YT YouTube2Text

YouTube2Text → Transcripts → Augmenta – The AI That Turns Weeks of Electrical Modeling into Hours

Augmenta – The AI That Turns Weeks of Electrical Modeling into Hours — Transcript

by Stjepan Mikulić · 18,010 words · 2,590 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00[music]
0:05[music]
0:14[music]
0:21Heat. Hey. Hey. Hey.
0:23[music]
0:28>> [music]
0:31>> Hi everybody. Uh, welcome to a new demo
0:35day with Augmenta. No, this one is not
0:38getting rescheduled or cancelled.
0:40Everything's fine. Everything's working.
0:42I had a bit of a technical difficulty in
0:44the back background right now, but
0:46everything's fine working as you see.
0:47So, it's all good. Thank you for your
0:49patience with a couple of reschedules
0:51this year. Um, you know, life happens.
0:54what can you do? Um, but here we are.
0:56Um, so again, if you're first time
0:58watching this, my name is Stephan. I go
0:59by any version of Stefan. That's
1:01completely fine. I am from Croatia, but
1:05I live in Los Angeles right now and I
1:07have two companies, AINAC and NeoStack.
1:10And I just want to tell you shortly
1:11about AIAC. Um, so if you just go to our
1:15website, aac.com,
1:17it's your best resource for learning AI
1:19and knowing how to implement AI as well
1:22within your firm. Um first part you can
1:25get the free ACI hub which is the
1:27world's largest database of AI tools
1:31um that exists in the world and we just
1:34made a big update to it. We added a
1:36bunch of dates, we added a bunch of
1:38categories, a bunch of new tools and
1:39it's getting updated um like regularly.
1:43However, if you want to learn something
1:45uh too, like how to use AI, how it all
1:48works, what are the best productivity
1:50apps, how can you save yourself 12 hours
1:52a week at least. I mean, this is just
1:56marketing. Don't get me please, you
1:58know, it's America. I know people can
1:59sue each other for a bunch of random
2:01stuff, but anyway, sorry. You can uh
2:04join discovering NAC. It's our
2:06fundamental course and we are actually
2:08moving toward a um subscription based
2:13model. So this is still where you can
2:14get it uh once for a lifetime. We will
2:17keep this model. Don't worry, it's not
2:19only subscription based, but we are
2:21opening up a community like so to speak
2:23a membership. Um so this is going to be
2:26something that you will be able to join
2:28and you will get as well if you join
2:30today. We're going to be launching in a
2:31few months. Um, so yeah, this is I guess
2:34just a bit little bit little bit about
2:36the business and um, let me just see
2:39here. This was pretty much it what I
2:41wanted to share. So again, thank you for
2:43the patience. Hey Nathan, thank you for
2:45commenting. I see here I'm also very
2:47excited to be here. So please write
2:49where you're calling from. I would
2:50really love to know that location. And
2:54while you write that, I would love to
2:56invite Erin to the stage. Hi Erin.
2:59>> Hey Stefan. Thanks for having me on. Of
3:02course. Nice to meet you and thank you
3:03for the patience and waiting for another
3:05couple of weeks for this.
3:06>> Of course. Of course.
3:07>> Yeah. Erin, please tell us a little bit
3:10about yourself. Who are you?
3:12>> Um, what is Augmenta? Like, you know,
3:14let's start with you first and then we
3:16can talk about the company.
3:17>> Yeah, sure. So, um, I'm Aaron. I'm the,
3:20uh, co-founder and the chief product
3:22officer at Augmenta. Um, bit about my
3:25background. So, I didn't actually come
3:26from construction. I came from
3:28industrial design. So I used to design
3:30Blackberries at BlackBerry but I moved
3:33into software uh moved into digital
3:36product design research with really a
3:38focus on 0ero to one products and a bit
3:42about how I got into this domain is um I
3:46met my co-found co-founder at Autodesk.
3:48So at Ardesk uh Francesco Yoro was the
3:52director of computational research there
3:54and he invented and then spearheaded the
3:57generative design program there. So
4:00generative design at the time this was
4:02around 2015 or so was actually very much
4:05focused on advanced manufacturing and
4:07that was of course much closer to my
4:08background as well and I worked with him
4:11there to really develop the principles
4:13of how people would interact with these
4:14systems. So it's really sort of pre the
4:16current um you know breakthroughs and
4:19attention around ML uh and it's really
4:21one of the first applications of using
4:24computers to design solutions for you as
4:26opposed to you know traditional CAD
4:28interface where you are drawing a
4:30solution that you've developed in your
4:32head. So there are a lot of new problems
4:34to solve. You know no one have really
4:36sort of thought about how to do this
4:37before. I worked with him to really
4:39develop the principles of how people
4:41should define a problem. what do I do if
4:43this thing's presenting tens or hundreds
4:45of results back to me? How do I make
4:47sense of that? How do I iterate on that?
4:49Provide feedback to it. And you know,
4:51one thing led to another. We sort of
4:53always felt that there was so much more
4:56we could do with this approach. Uh we
4:59thought AC was really the right place to
5:01approach it. So eventually we decided to
5:03create Augment to uh solve that problem.
5:06That's what we've been doing now for a
5:07couple years.
5:08>> Okay. So tell me a little bit about that
5:10uh jump you guys made. So Franchesco and
5:13yourself met out as you said it.
5:16>> Yeah.
5:16>> And who's who said it first? Who were
5:18like hey let's open a company like let's
5:20let's get out of there out of here.
5:22[laughter]
5:24>> Um it was a very slow process very
5:26organic process. I think that you know
5:29we just sort of felt that um
5:34how much to say here? um you know
5:36Autodesk Autoesk is in a different
5:38position within AEC than they are in
5:40manufacturing right within AC they have
5:42a sort of very much a dominant position
5:45whereas in manufacturing they don't
5:47they're competing against very uh
5:49entrenched you know large players so you
5:52have a company like the so who is far
5:55more sort of um the standard in advanced
5:58manufacturing sort of talking aerospace
6:00automotive and so on than someone like
6:03autoes so for autoes I think that they
6:05were much more motivated to really try
6:08and push generative design as this new
6:11potentially disruptive model in that
6:13space but less so in construction. You
6:16know, I think in construction they're
6:19doing very well for themselves as as we
6:21see there's not as much internal
6:23incentive to try and rock the boat too
6:26much. I think that is changing now to
6:27some extent. Um and of course you know
6:30they have they have their own factors
6:32that they have to contend with but what
6:33it meant for us ultimately was that you
6:35know on the ground is us building this
6:37technology and being very excited about
6:39it. Uh really believing that this could
6:41fundamentally change how people use
6:43computers to solve very complex
6:45multi-variable problems which is what
6:48construction is. Uh we were just
6:50frustrated. We were increasingly
6:51frustrated that you know we couldn't
6:54really get traction
6:56applying this idea to that domain. So
6:58the conversations happened actually over
7:00a course of a couple years that sort of
7:01just it was not sort of like a big flip,
7:04but it just sort of was this slow sort
7:06of realization that okay, we were going
7:07to we were going to have to do this on
7:09our own for real if we really wanted to
7:11see this through and really see it to
7:12the extent that we wanted to see it. You
7:15know, we never wanted it to be just
7:17another plugin, just another drop down
7:20or a button on a on a toolbar somewhere.
7:22We really thought you know if you
7:24approach this problem from a first
7:25principles perspective and treat the
7:28design of buildings like the design of a
7:30integrated circuit or or a chip you know
7:34um it is a there's a real chance to
7:37introduce a new paradigm where you are
7:39in fact using computers to take in a
7:42bunch of requirements and literally just
7:43design an entire building single shot.
7:46Actually doing that is very very hard.
7:48It's going to take a while to get to a
7:50point where we're doing that with entire
7:51buildings, but it is feasible. It is
7:53possible and we built this company and
7:56started this company to really be the
7:59ones to try and make that a reality. So,
8:01that's kind of the northstar that we've
8:02been working towards.
8:04>> Okay. Yeah. Thank you for that
8:05background story. I'm so happy you
8:06asked. Um,
8:08>> yes. Listen, we have people from all
8:09over the world. Uh, hottie here is very
8:12excited. We have Tim. Hey, Tim. Thank
8:14you for joining us always from uh
8:16Minneapolis. um from Minnesota, minus5
8:20degrees Fahrenheit or 5 degrees
8:22Fahrenheit. And then we have Sy from
8:24Oricius for 25.
8:26[laughter]
8:28>> Yeah. Um Yes. Where are you actually
8:31based? Is where is
8:32>> Yeah, I'm in Toronto.
8:34>> In Toronto. Okay. Awesome.
8:36>> Yeah. Yeah.
8:36>> So, we're we're found in Toronto, but
8:38we're a global team. We're fully remote.
8:40So, we've got folks we got a large team
8:41in the US, bunch of folks here in
8:43Canada, bunch of folks throughout Europe
8:45as well. So very much a distributed
8:47firm.
8:48>> Yes. Uh listen, I know you have a had a
8:50presentation. Um yeah, you can go ahead.
8:52Should I pull it up?
8:53>> Yeah, sure. Let's do that.
8:54>> Yeah. Okay. Awesome. Go ahead.
8:56>> All right. So, what I'm going to do here
8:58is kind of just walk everyone through
9:00sort of a high level introduction. I
9:02know sort of the goal of these sessions
9:04is the demo. So, we will focus on the
9:06demo. Um so, I'll treat this more as a
9:08table of contents. So we'll kind of talk
9:10about what we'll show you, but we'll go
9:12through all that live and you guys can
9:14hop in with any questions you have at
9:15any point. But where we are today as
9:18Augment is we are focused on electrical
9:20as you kind of might you know um as I
9:24sort of suggested previously our
9:25ambition is to automate the design of
9:28entire buildings. That is what we're
9:29working towards. But right now we've
9:31started off with electrical. Number of
9:33reasons for that. But fundamentally our
9:35goal today is to help electrical
9:38contractors and electrical engineers
9:40dramatically accelerate their modeling
9:43and coordination process through the use
9:45of an AIdriven design system.
9:49So a bit about myself. We already
9:51touched on this, but I'm Aaron Samansky.
9:53I'm the chief product officer and one of
9:55the co-founders at Augmenta.
9:58What we do is we build these systems
10:00that automate the design of really the
10:02core engineered system. So we really
10:04care about the things that are behind
10:06the drywall that really drive the cost
10:08and coordination or you know cost and
10:10complexity of construction projects. So
10:12your mechanical, electrical, plumbing
10:14systems, your structural systems. This
10:17is really the hard part of uh the
10:19building design process.
10:22So where we are today is um we've got
10:25this product that automates the routing,
10:27modeling and coordination of this
10:28conduit or of electrical raceway with a
10:31focus on conduit and most of our users
10:33today are electrical contractors
10:36primarily in the United States and
10:38they're using us to get from zero to one
10:40very very quickly. So you've just kicked
10:42off a project. You've got your assets.
10:45You've got your Revit model. There's
10:47nothing in there. And you need to get to
10:49that first coordination meeting, which
10:50is, you know, next Friday very, very,
10:53and you need to get that model populated
10:54very, very quickly. So that's really our
10:56sweet spot right now. So we'll take that
10:58model from zero to something that is
11:00fully populated. It's coordinated. You
11:03can bring that into that coordination
11:04meeting uh with a lot of certainty
11:06around where everything's going to go.
11:08uh the quantity the size I'm going to
11:10need for you know for me to fit all that
11:12feeders and my branch and so on and it's
11:15really about leaprogging you from that
11:17start point to a fully you know modeled
11:20uh system very very quickly and it's
11:23really fundamentally the value is that
11:24you can do that in way less time than it
11:26would take to do that manually and that
11:27sort of unlocks other benefits as well
11:29that we can talk about so I won't spend
11:32a whole lot of time on this because I'll
11:33just hop right into a demo but again
11:35just kind of think of this as sort of
11:36like a table of contents
11:38uh of the demo. Fundamentally, the way
11:41it works is that you define your
11:43problem. So you give it a Revit model,
11:44you give it your sources and
11:46destinations in that model. You can
11:48provide your preferences or the context
11:50you have around where you can or can't
11:52route in that model. You provide a
11:55description, a conduit run schedule that
11:58describes the the runs that you want to
12:00or the equipment that you want to
12:01connect. So your two and froms in a
12:03given project. You can specify
12:05additional design rules like your
12:06elevations, your supporting methods,
12:08your clearances. And once you've set all
12:11that stuff up, you hit generate and
12:13you're kicking off a generation process.
