32Throughout this morning, we have heard
3:14:33from institutions and individuals in the
3:14:36Philippines and across the Asia-Pacific
3:14:38region from both the public and private
3:14:40sectors. We have also seen AI
3:14:43applications and solutions that
3:14:44illustrate what is already possible.
3:14:48As we bring these discussions together,
3:14:50I would like to return to the central
3:14:52premise of this forum. What does it
3:14:55really mean for government to move
3:14:56beyond DX?
3:14:59I would like to draw out some of the
3:15:00insights from today's proceedings. What
3:15:03we have learned, what these insights
3:15:05might mean for the public sector, and
3:15:08how we might translate them into action.
3:15:10First, what have we learned?
3:15:13One recurring message is that the value
3:15:15of AI should be measured in public
3:15:17outcomes, not simply in the number of
3:15:19technologies adopted.
3:15:22The perspectives from Thailand and the
3:15:24Republic of Korea bring us back to the
3:15:27experience of citizens.
3:15:29More responsive services, less
3:15:31inconvenience, and potentially
3:15:34government that can anticipate needs
3:15:36rather than
3:15:38only react to requests. Our keynote
3:15:40speaker offered a useful distinction.
3:15:43Digitization asks whether a service is
3:15:46online.
3:15:47Digital transformation asks whether it
3:15:49is seamless. And AI transformation asks
3:15:52whether it is better for people and
3:15:54whether they trust it.
3:15:57This is an important distinction because
3:15:59AI transformation is different in kind.
3:16:03AI can influence decisions, reshape how
3:16:06public servants work, and change how
3:16:08government interacts with citizens.
3:16:11Its value therefore extends beyond
3:16:14efficiency to better services, better
3:16:17informed decisions, more meaningful
3:16:19public work, and greater public trust.
3:16:22The AI solutions demonstrated today also
3:16:25gave us concrete examples of how these
3:16:27possibilities are already being
3:16:28translated into practical applications.
3:16:33Another message is a pro is a promising
3:16:35application is not the same as a
3:16:37successful government service.
3:16:40The USDST's perspective draws draws an
3:16:43important distinction between a working
3:16:45prototype and a solution that can be
3:16:47deployed reliably maintained and scaled.
3:16:52Dependent and Taiwan likewise underscore
3:16:54the that governance safeguards and
3:16:57infrastructure and actual institutional
3:16:59needs are not peripheral concerns. They
3:17:02are part of what makes adoption possible
3:17:05and trustworthy.
3:17:07And as the keynote reminded us, when
3:17:09government increasingly relies on AI
3:17:11systems, questions of security,
3:17:14reliability, and accountability become
3:17:16even more important.
3:17:18A third message concerns people and
3:17:20organizations.
3:17:22Japan's perspective reminds us that to
3:17:25distinguish work that AI can support
3:17:27from work that continues to depend on
3:17:29human judgment and relationships.
3:17:32related to one of the principles of
3:17:34DAP's AI capability program. Training
3:17:37individuals is necessary, but
3:17:40institutions must also be ready to use
3:17:42their capabilities.
3:17:44Leadership, workflows, decision making,
3:17:47and organizational culture matter just
3:17:50as much as familiarity with the tools.
3:17:54Now, what do these insights mean for the
3:17:56public sector?
3:17:58When we put these perspectives together,
3:18:01they suggest something more fundamental.
3:18:03Beyond DX is not simply a technology
3:18:06agenda. It is a question of how
3:18:08government itself may need to change.
3:18:11We should be willing to reimagine
3:18:13services, not merely automate the steps
3:18:15that already exist. If a procedure is
3:18:18confusing or burdensome for citizens,
3:18:22making it faster may help, but
3:18:24redesigning it around their needs could
3:18:26matter more.
3:18:27We also need to think of AI readiness as
3:18:30an institutional capability.
3:18:33A trained workforce cannot deliver
3:18:35sustained results without appropriate
3:18:37data, systems, safeguards, leadership,
3:18:41and room to improve how work gets done.
3:18:44And importantly, readiness can't be a
3:18:47one-time achievement. Technology, risks,
3:18:51and public expectations will keep
3:18:52changing. Government needs the capacity
3:18:56to test ideas, assess their effects,
3:18:59learn from what works and what does not,
3:19:01and adapt continuously.
3:19:04So, how do we move from insight into
3:19:07action? Of course, not every agency is
3:19:10starting from the same place. Some are
3:19:13still strengthening their digital
3:19:15foundations. Other are experimenting
3:19:17with AI, while some are beginning to
3:19:20integrate it into their regular
3:19:21operations.