12:15And because you won't really see what's
12:17happening behind the scenes in the demo,
12:19I do want to spend a little bit more
12:20time on this, but fundamentally what is
12:22happening is that our system, which
12:24operates in the cloud, is extracting
12:26site geometry, the information that it
12:29needs about your site, and it's
12:30performing a spatial analysis on that
12:32site. And essentially what we're doing
12:34or the way we explain is that we're
12:35trying to look at your site the way a
12:36foreman would look at it where our goal
12:38is to produce solutions that are not
12:42just sort of generically routed but are
12:44actually constructable. So we perform a
12:47supportability analysis based on the
12:49types of supporting methods you specify.
12:51So that can be a floor stand, trapze, uh
12:54you know wall standoff and so on. It's
12:57going to look at the site, identify the
12:59spaces in that site that can be
13:00supported with those methods. So, for
13:02example, it's going to prioritize areas
13:05with open slab above over let's say, you
13:08know, directly underneath a duct where
13:10there would be some obstacles to get
13:11that threaded rod around and and get
13:13that thing supported. We're looking at
13:15the penetrability of walls. We have an
13:17understanding of the cost of
13:18penetrations based on, you know, the
13:20type of the wall, the um amount of
13:23conduit that's going to be running
13:24through that wall and so on. So there's
13:26a number of different analyses that the
13:27system runs and then within that
13:29understanding of the site it generates a
13:31number of designs and what it's doing is
13:33it really is singleshot full uh building
13:37routing and and coordination. So by
13:39coordination I mean it's producing clash
13:41free designs in a single shot and it's
13:43really solving a number of problems
13:44here. So it it's obviously finding the
13:47optimal pathways for every single source
13:50and destination. So every run that you
13:52need to run in that project. It's
13:54figuring out optimal racking strategies.
13:56So, it's going to figure out how do I
13:57group these together to minimize the
13:59amount and the cost of the supports in a
14:01given project. How do I minimize the
14:03bends to reduce the difficulty of
14:06installation to reduce my total degree
14:08of bend? Um, it's figuring out the
14:10optimal layout within that rack. So, how
14:13many tiers should it be? It will vary
14:15that based on, you know, external
14:17factors like the space that's available.
14:19Uh and again it's trying to reduce the
14:21cost of the supports that would be uh
14:22used on a given design. It's figuring
14:25out where to peel off and so on. So it
14:26really is solving a number of different
14:29levels of problems that BIM and DDC
14:32teams have to solve every single day.
14:34And that's ultimately what makes this so
14:35hard to do for those teams.
14:38Once you've done that, you can review
14:40the designs. Uh you can review them in
14:423D, but you can also review the analysis
14:44of those designs. you can learn
14:46something about the performance of these
14:48different designs. So for example, I can
14:50look at any one of these solutions. So
14:52here we've got a 3D view. We've got a
14:54plot chart and I can look at what's the
14:56estimated cost of this, what's the
14:57estimated time to install, what's the
14:59amount of material, the degree of band.
15:01And where this is really useful is that
15:03you know in just a couple hours I've
15:05just routed an entire hospital floor. So
15:08if I'm an electrical engineer, I'm say
15:10working on a design build contract or
15:12I'm sort of approaching a value
15:14engineering problem. Well, now I can go
15:16and compare what would happen if I ran
15:18this with aluminum versus copper
15:20feeders. There's a size difference in
15:21those. There's a cost difference in
15:23those. I can just run both in parallel.
15:24And now I can get this very informed
15:27feedback on what is the optimal design
15:29for a given project. This is
15:31particularly useful for folks working at
15:33the earlier stages of the design
15:34process.
15:36Once you're happy with the design, you
15:38can push it back into Revit where we've
15:40got additional tools for QA and QC uh to
15:43really help you make sense of that
15:44design because you've really just gone
15:46from zero to one, right? So, you started
15:47with an empty model. Now, you've got
15:49hundreds of runs populated. We need to
15:51make you we need to help you make sense
15:53of those designs so you know what you're
15:54looking at and can verify that
15:56everything's routed the way that you
15:57want to. But, you know, there's another
16:00reason we need this tooling that is that
16:01our designs are not perfect, right? So
16:03it's just we're very much in the early
16:05days of really applying ML to the design
16:10of um well to the creation of designs in
16:14a complex 3D context. It's much easier
16:17to apply to 2D. You know if we're
16:19producing sort of highle sketches or
16:22highle routes over a floor plan that's
16:24much more of a solved problem than it is
16:26in 3D. And so the result is that today
16:29you know there are going to be things
16:30that you have to fix. there's going to
16:31be some soft collisions, extra bends,
16:33and so on. I'll show some of those in
16:35the demo just to kind of help you, you
16:37know, characterize those and get a sense
16:39of um, you know, what that looks like in
16:42practice.
16:44But ultimately, the way we measure
16:46whether this is helping our customers or
16:48not is by comparing the time it takes
16:51for you to set up. So, let's say, you
16:52know, create that schedule, get the
16:54Revit model set up, the time it takes to
16:56generate, which because it happens in
16:58the cloud is sort of passive time. you
17:00can go work on other things while that's
17:01happening and then the time for you to
17:03do that review to make edits whether
17:05because you're seeing issues in uh from
17:08a solution quality perspective or you
17:10just simply want to approach that design
17:12in a different way because of something
17:14you've learned during coordination. And
17:15so we sort of include all of that and
17:17generally what we're seeing is that
17:19teams using this tool are cutting their
17:21time by about 60%. Or so and that rate
17:25of time savings has increased in a very
17:28linear fashion over the last several
17:30years. So again, it's like any other ML
17:32tool. We're making significant
17:33advancements in the performance of the
17:35system over time. Our solutions are
17:37getting better and better. There's less
17:38and less editing required. The setup
17:40gets faster and easier and so on. So
17:43even what I'm going to show you today is
17:44really just a snapshot. You know, if if
17:46we were to give you this demo even a few
17:48months ago, it would be a 50% reduction
17:51in time. Today it's around 65%. And
17:53that's, you know, continuing to
17:54increase. So we're really targeting to
17:56get folks to about a 75% reduction. So
17:59about 4x faster by the end of this year.
18:01And we're definitely on track to do
18:03that.
18:04So just some examples of what that looks
18:06like on some more complex projects. You
18:08know, this is a hospital. So I've sort
18:10of hidden all the background geometry
18:11but this was against fully you know
18:13detailed and coordinated mechanical and
18:16plumbing backgrounds. Uh same for a
18:18science lab. So very similar level of
18:19complexity but on average we're able to
18:21about cut that in half to down to about
18:24a third.
18:26So what we're talking about today is
18:28electrical but
18:31our goal is to do this not just for
18:34electrical but for mechanical and
18:35plumbing as well. And it's not simply
18:38about just copying the workflow I'm
18:39going to demo over to those trades, but
18:42it's about being able to simultaneously
18:44coordinate those systems in a single
18:46shot. So imagine everyone setting their
18:49own requirements, their own schedules,
18:51their own preferences around things
18:53around where things route and you hit a
18:56button and you're gening a fully
18:57coordinated MEP system in a single shot.
19:00Really what we're trying to do is we
19:01want to eliminate the coordination
19:03meeting. It's going to take us a little
19:05bit of time to get there. Obviously,
19:06it's a very complex problem to solve,
19:08but like I alluded to before, this is
19:10very much a solvable problem. And that's
19:12really why we built this company because
19:14once we can start to do that, that's
19:16when we can really start to take weeks
19:18or months off of construction schedules.
19:20Uh whereas with electrical today, we're
19:22helping the electrical contractor a lot.
19:24There's certain types of projects where,
19:25you know, a data center, electrical is
19:27the critical path, but for us to really
19:30bring this benefit to the whole
19:31industry, we need to at least be hitting
19:33these MEP systems simultaneously. And
19:36that's what we're working towards as
19:37well.
19:39Let's see. Um, and then we got a
19:41question. Can you plan to penetrate
19:43steel beams along the neutral axis where
19:46the above ceiling space is limited? So,
19:49you can, and what I can do is I'll touch
19:51on that in a little bit in the demo. So,
19:52by default, the system will not
19:54penetrate steel beams unless you give it
19:56explicit permission to do so. So, we
19:58sort of have um I'll I'll show you, but
20:00we've got different routing boxes where
20:02you can indicate areas where
20:04penetrations are possible and that gives
20:06the system the option to penetrate
20:08through there. So, that's really how you
20:09would do it in those cases. But by
20:11default, we are always going to avoid
20:14any solid geometry in that site. The
20:17exception would be walls and floors and
20:19things like that where you can set more
20:20general rules allowing for penetration.
20:23Um, but anything a little bit more, you
20:25know, anything in the structural
20:27categories or the mechanical employment
20:28categories, it really is up to you to
20:30make that explicit indication of where
20:32you can and can't penetrate.
20:36All right.
20:38All right. So, that's sort of a high
20:40level overview. Um, yeah, would love to
20:42hear any any more questions otherwise we
20:44can hop into a demo.
20:46>> Um, yeah, this was awesome. Thank you
20:49for presenting it uh so well. So um um
20:52yeah, now it makes sense how you got got
20:53started with electrical. I was about to
20:55ask you like because you you showed some
20:57projects that probably going to like is
20:59it I was about to ask is it only
21:01residential that you're covering but you
21:02showed healthcare, you show data centers
21:04like is it is there any limits to what
21:06this can do? Is there any building shape
21:09that this like augment can work with?
21:13>> We are not currently limited to any
21:16building typology. We perform best
21:19actually on more complex projects. So
21:22we've run multiple data centers,
21:24multiple hospitals. We actually have not
21:27run I don't think any res residential
21:29projects. I think we actually have one
21:30going on now that's a a condo building.
21:33Um but we find really the biggest
21:34application of of this is is in those
21:36more complex projects because those are
21:38the ones that need it, right? If you
21:39really have hundreds of runs that need
21:42to be woven through these very tight
21:44spaces. Um you know basically what we
21:46find is that the more complex there's
21:50sort of a in that initial setup period
21:52of setting up ACP there's like a linear
21:55increase in setup time but a uh sort of
21:59uh far more aggressive rate of time
22:01savings. So that means on a very small
22:03simple project, we've seen this like um
22:06like a clinic, you know, it's a it's a
22:0810 room clinic. To me, that's a
22:11threshold of, you know, given the time
22:13it would take for you to set this up,
22:14you might just want to model that
22:15yourself. We might save you some time,
22:17but it's not it's not you're not going
22:19to get that enormous time savings.
22:21Whereas on a data center or hospital
22:24where you are routing um you know this
22:27is weeks weeks weeks of manual modeling
22:30and coordination that's where you see
22:32those really significant time savings.
22:34So I would say that our most common
22:36project types are data centers. We've
22:38got multiple running right now. Um
22:41hospitals we've seen a lot of hospitals.
22:43The that seems to have quieted down a
22:45little bit in the last sort of quarter
22:47being taken over by data centers. We've
22:49had several manufacturing facilities I
22:52would say. So we do manufacturing
22:54facilities. Well, one thing that we
22:57don't support quite so well yet is heavy
22:59industrial facilities. And the reason
23:01for that is you tend to see a lot of
23:03very specific support methods. You know,
23:06folks are often building custom support
23:09to route through those sites and we're
23:12not as good as interp at interpreting
23:14sort of the geometry in in those sites
23:16quite yet. Um whereas something like a
23:19manufacturing facility we can we can
23:20handle. So I'm talking about like a oil
23:23refinery versus a manufacturing
23:24facility. We can do the manufacturing
23:26facility. Oil refinery would be a little
23:27bit more challenging for us. Um we've
23:30done a number of commercial projects but
23:32as we all know commercial is not um you
23:35know there just hasn't been as many
23:36commercial projects recently.
23:39Um so it really is quite quite diverse.
23:42Um and we have a question about the um
23:45uh routing underground.