3:19:23The experiences shared today are
3:19:25therefore not templates that every
3:19:27institution must copy. They offer
3:19:30principles we can adopt to our own
3:19:32circumstances.
3:19:34Our keynote speaker also proposed a
3:19:36practical 12-month agenda from
3:19:39establishing ownership and identifying
3:19:41public service problems to testing
3:19:43solutions with safeguards, measuring
3:19:45outcomes, and scaling what works. What
3:19:49is significant about this agenda is that
3:19:51it begins not with choosing a technology
3:19:53but with identifying a problem and
3:19:56defining what success should look like.
3:19:59Perhaps you can take three questions
3:20:00back to our respective institutions.
3:20:03What public problem are we trying to
3:20:05solve? In what improvement would
3:20:07citizens actually notice?
3:20:10What institutional capabilities and
3:20:12safeguards would make that improvement
3:20:14possible?
3:20:16And what can we begin doing now while
3:20:18preparing for the changes we may face
3:20:20over the next three to five years? That
3:20:22to me is the connection between insight
3:20:24and action.
3:20:26Our starting points may differ, but the
3:20:28direction can be shared toward a
3:20:31government that is more capable,
3:20:33responsive,
3:20:34accountable, and prepared to serve
3:20:36people in a changing world. And perhaps
3:20:39that is what moving beyond the X
3:20:41ultimately means. not simply keeping
3:20:44pace with techn technological change but
3:20:47shaping how government evolves and
3:20:49ensuring that the changes we make
3:20:51genuinely pro improve the lives of the
3:20:53people we serve. Thank you very much and
3:20:56have a good day.
3:21:06Thank you, Director Seratus, for
3:21:08bringing together today's insights and
3:21:10pointing us towards action. Now, we now
3:21:13present certificates of appreciation to
3:21:17our presenters and exhibitors for
3:21:19sharing their innovations and expertise
3:21:22at Beyond DX. May we invite our
3:21:25officers, President Sebastian, executive
3:21:28vice president Mendoza, [music] and uh
3:21:32uh senior vice president Kaluen, and uh
3:21:34director Saratus to be here on stage as
3:21:38we present [music] uh the certificates
3:21:39of appreciation.
3:21:52Heat. Heat.
3:22:30First up, we have uh the presenters of
3:22:33Tala by Anti- Red [music] Tape
3:22:35Authority.
3:22:56Next we [music] have a
3:23:29Thank you. Tala by Arta. And now we have
3:23:32the department of budget [music]
3:23:34and management.
3:23:36Thank you for presenting project Bella.
3:24:12Thank you DBM. We have lifesaver.
3:24:13[music]
3:24:27Oh,
3:24:40next up we have Haya.
3:25:13Next up we have front learning. ers.
3:25:43Hey
3:25:49I call on stage.
3:26:00I call on stage once again life savers.
3:26:12the
3:26:32baby.
3:26:43Thank you, Lifesavers.
3:26:46Next up, we have Sirmo by Mobility
3:26:48Vision Plus.
3:27:19Next up, we have Tala.
3:27:59Next up, our presenter [music] from DAP,
3:28:03Transwatch 2.0.
3:28:52Next up we have the Yes.
3:29:09You have a very supportive uh bunch
3:29:11here.
3:29:13Join forces COE, CFG,
3:29:16LGDO.
3:29:30Thank you to [music] class. And lastly
3:29:32from DAP, Project Oracle.
3:30:11Thank you, Dean. Next up, our [music]
3:30:13exhibitors. First up, Envycom
3:30:16Corporation.
3:30:50Thank you. Next up, Chat Genie.
3:31:23Next up, DV code technologies.
3:31:33I am.
3:31:50And lastly,
3:31:52we have the DOSD [music] Advanced
3:31:53Science and Technology Institute.
3:31:59Heat. Heat.
3:32:33Yep. And we have Haya, but I think he
3:32:36already left. Uh, Sir Luis Benaventura.
3:32:41Yeah. We'll just uh give a copy [music]
3:32:44of him to this.
3:32:49Now, thank you all to the presenters and
3:32:51exhibitors for making AI in action a
3:32:54highlight of Beyond EX forum. And now to
3:32:58deliver our closing message, let us
3:33:00warmly welcome the senior vice president
3:33:02for programs, Emldda C. Kaluen.
3:33:06>> Thank you
3:33:08our distinguished speakers, partners,
3:33:13exhibitors, fellow public servants and
3:33:16colleagues.
3:33:18Thank you for being part of today's
3:33:20forum.
3:33:22Through our discussions and
3:33:24demonstrations, we have explored how AI
3:33:28can support public service and what it
3:33:31takes to move from promising ideas to
3:33:34practical applications.