23:47So we do route underground in a limited
23:49fashion right now. It's very much in a
23:51in an alpha stage. Uh so we route
23:53underground but grid aligned as in you
23:56can route but it's going to be you know
23:5990s uh kind of like a snake right. Uh
24:02what we don't currently support is
24:03pointto-point routing sort of as crow
24:05flies which is what you get uh quite
24:08often in these underground projects. So
24:11that's sort of where we're at. It's
24:12something that we are working on this
24:14year is sort of true uh underground
24:16routing. Um
24:20so keep an eye out for that if that's
24:22something you're interested. We we'd
24:23love to talk about that. Uh working with
24:25existing conditions. So not yet. We
24:28don't currently work with point clouds
24:29particularly well. Uh the challenge with
24:31point clouds is that a human being can
24:34look at a point cloud and they know
24:35what's what. It's actually hard for our
24:37system to extract meaning out of that
24:40type of geometry. So that's something
24:41that we're that we're working on and I
24:44can't say we officially support quite
24:45yet unfortunately. So generally this is
24:48something that's going to work better on
24:50green field projects for that reason.
24:53Can we design modular hospitals? Um well
24:58yes. I think that
25:00there's nothing particularly unique
25:03about from our from our software's
25:04perspective if okay I guess I suppose it
25:07depends on what you mean by modular.
25:09We've had customers using us on modular
25:11projects where there are you're putting
25:15the building together with these
25:16different components but the feeders
25:19that are feeding those different
25:21sections are still being routed almost
25:23like on a per project basis. So that's
25:25one application that we've seen. The
25:27other is within those modular
25:29components, I'm trying to design modular
25:32electrical rooms and so on. We actually
25:33have a customer using us to do that as
25:36well. So you'll run our our system to
25:38produce that design and of course you're
25:39then replicating that design afterwards.
25:42Um and then the last question for how
25:44much it costs is hard for me to answer
25:46here because it really depends on the
25:47size of your team and and your and how
25:49much volume you run through. So we
25:51charge on a uh well our our licensing is
25:54companywide. So we have unlimited users,
25:56unlimited projects. We don't really sort
25:59of uh you know there's no limitations
26:01there. Uh if you kind of think about it,
26:03our biggest cost is when you hit
26:04generate. The compute is extremely
26:05expensive. So what we do is we take in
26:07an estimate of what's the approximate
26:10volume that you're going to be running
26:11through the system and um uh then the
26:15pricing is based off of that. So that's
26:16something that would just need a follow-
26:18on conversation um if you want to get
26:19more detailed pricing.
26:22modular plug-and-play connectors and
26:24different cable management levels. So
26:26the one thing um just around that is
26:28that we don't currently support tray
26:31which is sort of an important limitation
26:33to call out. So it really is right now
26:35on just the conduit side. So where
26:39people are generally using us is there
26:40well in the US many projects are run
26:43with conduit most still are. So people
26:46are running feeders they're running
26:47branch uh with our system tray is
26:49something that we're working on. And how
26:51do you feed it input? I'll show you that
26:53in the demo. But yes, it is conduit run
26:56schedules.
26:57>> Okay. Awesome. Erin, thank you, man.
26:59Like, you're doing my job for me. I
27:02That's awesome, man.
27:04>> I will I will let you ask the questions.
27:05>> No, no, no. Please, please. Hey, go
27:07ahead, man. It's your show. Listen, I'm
27:08just here as a as a messenger anyway.
27:11>> Um, so I'm not here, so don't worry
27:13about it. It's just that I'm I'm just
27:15joking, you know. So, anyways, please go
27:17ahead with demo. You see that I'm live.
27:19Please uh everybody in the audience ask
27:21your questions. Um you see Erin is
27:23amazing at covering them and answering
27:26them and yeah if anything sparks sparks
27:29your curiosity please just ask and we're
27:31going to cover it afterwards. Erin, go
27:33ahead. Stage is yours.
27:35>> All right. Great. So there is great
27:37demand for uh yes we do have plans to um
27:42move into the European market with
27:44greater support for cable trade.
27:48So, um,
27:51let's kick off. What we're looking at
27:53here is Revit. We all know that. And
27:56what I'm going to show you is how you
27:58would use ACB to produce what we have on
28:01the screen here. So, this is a large
28:03sort of commercial um, uh, commercial
28:06warehouse um, type, let's say, yeah,
28:09commercial sort of warehouse type
28:11building. Uh, so this is a real project.
28:13Um, I won't sort of zoom in too much on
28:15the building details, but I will zoom in
28:17on the conduit details. So, everything
28:20we're seeing here, uh, think about how
28:22long it would take you and your team to
28:25model and coordinate this manually. I
28:27won't sort of feed you with a with an
28:30answer. But when we, you know, ask
28:31different customers, this, they'll say
28:33something in the order of, let's say,
28:34one to two weeks. And this is something
28:36that was entirely generated single shot
28:38within ACP in about 40 minutes. So there
28:41was some setup time to go build out the
28:43schedule, but the actual generation
28:44process was extremely fast. And uh the
28:47output is something that we bring into
28:49Revit. And we can zoom in here. You can
28:51see that we're connecting to all the
28:53equipment. We're laying out those racks.
28:55We're coordinating around different
28:57obstacles. I'll I'll spend more time
28:59looking at the solutions in a little
29:00bit. We'll we'll look at it in ACP, but
29:02I just kind of want to start at the end
29:04here to show you what type of output is
29:07possible. And honestly by our standards
29:10this is a very simple project. So uh
29:13something that you can generate in 30 to
29:1540 minutes is at this point fairly easy
29:18for us to handle. Uh we're running data
29:21centers that can take up to 12 hours to
29:23generate. So you're running them
29:24overnight with you know we've run 18
29:28miles of conduit in a single shot
29:30overnight um in a single generation. So
29:33we ran all the feeders and almost all
29:35branch for entire data center uh in a
29:38single shot. So we've managed especially
29:41over the last year to make a lot of
29:42progress in terms of the scale site that
29:44we can handle. But so how do we get
29:47here? Well, let's start off with the
29:50setup that happens inside Revit. Okay,
29:52so what I'll do is I'm going to hop over
29:54to the ACP import view. So basically
29:56we've got a a special view inside AC
29:59inside Revit that controls what ACP
30:02sees. So by default ACP that's the
30:05augmented construction platform is going
30:07to route around like I mentioned before
30:10solid geometry. Of course it has special
30:12handling for things like walls and and
30:14certain elements. But generally speaking
30:15it's going to coordinate around anything
30:17you want it to coordinate around. And
30:19you control that by controlling the
30:20visibility of those categories or
30:22elements in this view. So, I've got my
30:25building open. I've maybe filtered out
30:27the things I I wanted to care or don't
30:29want to care about. But really, the most
30:31important starting point is that I've
30:32got my equipment. So, Augmenta does not
30:34place equipment for you. This is
30:36something that you've placed. It's
30:38either coming from the engineer model or
30:39it's something that you've placed
30:40yourself. And every piece of equipment
30:43here, you know, if I open up, let's say,
30:44a panel or so, it's got some kind of a
30:46unique identifier.
30:50Where is that panel name? DH2A. Right?
30:54So,
30:55not saying anything particularly
30:56revolutionary here. Every piece of
30:58equipment. So, that's my sources, my
31:00destinations, is going to have some kind
31:01of a unique name. And I'll show you how
31:02we're going to use that in a little bit.
31:05So, very first thing we have to do is we
31:07have to on this project define our two
31:09and froms. So, I have gone and I've
31:11placed all my equipment. I've placed my
31:13junction boxes, my home run boxes. So,
31:15all the sources and destinations have
31:17been placed. I'm going to launch ACP. I
31:19actually had it running back here.
31:25Okay.
31:27And I'm going to go and and while we're
31:30here, we're looking at ACP. So ACP
31:32itself runs in a web view that's
31:34embedded inside Revit. So we have this
31:37plugin here. And really the goal of this
31:39plugin is to read data out of Revit and
31:41then later to push a solution back in.
31:43We've got some tools here that I'll talk
31:46about in a little bit that really help
31:47you mark up this model, help you provide
31:50some of those preferences. And you can
31:51kind of see hints of those colored boxes
31:53back here. That's me providing
31:54information around where I can, can't,
31:56or would prefer that you route. We've
31:58got other tools for doing our QAQC after
32:01that I'll talk about. But really the the
32:03brains of the system is running in the
32:06cloud and it's accessed through this web
32:08app. So right now I'm inside our well
32:10this is our our demo environment, but
32:12this would be you know your company's
32:13personal environment. We're inside our
32:15projects and we're looking at this demo
32:17project here. And what we see here is a
32:20list of studies. So studies are really
32:22the key organizing concept inside ACP.
32:25And a study is basically a whatif
32:26scenario. It's basically saying with
32:28this particular setup of this site, if I
32:30were to use this conduit schedule, this
32:32site and hit generate, what type of
32:34solutions am I going to get out of that?
32:36So earlier when I talked about doing for
32:38example copper versus aluminum feeders
32:41um or as changes are happening to that
32:43project I can create these new studies I
32:45can generate against them. I can run
32:47multiple studies in parallel and then I
32:49can go and actually compare them with
32:51each other. I can compare the stats but
32:53I can also overlay them over each other.
32:54So I can see what is the routing or the
32:57space impact of one sort of design
32:59decision over another. So if I create a
33:02new study you know I just hit create new
33:03study. I have sort of a blank one here.
33:06And opening this up, the first thing we
33:09see is the conduit schedule and feeder
33:11schedule. So that's going to answer
33:12Sanjay your question. This is really one
33:15of our our primary inputs here. And
33:17what's cool is that, you know, I can
33:18build out this conduit schedule, which
33:20for those maybe less familiar, the
33:22conduit schedule simply is actually I'll
33:25create one and then we can talk about
33:26what's in it. So a conduit schedule
33:28defines the runs, the conduit runs I
33:31want to run between a given source and
33:34destination. So each one has a source
33:37and I can go in here and it's actually
33:39live linked to Revit. So it's pulling in
33:41all of the names that it can find in
33:43Revit. So I'll just pick one. I'll pick
33:45one just randomly here. Uh it's got a
33:47destination. I'll pick another one
33:49randomly. It's got a system. And systems
33:52for us are something that we can later
33:54use to target design rules to. And then
33:57it's got a feeder ID. And a feeder ID is
34:00basically just a reference to a separate
34:02schedule, which right now is empty. I'll
34:04load one in that defines the number and
34:07the sorry the quantity and the size of
34:10the parallel runs. So I'm basically
34:11saying from ATS2 to CL1B
34:14uh that belongs to the power system and
34:16I want to run you know two twoinch
34:18conduits between that. So if I went to
34:20the feeder schedule I can go and build
34:22out a new row. I just got to give it an
34:24ID. I can say you know one parallel run
34:26and I can pick a size and then I can go
34:29and reference that here. Right? Okay.
34:32And now I've got this run.
34:37Let's see.
34:39All right.
34:41So the other thing is like you just saw
34:43that when I load these in or when I load
34:46these names in, these are actually live
34:48linked to Revit. Right. So this ATS
34:50number two, if I click the link button,
34:52I can go and actually find that element
34:54in Revit. And the reason this is
34:56important is that if I change data in
34:58Revit, if I change the name of a panel
35:00in Revit, that conduit schedule is going
35:03to automatically detect that change.
35:04It's going to flag that, hey, something
35:06has changed here. It's going to suggest
35:07that there's something with a similar
35:09name. So, if I went here and let's say I
35:12just did ATS2,
35:15you know, it's going to go and say, hey,
35:17there's, you know, we couldn't find an
35:18ATS2. There's something very close
35:20though. Uh, let's go look at what that
35:22is. ATS number two, right? That's what I
35:25meant for that to be. I'm going to go
35:26ahead and change that here. So, this is
35:28basically a real-time verification
35:30system that makes sure that my schedule
35:32is always completely accurate against
35:34the model that I have. And this is
35:35really important. You know, this a here
35:37if I've got 40 runs, maybe that's
35:39something I can manage myself and keep
35:41in my head. But again, these complex
35:43projects can have hundreds of runs. We
35:45all know that things are changing in the
35:47backgrounds. And also, people make
35:48mistakes. They make these little
35:50changes. Maybe I wrote ATS2 in my uh
35:53Excelbased conduit schedule, but it's
35:55ATS, you know, pound sign two um in the
35:59model, and that's a link that's now
36:02broken, right? There's now a mismatch
36:03between that data. So, we've put a lot
36:05of work into building this experience
36:06where we have this sort of real-time
36:07two-way link between the schedule and
36:09and the model. But let's go ahead and
36:13clear this out and put in some real data
36:15so we can go ahead with the rest of the
36:17demo.
36:20All right. So the other way we can build
36:22out these schedules is most of the time
36:24you probably have a schedule to start
36:25with. Maybe you built it out during
36:26estimation. Maybe your planning team has
36:28built it out. So we support the import
36:31of CSV or Excel formats. So I've got,
36:34for example, a schedule here. Let me
36:36pull that up. You know that maybe I I
36:38pre-built here.