3:33:37We have also seen the importance of
3:33:39leadership, capable people and
3:33:43institutions prepared to manage change
3:33:47responsibly.
3:33:49So the next step is to bring these
3:33:51conversations into our respective
3:33:54organizations.
3:33:56For some this may mean identifying a
3:33:59process that could be improved. For
3:34:03others, it may involve helping
3:34:06colleagues develop foundational AI
3:34:10skills, examining the readiness of their
3:34:14data, or giving a promising initiative
3:34:18to support it needs to move forward.
3:34:24Each is a meaningful step towards an AI
3:34:27ready government.
3:34:29As we look ahead, the Development
3:34:32Academy of the Philippines will continue
3:34:36to advance the national capability
3:34:39building program on artificial
3:34:41intelligence,
3:34:43building on our foundational courses. We
3:34:47are preparing more specialized learning
3:34:50opportunities
3:34:52in planning, human resource management,
3:34:56financial management, and advanced
3:34:59applied AI.
3:35:01Our aim is to help public servants
3:35:05translate learning into improvements
3:35:08in the work that they do and the
3:35:11services they deliver.
3:35:14We invite your continued support to this
3:35:17effort. To our agency leaders,
3:35:21please nominate personnel whose roles
3:35:24align with our learning opportunities.
3:35:28Give them time and support to complete
3:35:31their training and create opportunities
3:35:34for them to apply what they have
3:35:36learned. Your encourage your
3:35:39encouragement can make the difference
3:35:42between an individual gaining a skill
3:35:46and an institution developing a a
3:35:50lasting capability.
3:35:52to our partners in the private sector,
3:35:56academ,
3:36:00continue sharing your expertise,
3:36:04practical experience
3:36:06and lessons from implementation.
3:36:09Work with us in developing relevant
3:36:13learning opportunities
3:36:15and helping government institutions
3:36:18address real service challenges.
3:36:23To our fellow public servants,
3:36:26remain curious,
3:36:28ask thoughtful questions and share what
3:36:32you learn.
3:36:34You understand the work of government
3:36:36and the needs of the people you serve.
3:36:40that knowledge is essential to deciding
3:36:44where AI can make a useful and
3:36:47responsible contribution.
3:36:50For our part, the academy will continue
3:36:55to listen, learn, and strengthen the
3:36:58program. We welcome your feedback on the
3:37:02capabilities
3:37:04your agencies need,
3:37:07the barriers you encounter and the
3:37:09support that would help you move
3:37:12forward.
3:37:14Our ambition is a public sector where
3:37:18people have the competence and
3:37:20confidence to use AI,
3:37:24where leaders can make informed
3:37:26decisions about the adoption and where
3:37:30improvements are measured by their value
3:37:33to clients.
3:37:35Let us carry today's exchanges forward
3:37:39through sustained partnerships,
3:37:42concrete initiatives,
3:37:44and a shared commitment to better public
3:37:48service.
3:37:50On behalf of the Development Academy of
3:37:52the Philippines, thank you to our
3:37:55speakers, exhibitors,
3:37:57private participants, and the team who
3:38:00made this forum possible.
3:38:09atong
3:38:12parabatipino.
3:38:17[applause]
3:38:19>> Thank you, Mommy. And as
3:38:25to our evaluation forms, uh there
3:38:27there's just two uh and you can respond
3:38:30to them within five minutes. It's okay.
3:38:33Thank you so much.
3:38:51Invest
3:39:10It's
3:39:27down.
3:40:17alone.
3:40:41Hey,
3:40:46hey, hey.
3:41:02down.
3:41:12Hey
3:41:24Hey,
3:41:34hey, hey.
3:41:40Heat. Heat.
3:42:26Heat. Heat.
3:42:47One.
3:43:11Hey. Hey.
3:43:46And we have one last surprise for you
3:43:50and then we can all have our lunches
3:43:52outside buffet. By the way,
3:43:58shut off the [music] lights, please.
3:44:29It seems
3:44:31surprise. We can all [music] now have
3:44:34our lunches outside. Uh, and then
3:44:37balikto
3:44:39tables.
3:44:44But we will play it over while we're all
3:44:47eating.
3:45:46Hey.
3:45:48Hey. Hey.
3:46:52favorite.
3:47:04Hey
3:47:57Hey,
3:48:06hey, hey.

About this transcript

This page contains the full transcript of Beyond DX: AI Transformation Forum for the Philippine Public Sector by Development Academy of the Philippines, generated from the public captions YouTube serves with the video. The transcript has 23,614 words across 4,060 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.