36:40Uh it's got a different format than the
36:42ACP one, but that's totally fine because
36:44we've got this tool here to map those
36:46together. So if I'll go here and I'll
36:48load that in and basically I just need
36:51to map up. So source that that's my from
36:53destination that's my two column system
36:56is my system column. Feeder feeder
36:57perfect match. We have this guidance
36:59column that I'll talk about in a second.
37:02But I don't really have an equivalent
37:03guidance column in my you know feeder
37:06schedule or my conduit schedule that I
37:07built in Excel. So I'm going to skip
37:10that and I can just go ahead and import
37:12that schedule.
37:13So, first thing you see is that there's
37:15a bunch of uh errors found here, and
37:18that's because we haven't loaded in a
37:19feeder schedule. So, let's just get that
37:21loaded in as well.
37:24And a lot of the time we see that, you
37:26know, folks who are using other systems.
37:30So, they're using uh various they've got
37:32an estimation or they've got a PLLD
37:34workflow that's creating these
37:35schedules. Maybe they're exporting them
37:37out of Blue Beam. Maybe they're creating
37:38them in Evolve. So, we really don't care
37:41where that schedule comes from. We've
37:42sort of designed this workflow to work
37:44with your process, the data that you
37:47already have. We really try to impose as
37:49few process changes as possible. And
37:51what this allows you to do, of course,
37:53is to well work with that, right? So,
37:55we've got this mapping workflow that
37:56really should take in any schedule
37:58format that you might have. So, we've
38:00loaded in the schedule and we see an
38:02example of what I was talking about
38:03before where we've got this mismatch.
38:05So, here someone typed in in that Excel
38:07C-1, but in Revit, it's actually called
38:10C1. We can go and verify that. We can
38:14change it. But we might actually say,
38:16you know, actually no, we want the
38:17schedule to be the source of truth.
38:20There is supposed to be a C-1 that's,
38:22let's say, different than C1. So the
38:24other thing that we can do is place it
38:26in Revit. And what this is going to do
38:28is it's going to open up this other tool
38:31where I can actually go and place
38:33devices off of my schedule. So I think
38:34where this is really cool is again a
38:36workflow where you've got a planning
38:38process, your PLLD team, you know, is a
38:40little bit more contractor specific.
38:42They've planned out their home run
38:43boxes. You've got those in a schedule.
38:45They're not in the model. I can go and
38:47systematically get all that content
38:48placed. Um, and make sure again that
38:50I've got a perfect match up with that
38:52model.
38:55But for now, we'll just change it here.
38:56We'll do that with this one, too. Okay.
38:59So, we've got our schedule all set up
39:01here.
39:03Next thing we need to do and uh let me
39:07just answer a couple of these questions
39:08that have come up. So Tim, you've got me
39:10a couple here. Will understand maximum
39:12cable length for cat 6? Uh no, not right
39:15now. So we don't sort of impose any hard
39:18limits based on cable length. What we're
39:20trying to do is we run a global
39:22optimization process that is trying to
39:23overall reduce overall length. We are
39:26building some tooling, some analysis
39:28tooling that's going to help you
39:29identify maybe where you might hit those
39:31limits. But really sort of the way to
39:34think about this is that the system is
39:36kind of working more towards goals. It
39:38doesn't sort of have hard limits around
39:41something like angle or length. It's
39:43really just trying to always minimize
39:44those as much as it can. Does it have a
39:47level of development that will reduce
39:49the model size? I'm going to assume you
39:52mean on the export side. So there's sort
39:55of the level of development on the
39:57import side as in what LOD is this
40:00background model. HP doesn't really
40:02care. It just needs the basic geometry.
40:05It needs to know where the walls and the
40:07floor is. That's actually technically
40:09all that you need. And you just need
40:11equipment in there. On the export side,
40:14today we export our designs just using
40:16generic um Revit system conduit. I'll
40:18show you an example of what that looks
40:19like. So we're not currently placing
40:21couplings and supports and things like
40:22that. So that's going to be um probably
40:26the lowest level of development that
40:28you'd really be interested in working
40:29at. Um and then from there we're really
40:32trying to make our output as agnostic as
40:34possible so that folks can use their
40:36existing modeling tools. So so you are
40:38using evolve. We make it very easy to
40:40flip over to a tool like that so that
40:42you can go and detail it yourself. And
40:44then would the contractor use the model
40:47to do coordination drawings where
40:48changes happen during coord
40:50construction? So that's a really good
40:52question and I would say that it kind of
40:54depends on the how big that change is.
40:57So a lot of our contractor customers,
40:59they're using this to do that initial
41:01population, right? And particularly, you
41:03know, we know this is sort of unique to
41:05electrical. They're kind of having to
41:07catch up the most compared to the
41:08mechanical and plumbing trades who
41:10generally have more to start with. So
41:11the electrical contractor, they're
41:13getting started. they're often starting
41:16with nothing uh modeled um and they need
41:19to get this populated very very quickly
41:21and I would say that most of the value
41:22for them is is that it's getting that
41:25initial model populated from that point
41:27on as the subsequent coordination
41:29process occurs there are going to be
41:31changes if those changes are localized
41:34like I just need to shift something from
41:36the left side to the right side of the
41:37hallway we would suggest that you just
41:39do that manually um right now it really
41:42is more of a um whole building design
41:46tool rather than sort of like a
41:48localized you know sort of a editor if
41:52that makes sense. But we have had
41:54projects where there were big
41:55coordination changes like turns out we
41:57can't run down the south side of this at
41:59all. we actually have to reroute
42:00entirely to the north where it's
42:01effectively a redo and for that it's
42:04perfect right because again you can go
42:06and take something that would have taken
42:08weeks to to do and now I can go and flip
42:10you know to this radically different uh
42:12path uh you know set of pathways very
42:15very quickly so yeah I guess to answer
42:17your question it really depends on the
42:18severity but really the big way to think
42:20of this is it's a global optimization
42:22process is a little bit more sort of
42:25whole building one shot so you sort of
42:27want to be using it where it makes sense
42:30to do so.
42:32Yeah, very good questions.
42:35All right, so we've got our schedule
42:37built out. So now HP knows what connects
42:40to what and what it needs to run between
42:42them. The next thing I want to do is
42:44give it a little bit more context around
42:45where I can or can't route in a project
42:50using the Vulve. Um Ariel, I'll answer
42:52that question in a second. Um but we um
42:56don't have that issue is is the short
42:58answer. So the
43:02way I do this is okay, I've I've I've
43:05got this project. There's things that I
43:07know about it that I need to tell ACPs,
43:09right? So ACP generally works with a
43:11fair bit of design freedom. I could sort
43:14of take what I've shown you so far,
43:16equipment, um, conduit schedule, and
43:19just hit go and it's going to go and
43:21route that entire building already with
43:22no additional guidance. But quite often,
43:25I have some idea in mind of where I want
43:27things to go. And we can specify that
43:29with these guidance tools. We've got two
43:31types, but they kind of work the same.
43:33We've got guidance boxes. The first is a
43:36preferred routing box where this is
43:38basically a way for me to specify my
43:40preference, right? So it looks like um
43:44it it's effectively uh I'm sketching out
43:47a magnet that draws the you know routes
43:51to that pathway. I'm sketching out may
43:54route boxes. Those are those ways I
43:56would indicate those penetrations. So
43:58again I want to indicate where I can
44:00penetrate through the beam. I can
44:03indicate where I penetrate through the
44:05keepout boxes and so on. and or sorry uh
44:08with the keep out boxes I I is sort of
44:10the inverse of that right so a keepout
44:12is a way for me to specify where I don't
44:14want you to go a keep out is a way for
44:16me to say look in this room this
44:18mechanical room you cannot run any rack
44:20through here but you're still allowed to
44:21hit equipment inside that space and a
44:24guidance space is basically the same
44:26behaviors except I can apply them to
44:28spaces instead of drawing these boxes
44:30manually but that's really what we see
44:32here right so we've got keep outs maybe
44:34we blocked out some shafts or some areas
44:36that we don't want any racks running
44:37through. We have uh we have different
44:40routing boxes. So we've got a
44:42combination of may route and force route
44:44boxes which are those preferred routing
44:46boxes. And this is a way for me to kind
44:48of provide this highle guidance to the
44:51system.
44:53Related to that is if I want to get
44:55really precise I can actually go and I
44:58can assign some of these runs to those
45:01uh preferred routing boxes. And what
45:02this will do is it will force that run
45:04to pass through that box. So I can set
45:09up these boxes essentially as way points
45:11for uh for those runs to pass through.
45:16Okay.
45:18So now I've gone and specified my to and
45:21froms where I want things to go where I
45:23don't want things to go and really at
45:25this point the next step is just to get
45:27that data up into ACP. So, I won't sort
45:29of show this live just because it takes
45:30a couple minutes, but this is really
45:32kind of the only syncing that happens to
45:35answer Harry's question where I would
45:38need to upload the geometry from that
45:40site over to uh over to Augmenta. This
45:44process generally takes a couple
45:46minutes. Um, but it is one shot. Once
45:48you actually hit that generate button
45:50and you start generating the solutions,
45:52it is running in the background. It's
45:54running in the cloud. So, it's not
45:55affecting uh Revit whatsoever. So
45:58generally this is something where you'll
45:59set it up in the morning, you hit
46:01generate, it's going to run in the
46:02cloud. We're not touching Revit at that
46:04point. You can go and continue working.
46:06Um you know after that at the end you're
46:09just pushing a solution back into Revit.
46:11That export process takes a couple
46:12minutes but there's no sort of um
46:15background syncing that happens. There
46:17is that sort of syncing or live
46:20connection that happens in the conduit
46:22schedule, but it's extremely light.
46:24We've put a lot of work into optimize
46:26it. So, we've never seen or heard about
46:28our customers having any kind of
46:30performance issues as a result of that
46:32running. And again, it's only running
46:34when you have that view open as well.
46:37So,
46:39so we we don't sort of have that problem
46:40where we're not we're not sort of live
46:42updating data in in your model uh as
46:45you're working through it. So, in this
46:47tab here, I've gone I've uploaded the
46:50electrical model and the background
46:51geometry.
46:53And then the last pane here is my design
46:56rules. So
46:58here I won't go kind of go through all
46:59of them. I'll touch on maybe some higher
47:02level than others. The first is
47:04supports. Again, I won't sort of dive
47:06into this in too much detail, but
47:07essentially what we're doing is even
47:09though we're not yet placing the
47:11supports themselves, we are very much
47:14considering the types of supporting
47:16methods that you specify when we're
47:18generating these routes. So in a
47:20separate part of the application, I can
47:21go and I can set up my support
47:24specifications. So it's literally just a
47:26prioritized list of, you know, trapeze,
47:28floor stand, uh, wall standoff, and so
47:31on. Again, I can and I can scope that
47:34down to different parts of the site. So
47:35I can say I've got, let's say, a
47:37manufacturing facility. This area is the
47:40um, uh, you know, floor stands are are
47:43acceptable here. Everywhere else is just
47:45overhead trap. In all the office areas
47:48we have different clearances from
47:50different elements. So we can specify
47:52clearances from walls, railings and
47:54things like that. We have our conduit
47:56spacing. So this is edgeto edge spacing
47:58uh horizontally and vertically.
48:01And then we have our elevations. This is
48:03really probably one of the most
48:04important types of uh rules that you're
48:06going to set. So we have a lower
48:09clearance. So basically what's the
48:10minimum height that this can run at?
48:12What's the maximum height that this can
48:13run at? That's my upper clearance. And
48:16then within that range, which might
48:17still be fairly big, what's my preferred
48:19height? Right? So we sort of can set
48:21this preferred elevation in that sort of
48:24allowable range. Uh ASP can still exit,
48:27enter and exit this range of course to
48:29go and actually connect equipment. But
48:30generally, you know, I want to run above
48:32the ceiling. Uh this is the allowable
48:34range above the ceiling and I want to
48:36try and stay below duct. So this is my
48:39preferred height is below where the duct
48:40gets run. Right? So then within that we
48:44are basically working on a very similar
48:46concept to that preferred routing box.
48:48It's basically a magnet. So the system
48:50will try to run at that height. It can
48:52still deviate where it needs to, but
48:54I've got, you know, say I've got a clean
48:56shot at here 11 ft above finish floor.
48:59There's something in the way. I can
49:00offset down around it and then get back
49:02up to that elevation.
49:06Other than that, we've got some general
49:07rules around routing through walls. So
49:09this just gives us permission to route
49:11through walls. We identify which walls
49:12are interior and exterior so that we're
49:14not punching and and exiting the outside
49:16of the building. And then same for
49:17routing through ceilings. Generally, you
49:19want to leave this on because you want
49:20to be able to sort of drop down and and
49:22hit equipment if it happens to be placed
49:24below a ceiling.
49:27All right. So, at this point, and I and
49:29I will pause and answer questions in
49:30just a second. Basically, this is this
49:32is the setup, right? So, this is
49:33everything that I need to do to define a
49:36problem for this system to go and solve.
49:40At this point, I would just hit
49:41validate, generate, and it will kick off
49:44that generation process. As it does so,
49:47then again, like I mentioned before,
49:48it's just running in the cloud. It's not
49:49affecting my Revit instance, and I'm
49:51free to, you know, um, work on work on
49:55other things. All right,
49:58so let's see questions from Okay, so
50:01Harry, I think I touched on your
50:02question.
50:05uh Marie looks super useful regardless
50:07of how it works but seems more like a
50:09mathematical optimization than AI if not
50:11what is the AI part it's a really good
50:13question and underneath the hood we're
50:16using a combination of more let's say
50:19traditional mathematical optimization
50:22and prove sort of datadriven machine
50:24learning the reason that this is not so
50:27this is not a entirely
50:30um let's say datadriven machine learning
50:34system. The reason for that is that
50:36particularly within a 3D context, we are
50:38not able today with present- day
50:41technologies to arrive at the level of
50:44accuracy that is sufficient particularly
50:47for contractor applications where they
50:50need something that they can bring into
50:51actual coordination meetings and get
50:53actual design sign off on. So, we knew
50:55that going in. Uh, so our system is a
50:58hybrid. I won't quite go into the
51:00details of where we sort of use
51:01different things, but essentially we've
51:03got a solution generation pipeline. Some
51:06parts of that solution generation
51:07pipeline are driven by machine learning.
51:10And you can sort of think of this as a
51:13sort of like a a higher order
51:16coordinator, let's say, that identifies
51:19uh sort of the highle strategy for
51:21routing. Whereas the actual details of
51:24how that route is executed is done using
51:27more sort of traditional uh mathematical
51:30optimization and computational geometry
51:33approaches so that we can get to the um
51:37consistency and accuracy that we need to
51:39for for this application.
51:43Okay. And then um Tim
51:47like areas below ring water leaders for
51:49the roof drainage.
51:52Tim I think I might need a bit more
51:54context on on that question.
51:58And then uh where do we upload the model
52:01data to? So the depending on sort of how
52:04you mean that question um what we do is
52:07we extract a limited subset of the site
52:10geometry. This is what we need to be
52:11able to run our system against. So we
52:13are not uploading the entire BIM model
52:15or or the Revit model or anything like
52:18that. The data itself is hosted on AWS
52:21entirely on US servers.
52:25Let me know if that is not what you
52:27meant with that question.
52:29>> Um can I just jump in with my question
52:31as well?
52:32>> Yeah. Um, so I mean many of the these
52:35questions are very technical, but I'm mo
52:37mostly looking at at like the process of
52:39this because like in the beginning you
52:40showed a presentation where you said
52:43okay we it was hard to kind of estimate
52:46because like you said so many hours like
52:48150 modeling hours but like it gets down
52:50to like I don't know like by 60%.
52:52>> But I don't know like how big the
52:53project is or is that a typical project
52:55and so on. So that I missed a bit of
52:57context, but I'm wondering like because
52:59the most time was actually spent. Okay,
53:02preparing is one thing. I understand
53:03that you need to put the boxes
53:05everything you explained right now.
53:07>> Um the generation is another thing that
53:10takes a bit of time. I also understand
53:12that. But then the QAQC part, is there
53:15like a typical checklist that you would
53:16recommend people like what should I be
53:20like careful about? What should I watch
53:22for? What should I check? because at
53:24first glance when you look at the 3D
53:25it's like a okay it looks very nice it
53:27looks amazing but what should I really
53:29pay attention to
53:31>> yeah that's a good question so let me
53:33keep that in mind once I go back to the
53:35solution itself and we can talk through
53:37that
53:38>> um and and we can talk through sort of
53:39what we're doing about that as well from
53:41a product perspective so um okay going
53:44back to the studies I'm going to go back
53:46to my list of studies so let's imagine I
53:48hit generate I hit validate when I do
53:50that that button turns into generate I
53:51hit generate
53:52Uh let me load up a previous solution.
53:54So this one ran in 26 minutes and we uh
53:59let me just jump right to the solutions.
54:00It's basically on the same setup. So
54:02these are a number of solutions that
54:04were generated against that site input
54:06that we just went through. This is that
54:08plot chart I talked about before, right?
54:10We can load in and view those solutions
54:12in plot chart form where we can compare
54:15uh estimated cost and construction time
54:17or any other sort of quantitative uh
54:20factors uh or evaluations of a given
54:23design.
54:26And then otherwise let me just open what
54:29one of these guys. I'll just go number
54:31one. It's going to load it into this
54:33viewer. Let me just make this bigger
54:36that you guys can see what we're looking
54:37at. Okay. So here we've got a 3D view of
54:41the building and I'll just hide the
54:43building so we can see the solution for
54:45a second. So this is the solution that
54:47was generated. Um so you can see here
54:49it's avoiding the keep out areas. It's
54:51drawn to those preferred pathways that
54:53we specified and kind of zooming in. Um
54:57just sort of touching again on the
54:58problems that it's solving. So we think
55:00of this as uh overall routing problems.
55:03So finding the most efficient pathways
55:04from sources destination in a way that
55:07minimizes total degree of band and total
55:10material cost. So which is primarily
55:13driven by the
55:16um length but it's also driven by for
55:18example the size and quantity. So it's
55:21more expensive to run a lot of or sorry
55:24to have a lot of um large conduit and
55:28introduce a lot of bends in large
55:29conduit than is smaller conduit.
55:32we are hitting in to the top of the
55:35equipment. We can drop out to the bottom
55:37of equipment as well. But that's sort of
55:40the high level overview of the solution.
55:41Okay. So what we can do now though is
55:43start to talk about some of the details
55:45at a quality level. And there's a number
55:47of things that we're working on here. Um
55:49so from a QAQC perspective, let me see
55:51here.
55:53You'll see things like this where this
55:55sort of bend just doesn't make sense,
55:57right? like it really should have gone
55:58out the front and uh and merged into the
56:01rack there. Uh let's see what else.
56:04There's
56:06going to be areas where it'll still sort
56:08of produce these slightly less than
56:10optimal sort of decisions around for
56:11example how it's ordering on the panels.
56:14So we generally find that there are a
56:15lot of these like it's still sort of
56:17these localized
56:19um things that you need to look at. It
56:21really is probably the way to think
56:23about it is that you're looking at it as
56:25if you had a junior modeler prepare this
56:28model for you and you now need to go
56:30ahead and review it. And really a lot of
56:33these things are what we're focused on
56:34improving. Um and where where that it's
56:39a primary it's sort of the primary
56:40factor that has led to that rate of time
56:43savings improvement or that productivity
56:45improvement has just been a very
56:47concerted and sort of systematic effort
56:50on the part of our engineering team to
56:52reduce these issues. Uh here we see
56:54cases where for some reason ASP could
56:56not figure out how to place that band.
56:58Right? So we call these an unknown part.
57:00Like we know that there's a connection
57:01here. Um, this doesn't look particularly
57:03complicated probably for a human being
57:05to solve, but there's just something
57:06about it where some constraint could be
57:08met and we couldn't quite figure out how
57:10to place that part. Um, the last thing
57:12that you'll kind of see is maybe maybe I
57:14can find some here.
57:18Sometimes we get some self collisions as
57:20well within generally within racks.
57:23Let me see if I can find any.
57:30Maybe there's something here.
57:36sort of funny to be
57:39going through a demo trying to find
57:40what's wrong, but I do want, you know,
57:42we're very open about where the system
57:44is.
57:48Um, well, you'll have to trust me.
57:49Sometimes we do get self collisions that
57:51you'll have to fix. So, what we've done
57:53though to help people do that is, you
57:57know, at this point, we've got this
57:58analysis tool. when you're happy with
58:00the solution. Actually, before I go jump
58:02into that, I just want to touch very
58:03quickly on being able to compare because
58:04I think this is where it gets useful
58:06from a value engineering perspective.
58:08So, I can load in other studies. Let me
58:10make this a bit more visible or
58:12solutions from other studies and I can
58:13compare them against each other. Right?
58:15So, um here we've got some very
58:17different pathways being selected. So,
58:19even within a generation, the system is
58:21exploring different ways of solving that
58:23problem. So it is going to explo uh you
58:25know different um well it's going to
58:27find different ways of of sort of
58:30optimizing for those different goals. So
58:32sometimes you'll have solutions that
58:33have less install time but greater
58:36length because of the materials sort of
58:38involved and so on. And where that's
58:41particularly useful is okay I've got a
58:43big coordination change let me evaluate
58:45the impact of that coordination change.
58:46What does that mean from a routing
58:47perspective but also from a cost and
58:49time perspective? uh I'm at the
58:51engineering stage. I want to evaluate
58:53the uh impact of a potential design
58:56decision that I'm making. Where should
58:57the electrical room go? Where should the
58:59riser go? You know, those big sort of
59:01design decisions that today generally
59:03are made through intuition and and
59:06expertise. Uh now you can prove it to
59:08yourself uh with a lot more confidence
59:11just by running those different setups
59:12in parallel and then going and comparing
59:14them against each other. So I can
59:16compare visually. I can also compare the
59:18statistics of those different solutions
59:20to really understand the impact of one
59:22over the other. We think that's really
59:24quite powerful. So
59:28once I've let's say landed on that
59:29solution, I would go export export to
59:31Revit as system conduit. Um that'll take
59:34a couple minutes, but the output will
59:36basically just look like what we saw
59:37here, right? So this is an outputed uh
59:40uh design out of ACP. So what we're
59:43doing here is we output in um just a
59:46very generic uh we use system conduit
59:49for the straight sticks and we use a
59:51very very generic conduit fitting for
59:53all the bends. We have built a tool that
59:56allows you to convert all those parts to
59:58whatever you want. So if you are using
1:00:01evolve, you're using uh bend works,
1:00:03you're using syscq, you can just open
1:00:06this up and it will convert all of uh
1:00:09our parts to whatever part you want. and
1:00:12we just map or you you help it map the
1:00:14parameters that you want to map and
1:00:15generally we can convert an entire site.
1:00:17Again, we've run entire data center
1:00:19sites in a couple minutes and got all
1:00:21those parts converted. So, it's really
1:00:22really fast process and we've done that
1:00:24intentionally because we want to be very
1:00:26agnostic with the designs that we
1:00:28output. We know people use different
1:00:30systems. We are just trying to get that
1:00:32conduit populated for you. We're trying
1:00:34to get all that data into that conduit
1:00:36in a way that's useful for you. And then
1:00:38it's really up to you to use whatever
1:00:40workflow you want to go and do that
1:00:41final sort of assessment. But to aid
1:00:44with that assessment, we have built some
1:00:46tooling to help folks do that QAQC. So
1:00:49if I putselves in the shoes of someone
1:00:52using the system, I've just gone from
1:00:54zero to one. I need to make sure that um
1:00:57well, we would we would identify for you
1:00:59if something wasn't routed. So you know
1:01:01that everything that was in that
1:01:03schedule is being routed, but I still
1:01:05want to verify that, right? So we've
1:01:07built tooling here that allows you to
1:01:09visually analyze the output based on
1:01:12different parameters or different
1:01:14factors. Right? So here I can color
1:01:16those sources and destinations by the
1:01:19source. So all the conduits leaving this
1:01:22panel are going to be this color and so
1:01:24on. And that really allows me to trace
1:01:26down those designs to make sure that
1:01:29everything is routing where I need it
1:01:30to. This is extremely useful uh in
1:01:32practice. We can sort by destination. We
1:01:35can sort by, for example, system. I
1:01:38don't know how many systems are in this
1:01:39project. Okay, we've got a bunch of
1:01:40systems. So, we can sort by system.
1:01:43And just in practice, the way this works
1:01:45is that um folks will review the
1:01:48solution. They'll spend some time
1:01:50reviewing the solution, walking through
1:01:52each um you know, generally folks have a
1:01:55a favorite of what they're going to
1:01:56filter by. So, that could be, for
1:01:57example, by source, it could be by run
1:01:59ID. Those are kind of the most common.
1:02:01Uh we make it easy for people to isolate
1:02:03these. So they can go and isolate and
1:02:06that's only going to show there it is
1:02:08the run that belongs to for example that
1:02:11source
1:02:14undo that
1:02:18and even from editing process so this
1:02:21site is a little bit cleaner but let's
1:02:22say I did have um some of those okay so
1:02:26maybe there there are things here okay
1:02:28here we go here's the self collision oh
1:02:30no it's not a self collision
1:02:33um Well, here's an issue that we saw
1:02:35before where we've got the bend sort of
1:02:37wrapping sort of the long way around.
1:02:39So, something I can do, you know, is
1:02:41really I can just focus down just
1:02:44isolate the conduits that belong to this
1:02:47panel. That's LTP UB1.1
1:02:55isolate by that. Make those edits,
1:02:57unhide that, and go and and look again
1:02:59at sort of uh each individual rack. So
1:03:01in cases where people are doing that
1:03:03review, people have different methods.
1:03:05So you can do it by source. A lot of
1:03:07folks will go by destination and they're
1:03:09just checking, am I happy with the
1:03:11layout of these racks? Am I happy with
1:03:13um are are in fact my parallel runs
1:03:15being run parallel with each other? So I
1:03:18think a lot of this is simply that even
1:03:20if we were producing perfect designs,
1:03:22there's always going to be some amount
1:03:24of time dedicated to that QA QC process
1:03:27simply because you are going from zero
1:03:29to one. And that's going to increase
1:03:31linearly as the complexity of that
1:03:33project increases because
1:03:36again I'd maybe gone from nothing to 400
1:03:40runs all in one shot. I'm seeing all of
1:03:43them for the first time. So, I know like
1:03:46I we can guarantee that someone is going
1:03:48to want to work through each of those,
1:03:50verify that they're happy with it before
1:03:52it's something that they're comfortable
1:03:54bringing into bringing into
1:03:56coordination.
1:04:00All right.
1:04:04So, with that, uh, let me go through a
1:04:07couple more questions.
1:04:10All right, Tim.
1:04:13>> Mhm. Yeah, I meant to recognize areas
1:04:15needed to keep electrical equipment and
1:04:17panels away from areas where plumbing
1:04:19passes above. Oh, the equipment itself.
1:04:23So, we don't help with that. We don't
1:04:25currently help with anything around
1:04:26equipment placement. We really just
1:04:28handle the raceway itself. So, ASP will
1:04:31avoid the those systems. Um, but we're
1:04:36not sort of flagging for a user right
1:04:37now that hey, you've put a panel under
1:04:39here or the plumbing needs to move
1:04:41because it's running over there. Um, and
1:04:44that's in some ways sort of related to
1:04:45your next question. Does it work with
1:04:47codes like the NEC and plumbing codes?
1:04:49So, we did origin design. Well, we did
1:04:51design the system with NEC in mind,
1:04:52right? We're not demoing plumbing right
1:04:55now. So, when as we build out plumbing,
1:04:57it uh will be considering the plumbing
1:04:59codes and uh mechanical codes in
1:05:02general.
1:05:04From our perspective, the because right
1:05:06now we're solving primarily for routing
1:05:08and not so much for the actual
1:05:12engineering decisions. We're not doing
1:05:14wire sizing. We're not quite doing
1:05:16grouping yet though we are working on
1:05:17that. We are working on some of these
1:05:19these these factors. But right now in
1:05:21terms of what's commercially available,
1:05:23it really is primarily focused on the
1:05:25routing and the coordination process. So
1:05:27we do have considerations uh codebased
1:05:30considerations for example around
1:05:31supportability but really the biggest
1:05:33controls are going to be those
1:05:34clearances that I mentioned and those
1:05:36tend to be very project specific. So
1:05:38we're really relying on the customer to
1:05:41know what's right and what the
1:05:43requirements are for their particular
1:05:45project.
1:05:46Sanjay do we use complex bends like kick
1:05:4990s or rolling offsets to reduce degree
1:05:51of turns? do uh we do actually use some
1:05:53very complex bends. So we handle kick9s.
1:05:56We do occasionally use rolling offsets
1:05:58though we try to avoid them. Uh I can
1:06:01see if I can find any here. Let me see
1:06:03here.
1:06:06If I could have loaded up a more complex
1:06:08project for you guys like a data center
1:06:10which I unfortunately can't do for
1:06:12confidentiality reasons and I think we'd
1:06:14see more of those complex bends.
1:06:17But um of course it's going to try and
1:06:19use simple bends. it's going to try and
1:06:21use standard angles wherever it can. But
1:06:24um let me see here.
1:06:30In short, we do handle those um and we
1:06:32sort of have a hierarchy where we try
1:06:34and handle it with the simplest solution
1:06:37possible, but we in cases where that's
1:06:39not sufficient, we will resort to more
1:06:41complex bends and kicks and offsets.
1:06:45And then Marie, uh, would you be able to
1:06:48use it completely on your own or is it
1:06:49safer to use augment support from you?
1:06:51So, we don't really have an option right
1:06:53now to use it on your own in the sense
1:06:56of sourcing it locally if that's what
1:06:58you meant from a technical perspective.
1:06:59But if you mean as an enduser
1:07:01perspective, uh, this is built for you
1:07:03to use, right? So, we are we're not a
1:07:05BIM consultancy. We are a software
1:07:07builder. Uh, all of our customers are
1:07:09driving it themselves. We do have an
1:07:11awesome support team. I promise not to
1:07:13be salesy in this in this uh session. So
1:07:16I won't spend too much time on that but
1:07:17we have an awesome awesome support team
1:07:19all XBA managers and modelers
1:07:20themselves. Uh we get very hands-on with
1:07:23our customers when when they need to. So
1:07:25you'll be very well supported there. Um
1:07:27and that's generally our model right
1:07:28now. So we don't from a technical
1:07:30perspective there's no current model for
1:07:33um uh onrem hosting. So it's all going
1:07:37to be cloud-based and running off of our
1:07:38our servers. Let me know if that wasn't
1:07:41quite answering your question.
1:07:45>> Okay, awesome, Eron. That's uh Yeah, you
1:07:47covered a lot. Um, do you still have
1:07:50anything to to go through or I mean, we
1:07:53covered a lot. So,
1:07:54>> yeah, we covered a lot. I think we kind
1:07:55of hit all the key points. Um, no, I
1:07:58think I think that's about it. I think
1:08:00we've covered enough that everyone
1:08:01should have a general idea of of what it
1:08:03can and can't do. Um, and of course,
1:08:05happy to answer any other questions that
1:08:06you guys have. Okay, I have a question
1:08:10um on and this is a probably a tough one
1:08:13because like most people that I believe
1:08:15are watching aren't using Augmenta yet.
1:08:18That's why they're here. They want to
1:08:20learn about it and so on.
1:08:22>> And we had uh we have had Andra on the
1:08:25show
1:08:26>> um a few months ago
1:08:28>> and now we heard about like everybody
1:08:30heard about Conigi, you know, but they
1:08:32got about out. Okay. Sure. So you can't
1:08:33even access it anymore probably. I mean
1:08:35at this point. Yeah.
1:08:37>> Um, and so how do you handle this
1:08:40question of like what app to choose if
1:08:43you're interested in getting all of
1:08:44this? Like I mean MEP is everybody's
1:08:47dream, right? The whole right now you're
1:08:49doing electrical and there's doing fire
1:08:51but they're moving to electrical.
1:08:52>> So how do you approach this question of
1:08:55competition? Why should somebody go with
1:08:56you and not with your competitors?
1:08:59>> It comes down to what level of detail
1:09:03and certainty you are working at. We are
1:09:06the only system that outputs designs
1:09:09that contractors are actually using to
1:09:12build
1:09:14>> and that is the difference. So uh I love
1:09:17what Andrew is doing and if you look at
1:09:19their website it says for engineers we
1:09:21are building this for contractors and
1:09:23that is the fundamental difference. So
1:09:24everyone on this call who works on the
1:09:27trade size knows that the output from an
1:09:30engineering process is a design intent
1:09:33and it always gets reworked to make it
1:09:35constructable.
1:09:37Especially on the electrical side you're
1:09:39getting highle 2D drawings. You're
1:09:41getting some roughly modeled in feeder
1:09:44conduit that has no details considered.
1:09:47Every electrical contractor I've been
1:09:49talking to them for years. I've never
1:09:50met a single one that's using that
1:09:51output and actually building off of
1:09:53that. Like all of it gets thrown out and
1:09:55they're all remodeling it themselves. So
1:09:57we are working backwards by building
1:09:59systems that output actually
1:10:01constructable actually coordinated
1:10:03designs that follow the code and
1:10:05constructability rules that you can um
1:10:08actually execute on. And that is a very
1:10:10very hard problem. And really our unique
1:10:13angle is that we're building these tools
1:10:15that embody the experience of the of a
1:10:19foreman from every one of those trades.
1:10:21Our electrical tool build systems that
1:10:24output designs that a foreman would sign
1:10:26off on. Same for mechanical, same for
1:10:29plumbing. Um I know that the other tools
1:10:33do not do that. And ultimate and I and
1:10:35the thing is that there are real
1:10:37problems to solve. There's a lot of
1:10:38opportunity on the engineering side. I
1:10:40think that even without that level of
1:10:42detail, you can still provide a lot of
1:10:44useful information to an engineer and
1:10:47maybe you don't quite need that level of
1:10:49detail uh for them to do their work. But
1:10:52fundamentally, what we're trying to do
1:10:54is we want to shave months off of
1:10:57construction projects and the risk and
1:11:00the delay and all the waste comes during
1:11:02construction. It comes during well
1:11:04pre-construction and construction.
1:11:06That's what we're tackling. So to
1:11:08summarize, I would say that for a
1:11:10contractor,
1:11:12I love what those other tools are doing,
1:11:14but they're they're not contractor
1:11:15tools. Those are engineer tools. They're
1:11:17simply just not going to output designs
1:11:19that um are any better than what you got
1:11:22before and are just not going to be
1:11:24something that produces a useful output
1:11:26for you. Uh we will do that and that's
1:11:28that's our focus. For folks who are more
1:11:31on the engineering side, ultimately what
1:11:33we're doing is we're taking the
1:11:35expertise of again those foremen from
1:11:37the different trades and we're moving
1:11:38that upstream. So we already work with
1:11:41engineers who um are using our output
1:11:44because it gets them actual coordinated
1:11:46output that they can give to their trade
1:11:48partners and have high confidence that
1:11:50their trade partner will build it that
1:11:51way. And I think that's really really
1:11:54unique and that's sort of where we've
1:11:56decided to focus. And the reality of
1:11:58that is that this is a very hard
1:12:00problem. It's much harder to solve that
1:12:04problem and to get that level of
1:12:07accuracy in 3D than it is to, for
1:12:09example, work in a 2D context, which is
1:12:13generally where a lot of those tools
1:12:14operate, or in a simplified uh 3D
1:12:17context where there's really not that
1:12:18much you actually have to coordinate
1:12:19around and the output is it's okay if
1:12:22it's approximate. Those are
1:12:23fundamentally different problems. which
1:12:24are both very valuable to solve, but
1:12:26they're for different people in
1:12:27different stages of the process. So, um
1:12:30hopefully that sort of gives you some
1:12:31that's sort of like how we would sort of
1:12:33philosophically separate ourselves and
1:12:35and practically it's that you know um if
1:12:38you want to build then you we're the
1:12:40ones to work with. Um, so I mean
1:12:43focusing on contractors, I mean I
1:12:44understand if it's a turnkey contractor,
1:12:46then it's like everything's going on
1:12:47within that sphere, but but if you if
1:12:50it's a contractor that gets hired later
1:12:52after the all the design started and
1:12:54people are do you sometimes maybe
1:12:58experience that Augmenta is getting
1:13:01um implemented too late in the process?
1:13:05not for electrical because with
1:13:07electrical the electrical engineers
1:13:09generally don't model a whole lot at all
1:13:12um very little and the reason they don't
1:13:14do that is that generally the routing of
1:13:16an electrical system does not affect the
1:13:18performance of that system whereas it
1:13:19does for mechanical plumbing so that's
1:13:21that's an important difference and I
1:13:22think that even for us as we move into
1:13:25mechanical plumbing our product and our
1:13:27target user may be different as a result
1:13:29of that it's going to be more engineer
1:13:31focused because that's really where the
1:13:32value is so fundamentally when it comes
1:13:35to a actual the actual process of
1:13:38figuring out how am I actually going to
1:13:40build a particular system in electrical
1:13:42falls heavily on the contractor whereas
1:13:44it's more distributed in mechanical and
1:13:46plumbing I would say um so in practice I
1:13:49wouldn't I wouldn't say so you know
1:13:51every electrical contractor even on the
1:13:53most complex projects they're starting
1:13:54with a model that has nothing in it like
1:13:56I they do not have a model conduit that
1:13:59they're using so they are implementing
1:14:01it at the right time I you know again
1:14:04when we work with our customers who are
1:14:06engineers. I think that they're using it
1:14:08the right way. They're using it to
1:14:09inform those decisions, really decide
1:14:11where things need to go. And
1:14:13fundamentally, if you're looking to save
1:14:15significant amount of money on your
1:14:18construction projects, it's really going
1:14:19to be found in those early stage design
1:14:21decisions. So if you can use something
1:14:24like this to bring insight into that as
1:14:27in literally just understand what's it
1:14:29going to cost if I do X for Y if I put
1:14:32this lecture room here or lay it out
1:14:33this way or that way um that's really
1:14:36the right place to use this. So I would
1:14:38say that philosophically yes this should
1:14:40be used as early as it can be in the
1:14:42construction process but practically
1:14:44we've never actually run into a
1:14:45situation where a user has been sort of
1:14:47too constrained using this because in
1:14:50practice there's there's very little for
1:14:51them to start with.
1:14:53>> All right. Um we have a couple of more
1:14:54questions but before we get to that I
1:14:56want to just clarify something about
1:14:58Revit and so on. So um I mean this is
1:15:01working within Revit as you've shown.
1:15:04Are there any other tools that you are
1:15:07planning to open within? I mean not that
1:15:09it would make strategic sense so to
1:15:12speak because verity is king after all
1:15:14but u I mean you you can still cover
1:15:16more example like you know you can work
1:15:18with maybe with speckle you can work do
1:15:20you work with IFC how does it all work
1:15:22when it comes to like platforms? Yeah.
1:15:25So, um, our strategy has always been to
1:15:27be really agnostic. Um, we right now we
1:15:30work with Revit because of the obvious
1:15:32reason, right? Everyone works with
1:15:33Revit, at least here in in, uh, in North
1:15:35America. Um, like I mentioned at the
1:15:38start, there's not a whole lot of logic,
1:15:40business logic in this in these buttons
1:15:42up here, right? Everything is is in our
1:15:44system. We we've built our own geometry
1:15:46kernel. Our data model is not based
1:15:48whatsoever on Revit. So, we could
1:15:50integrate into any system. uh we could
1:15:53integrate into other CAD systems when we
1:15:54move to other markets and um in that
1:15:58sense we've been
1:16:00that solves a lot of problems for us
1:16:02because everyone integrates into Revit
1:16:04here. So as long as we integrate into
1:16:06Revit and we push back out in a very
1:16:08clean Revit format that doesn't impose
1:16:11you know a particular structure to the
1:16:12data then out of the box we sort of work
1:16:15really well with with everybody. Uh IFC
1:16:17if you can open it up in Revit it will
1:16:19work with with ACP. Um, so again there
1:16:23it doesn't need to be a native Revit uh
1:16:25file. We work read IFC just fine. Um,
1:16:29>> what about Revit families? I remember
1:16:31because like you don't know but I I
1:16:33worked for SECO for three years. Yeah.
1:16:35>> And one of the like a few months I spent
1:16:37actually modeling electrical stuff.
1:16:39>> I mean so I remember some some of it.
1:16:41Um, and what I remember is that we had
1:16:43um issues with some families that were
1:16:453D, some were just 2D because we didn't
1:16:47have we didn't have a need like the
1:16:49deliverable was a drawing. So, how does
1:16:52ACP work with um Revit families? Do you
1:16:55need to have it all built out? Does it
1:16:57have its own thing that you can bring
1:16:58in? Does it work with all what's already
1:17:01in the template?
1:17:02>> We so we have we have we provide basic
1:17:05family for our bands and things like
1:17:07that. Uh but that's very very minimal.
1:17:10>> Um otherwise we have we need the site to
1:17:14be 3D and you need to have modeled your
1:17:16electrical equipment. Um the good news
1:17:18is that especially in the US everyone
1:17:20has on the contractor side. So we we
1:17:21never run into problems where there
1:17:23isn't really enough fidelity in the
1:17:24data. Um so we we do have sort of very
1:17:28basic minimum requirements around the
1:17:30families what's in them as in they just
1:17:32need to be properly categorized. We need
1:17:34a panel name parameter. We need one
1:17:36electrical connector in there.
1:17:38>> That's it. So it's it's really quite
1:17:40light. Um
1:17:41>> okay.
1:17:42>> So it's yeah very very minimal
1:17:43requirements on that front.
1:17:46>> All right. We have a couple of questions
1:17:47again. Marie and Tim, thank you for for
1:17:49being so active.
1:17:50>> You guys are awesome.
1:17:52>> Yeah.
1:17:52>> In Germany, contractors are such high
1:17:54demand. They don't care about working
1:17:55with superior software. They just do as
1:17:58they always did. How's your experiences
1:17:59of contractors in Canada, the US?
1:18:02Um, okay. It pains me to say this as a
1:18:05Canadian, but uh the Canadian experience
1:18:08feels feels similar to your German
1:18:10experience, whereas it's less so in the
1:18:12in the US. I would say that in the US um
1:18:16we found that
1:18:18there's a very high rate of BIM
1:18:21adoption. There's generally been a lot
1:18:23of interest in um in new software. I
1:18:26think that a big driver for that is the
1:18:28labor shortage. So even though the
1:18:30contractors are yes in high demand,
1:18:32there's a lot of projects and the
1:18:36capacity to deliver on those projects is
1:18:38not matched by the size of the teams.
1:18:39Basically, everyone's trying to hire BIM
1:18:41moders. No one can find them. They're
1:18:43very expensive. Everyone's outsourcing
1:18:45more and more. So, there's a lot of
1:18:46internal pressure to try uh and and and
1:18:49sort of innovate. In Canada, I would say
1:18:51it feels a little bit quieter. We find
1:18:53that firms here are
1:18:55just not so far along in terms of BIM
1:18:59and technology adoption. Again, it does
1:19:01pain me to say that. Um so, I imagine
1:19:03it's a similar dynamic in in Germany
1:19:06because you're you're right. Like end of
1:19:07the day, I think that um I mean this is
1:19:11kind of going off track. I think that a
1:19:12big driver for BIM adoption in the US
1:19:14was the 2008 crisis. Uh we were hit by
1:19:18that here, but not as much for, you
1:19:21know, a multitude of reasons. And there
1:19:24was just less pressure, I think, for
1:19:26firms to sort of innovate or die the way
1:19:28there was in the US. And I think that
1:19:30was sort of a
1:19:32in retrospect a healthy push for for
1:19:34that industry. Um
1:19:37so yeah uh Tim can a smaller contractor
1:19:41use this or does they need to be
1:19:43sophisticated with full in-house Revit
1:19:44capability?
1:19:46So I think that okay we definitely have
1:19:49small contractors using this. We have
1:19:51contractors as small as one or two
1:19:54modelers but you do need a modeler right
1:19:56so I think that in practice you can see
1:19:59that it does sort of require uh well it
1:20:02integrates in the rabbit there's some
1:20:03basic level of rabbit knowledge required
1:20:05you need to be working in 3D you
1:20:08probably want to be working on projects
1:20:09where the other trades themselves are
1:20:11also coordinating and you're actually
1:20:13getting something to coordinate around
1:20:15though we've had folks use this on um
1:20:19projects where they would coordinate
1:20:20around the engineer models for the other
1:20:22trades and because they could get their
1:20:24stuff modeled so quickly they could get
1:20:25it fabbed and get design sign off on
1:20:27that so they can kind of beat the other
1:20:29trades which they weren't able to do
1:20:30before but I think fundamentally um
1:20:33there is sort of a baseline level of
1:20:36sophistication which isn't very high but
1:20:37you do need to have Revit in house and
1:20:39and know how to use it if not you know
1:20:41we we do work with um third parties
1:20:44right third parties are very much a
1:20:46great uh user of or sort of target user
1:20:50of something like this as well. Um, so
1:20:52if you don't do that in house, we can
1:20:54certainly work with your third party and
1:20:56they can use this to get you that same
1:20:58benefit faster.
1:21:00>> All right. Um, awesome. Thank you, Erin.
1:21:03Listen, you're one of my guests that is
1:21:06so well spoken and so detailed in his
1:21:08answer that I have basically an easy job
1:21:11today. So, thank you so much for that.
1:21:13>> Um, listen, I have three more questions
1:21:15on my end. Um,
1:21:18>> pretty much. So please, anybody in the
1:21:20audience that still wants something
1:21:21clarified, please ask. Uh, but I I guess
1:21:24the first question is, um, how do people
1:21:27get started working with this? Okay,
1:21:28they saw the demo, they like what
1:21:30they've seen, they need to just get
1:21:31started testing it on on a pilot
1:21:33project. What's the best way? Do they go
1:21:35to our website, reach out to you? How
1:21:38does it work?
1:21:38>> Yeah, just go to our website, reach out
1:21:40to us. We've got a form on the site. Um,
1:21:43we'll reach out to you. we can give you
1:21:45and your team a fuller demo so folks who
1:21:47haven't seen it you know can get caught
1:21:49up to speed and then we have different
1:21:50options uh if piloting is really
1:21:52important for you we're happy to support
1:21:54you guys on the pilot we like to work on
1:21:56live projects so that you can actually
1:21:58see the value and we work very closely
1:22:00with you through that whole process to
1:22:02make sure you're supported so we're
1:22:03extremely hands-on especially during
1:22:04pilots so you'll be in very good hands
1:22:07um through that process
1:22:09>> okay I just for folks on LinkedIn I just
1:22:12posted the link it's augment
1:22:14aai the link is and um I want to know
1:22:18what is your financial runway like? Is
1:22:20it an early stage thing or like are you
1:22:22comfortable right now growing or like
1:22:24how does it look like?
1:22:25>> Yeah, so we've been around for a while
1:22:27now. We've been operating for about 5
1:22:29years. Um turns out it's a very hard
1:22:30problem to solve. We are VC backed.
1:22:33We're very solidly funded. Uh we have
1:22:36customers. We have revenue. We're very
1:22:39comfortable into uh certainly late next
1:22:41year and we're going into our series A
1:22:43later this year.
1:22:45>> M okay. Exciting stuff ahead, man. Yeah.
1:22:48>> Uh how big is the team?
1:22:50>> We are a little bit over 50 people right
1:22:52now.
1:22:52>> Oh wow. Wow.
1:22:54>> Yeah,
1:22:55>> that's impressive. Wow.
1:22:56>> We're not kids in a garage. No, it's a
1:22:58>> Sorry,
1:22:59>> we're not kids in a garage. [laughter]
1:23:02>> No. Yeah. Maybe like probably like in 10
1:23:05years you're gonna have like a Steve
1:23:06Jobs, Steve Wnjak story. Like it's just
1:23:09the two of us started at Altoesk. We
1:23:11didn't get what we wanted so we needed
1:23:12to kick out on our own. So [laughter]
1:23:14>> that's right.
1:23:15>> Yeah. Um Okay. Um roadmap. Uh you
1:23:19mentioned of course you want to expand.
1:23:20Um but so let me let me ask you when do
1:23:23you expect to expand and is the first
1:23:25focus to nail this down first? I mean
1:23:28it's already working well
1:23:30>> but there's always like you know it can
1:23:32be faster, it can do this, it can do
1:23:33that. Um
1:23:35>> what's the focus right now for 2026 and
1:23:37beyond?
1:23:38>> Sure. So we think about expansion in two
1:23:40dimensions or a few dimensions. Um let's
1:23:44start off with electrical. Within
1:23:45electrical right now we over we
1:23:48primarily focus on overhead conduit. So
1:23:50we sort of think about how do we help
1:23:53people save more and more time and
1:23:55therefore let's target the activities
1:23:57that take the most time. The next step
1:23:59is undergrounds. Underground takes a lot
1:24:01of time to model. We do sort of like I
1:24:03mentioned before have sort of partial
1:24:05support for that but that's really big
1:24:06on for us on the road map. That's
1:24:08absolutely something we're working on
1:24:09this year because that would just add a
1:24:11ton of value. Uh and some some projects
1:24:14are only underground or effectively only
1:24:16underground. So we think about let's how
1:24:18do we expand the range of activities
1:24:20that we help accelerate so underground
1:24:23but then there's also opportunities for
1:24:24us to move upstream. So branch planning
1:24:27um the grouping so all those things that
1:24:29happen maybe at the very start of a
1:24:31project where you've just been awarded
1:24:32that project and it is sort of that
1:24:34initial sort of planning. So kind of
1:24:35left and right the sort of how we think
1:24:37about it. How do we do more scope and
1:24:39how do we start to move upstream to help
1:24:40people with with that initial planning
1:24:42that sets the inputs ultimately results
1:24:44in the inputs that you would feed into
1:24:46ACP. Um but in parallel to that we're
1:24:49working on mechanical and plumbing this
1:24:50year. So if anyone on the call is
1:24:52interested in those domains or has a
1:24:54background in those domains we are
1:24:55building out our network of design
1:24:57partners for those trades. So that means
1:24:59that we would work with you. You guys
1:25:00would be beta testing it really
1:25:02co-designing that future workflow
1:25:04together. So that's still early stage.
1:25:06We're going to be working on that
1:25:08through this year. Uh so it's something
1:25:09that our design partners will have
1:25:10access to later this year, but not
1:25:12something we'll probably be releasing
1:25:14commercially quite yet. Um so really for
1:25:16us, our primary focus is electrical. You
1:25:18know, it's still early days. Um trust
1:25:21for us is really really important in the
1:25:22industry. So we're going to be
1:25:24hyperfocused on not letting up the gas
1:25:27uh the slightest bit. There's a lot of
1:25:29work still to do. better solutions, more
1:25:32scope, uh more sort of helping with more
1:25:36and more activities, planning activities
1:25:38within electrical. So, sort of at a high
1:25:41level, that's sort of how we think about
1:25:42it. I'd say mostly electrical and then
1:25:44started to do our initial strategic work
1:25:46around mechanical and plumbing.
1:25:49>> Okay. Tim has a question. Uh will it
1:25:52interface with the civil 3D model?
1:25:54>> Uh not right now, but it's something
1:25:55that we would build out if we got a lot
1:25:57of demand for it.
1:26:00Okay. So, everybody who wants that
1:26:02demand just like I don't know put thumbs
1:26:04up or Yeah. reach out. Yes. Awesome. Um
1:26:07listen, my last question is uh basically
1:26:10like you know you've been I mean for me
1:26:12like I guess the whole journey of like
1:26:14you starting without the desk you're
1:26:16meeting Franchesca there you guys kick
1:26:18kick kick it out on your own because you
1:26:20realize okay we can have the freedom we
1:26:22can build what we want. Uh maybe we have
1:26:24the context so we can get the venture
1:26:25back. I don't know the whole story of
1:26:27course I'm just guessing here
1:26:28>> but it's um I think that that's a very
1:26:31interesting part and now you're at 50
1:26:32people which is very impressive and
1:26:34you're building the future so to speak
1:26:36so I think that's awesome and you're
1:26:38actually one of the first tools that uh
1:26:39ended up on my list ACI hub one of the
1:26:42first ones so that's awesome but in all
1:26:44of this in the whole journey what are
1:26:46you the proudest of
1:26:50>> I think that there's a couple things so
1:26:55There's there's two I want to highlight.
1:26:57I think one is that just very
1:26:59practically we've made a lot of very
1:27:01real progress on really the hardest
1:27:04problem to solve in in construction. Um
1:27:08there are probably other problems I'm
1:27:09not aware of in completely different
1:27:10domains within construction, but really
1:27:13doing 3D design is really really hard at
1:27:17this level of detail. It turned out to
1:27:19be a lot harder than we thought it would
1:27:21be when we started this company. Um,
1:27:23it's one of those things where you asked
1:27:24yourself, would I have done that if I
1:27:25knew how hard it would be? Um, don't ask
1:27:28me that. But, um, we we've now cracked
1:27:32it, right? We're now at a point where
1:27:33people are actually building real
1:27:34buildings with this. And that was really
1:27:36really hard to do. We've gone through so
1:27:39much work, so much thinking, uh, as a
1:27:42team. I'm extremely proud of the team
1:27:45that we've built, just really brilliant
1:27:47people that it took to build this. And
1:27:50um you know I think that's that's really
1:27:51notable. It really feels like what we're
1:27:53building here is infrastructure. It's
1:27:55not something that you can vibe code.
1:27:56It's not something you can just sort of
1:27:58build quickly or or spin up a proof of
1:28:00concept around really quickly. It really
1:28:02takes years of work to get to this
1:28:03level. Um so I think that's something
1:28:07that of course we're we're going to be
1:28:08proud of because that's sort of our our
1:28:10jewel. The other side though is the
1:28:12relationships that we've built in
1:28:13construction. I think that we've been
1:28:16able to I think build a really good
1:28:18reputation for just listening a lot. You
1:28:20know, I think I think particularly
1:28:21because Franchesco and I didn't come
1:28:23from construction, we knew that we we
1:28:25knew that we didn't really know anything
1:28:28and so we had to listen. So, we didn't
1:28:30really come in with firm ideas or
1:28:33bulldo.
1:28:34Um, we've been we've got to where we are
1:28:36because we basically co-designed this
1:28:39thing with a number of of design
1:28:41partners and it's something that we'll
1:28:43make sure to not lose track of as we
1:28:45grow and and move into those other
1:28:47industries because it's really I think
1:28:48quite precious and particularly in
1:28:50construction where these relationships
1:28:51matter. Trust matters. These things take
1:28:53years to build. Buildings take years to
1:28:55build. So does great software and um you
1:28:58know you need that trust um all the way
1:29:01through. So
1:29:03>> awesome. I think that that's an awesome
1:29:05answer, man. And you kind of reinstated
1:29:08my like my the whole journey we're
1:29:10experiencing currently with Neo Stack.
1:29:12So, it's um
1:29:13>> because they know it's like you you
1:29:15started to like, yeah, this is going to
1:29:16be easy. We're going to build it in a
1:29:18few months and then like it's always
1:29:19much harder than you expect.
1:29:21>> Yep.
1:29:22>> But then you you need to appreciate the
1:29:23people who appreciate you first, right?
1:29:25And early and so
1:29:26>> yeah.
1:29:27>> Um
1:29:28>> Yeah. Awesome. Let's good I mean good
1:29:31presentation, awesome demo. Um,
1:29:33>> thank you. Thank you for having me. It's
1:29:35been great.
1:29:36>> Yeah, you're like you're you're quiet,
1:29:38you know, but when asked, you go on and
1:29:41it's like go in depth. It's it's a
1:29:42really Yeah, you're really good at this.
1:29:44So, this is awesome.
1:29:45>> Thank you.
1:29:45>> Um, yes. Do you have anything at the end
1:29:48you want to mention you want to talk
1:29:51about or
1:29:52>> Well, I I guess I want to finish off
1:29:54with a reaction to your last statement,
1:29:56not about me, but about um your
1:29:58experience with this is that I always
1:30:00thought that the very first customers
1:30:03who even worked with us were just as um
1:30:08bold and admirable as any person
1:30:10building a startup. Like it really takes
1:30:13that vision, you know, I think to to
1:30:15look at something and say, "Okay, if I
1:30:17squint my eyes, I know you're just
1:30:19showing me a deck. You're just showing
1:30:20me a demo.
1:30:22>> I'll give this a try." And it's sort of
1:30:24interesting that I think especially in
1:30:26construction, it's it's hard, you know,
1:30:28you know, it's hard to sort of build
1:30:29that momentum up in this industry. It
1:30:31always takes time. And so, you really
1:30:34need those people. You really need those
1:30:35companies. And uh so this is just a
1:30:38general thank you to you know those
1:30:40folks who really started out on this
1:30:42with us years ago. We wouldn't be here
1:30:44without you. Um so thank you all if you
1:30:47happen to be on the call and if not we
1:30:49are uh looking for the same partners on
1:30:51mechanical plumbing you know. Um it's a
1:30:54it's a big journey and we're looking to
1:30:56find some partners to do it with. M yeah
1:30:59we we we have uh three of those right
1:31:01now on on for neo stack and it's just
1:31:04like the people there like you know
1:31:05they're they're fighting for you within
1:31:06their firm like you know
1:31:08>> um and they are willing to pay you
1:31:09without without you having a a feasible
1:31:11product even like you know like you know
1:31:14it's always buggy it's always like you
1:31:16know whatever but they're they're there
1:31:18with you they're giving feedback they
1:31:19really believe in you as a person and
1:31:21then the product as well
1:31:23>> and of course you want to pay that off
1:31:24as time goes on but this is it's it's an
1:31:26amazing part of this industry, finding
1:31:29these people and having these
1:31:30relationships. So
1:31:32>> that's
1:31:33>> awesome, man. Listen, uh thank you. Hey,
1:31:35thank you not only for being on the
1:31:36show, but thank you for making the AC a
1:31:38better place. I think
1:31:40>> we're doing our best.
1:31:42>> Yeah. Yeah. I think that people forget
1:31:43about this about entrepreneurs. Like we
1:31:46typically focus on, hey, was the success
1:31:48you generated for yourself? But in the
1:31:50end of the day, it's like if you weren't
1:31:52in this world, how worse off would it
1:31:56be? You understand? like you you you do
1:31:57something like from like to change it
1:32:00for the better and I think people
1:32:02sometimes forget that. So, thank you for
1:32:04doing that for me to see.
1:32:05>> Um,
1:32:06>> awesome. Listen, Erin, you can stay in
1:32:08the back stage for a second if you want
1:32:10to chat after.
1:32:11>> Um, I'm going to remove remove you now
1:32:12and just mention a few things for the
1:32:14guys on the call.
1:32:16>> All right. Bye, everyone. Thank you,
1:32:17Stefan, for having me. Um, hope to talk
1:32:20to you guys soon. Take care.
1:32:21>> Awesome. Yes. Great. Um, yes guys, that
1:32:25was Erin. Um, and we hope to have him
1:32:27back probably when they build some
1:32:29mechanical or plumbing at some point.
1:32:31We'll see how fast that's going to
1:32:32happen. And, uh, for everybody still
1:32:34here, again, I want to mention the
1:32:36courses. Uh, you can join them because
1:32:38we're changing to something really
1:32:39exciting, a membership. Um, so you can
1:32:42join them here and you're going to be a
1:32:44part of that um, in the future. But I
1:32:46also want to mention what who we're
1:32:47going to have next week on Wednesday.
1:32:50So, Wednesday, same time, we're going to
1:32:51have Ichi. Um, this is more for the
1:32:55architectural side, but they claim to
1:32:57have a like a lot of stuff within the
1:33:00app. So, anything from code compliance
1:33:02to automations to combining your files,
1:33:05you know, there. So, I'm also learning I
1:33:07mean also looking forward to learning
1:33:09more about that. Um, and I'm just going
1:33:11to like send this link to you on
1:33:14Streamyard and uh YouTube. You can see
1:33:18it and then uh you can sign up. If if
1:33:20you're on LinkedIn, you can just go to
1:33:21my profile. The event is published, so
1:33:23you can just click attend there. Um,
1:33:26yeah, I mean, this was this is it uh
1:33:28from us uh this week. Thank you so much
1:33:31for the patience. Apologies again for
1:33:33those of you who joined the first one,
1:33:34so it was kind of rescheduled for today,
1:33:37but here we are. It was an awesome demo.
1:33:39And yeah, for any questions, go to
1:33:41augmenta.ai.
1:33:43And um yeah, that's pretty much it,
1:33:45guys. Listen, God bless you all. Have a
1:33:48great day, evening. Um, and yeah, see
1:33:51you next week then. Bye-bye.

About this transcript

This page contains the full transcript of Augmenta – The AI That Turns Weeks of Electrical Modeling into Hours by Stjepan Mikulić, generated from the public captions YouTube serves with the video. The transcript has 18,010 words across 2,590 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 translate the transcript into 100+ languages.

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, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.