YT YouTube2Text

YouTube2Text → Transcripts → no one wants to hear this, but the claude code era is coming to an end..

no one wants to hear this, but the claude code era is coming to an end.. — Transcript

by Nick Puru | AI Automation · 5,105 words · 718 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00A year ago, I told you to stop learning
0:01N8N. Now that video's been watched
0:03almost a quarter of a million times.
0:04Well, the same thing is now true of
0:06Claude and honestly of whatever tool is
0:08going to be coming after Claude. Now,
0:10Claude isn't going anywhere or Chat GPT
0:13for that matter. So, if you've been
0:14spending this whole year just trying to
0:15keep up with every new Claude feature,
0:17this video is for you. I'm going to be
0:19teaching you what the next step actually
0:20is, whether you're automating your own
0:22business or selling this to other
0:24businesses. Let's get into it. So, to
0:25start, let's go back a few years. If you
0:27wanted two applications inside of a
0:28business to actually talk to each other,
0:30you typically used one of these three
0:33different tools right here: Zapier,
0:35make.com, or N8N. You typically just
0:37have a blank canvas. You just drag the
0:39box onto it for each application, and
0:41then you just simply wired them
0:43together. So, maybe a lead fills out a
0:44form, it lands in the CRM, and the sales
0:46rep gets a message. That's a workflow.
0:48Honestly, getting one of those to run
0:50reliably was real technical work where
0:52you needed credentials for every
0:54application. You need to be mapping
0:55every field, handling all of the errors,
0:57and testing it until it stopped
0:59breaking. That's what businesses were
1:01paying for. Now, N8N, as we all know,
1:03got really popular and for a long time
1:04it was most of this channel. Four of my
1:06five like watched videos have been N8N
1:09inside of the title. Four of my top five
1:11watched videos have N8N in the title.
1:13And then a little bit later, the tools
1:15started doing that part themselves. Like
1:16this summer, N8N put out the N8N
1:18assistant. You just describe the
1:20workflow in plain words and it builds
1:21it, it tests it, and it fixes the steps
1:23that actually failed. And it's still
1:24labeled a preview, but Make and Zapier
1:26have their own versions of this as well.
1:27So, building the workflow by hand
1:29stopped being the hard part. And by
1:31then, a lot of people had already moved
1:32on to these coding agents, and the big
1:34one, as we all know it, is Claude Code.
1:36Now, Claude Code, if you're unfamiliar
1:38and you live under a rock, this is just
1:39an agent that works directly on your
1:41files where you just give it the job in
1:43plain words, and it reads it, it writes
1:45it, and it runs the things. It checks
1:46its own outputs, and it keeps going
1:48until the job is actually done. So,
1:50there's no workflow to be building here.
1:51You just hand over the work. And this
1:53got really popular really fast. In the
1:55first week of January, like more people
1:57were searching for Claude code than for
1:59N8N and it's been ahead every week since
2:01that point. So, I moved over as well.
2:03Majority people have as well. Now, I
2:05went from teaching N8N to teaching
2:07Claude code and the Claude videos on
2:08this channel have been watched over
2:10600,000 times just this year. But, more
2:12importantly, this a little bit later
2:13moved out of the terminal because
2:15earlier this year Anthropic put the same
2:17agent inside of the regular Claude
2:19desktop application and they just called
2:20it Code Work. And on the team and
2:22enterprise plans, an owner can just
2:23simply upload a skill which is a written
2:25set of instructions for just one job and
2:27everybody in the company, they have
2:28access to that right away. And it's not
2:30only Claude, either. By the way, like by
2:32June Open AI said that Codex had passed
2:345 million people using it every week and
2:36about one in five of them aren't even a
2:38developer. I'm not a developer myself. I
2:40hardly consider myself to be a
2:41developer. So, operating the tool used
2:43to take somebody technical, but now it
2:45just takes somebody who can actually
2:46type a sentence on a keyboard. Now,
2:48really quick, look at my channel. In the
2:49first half of the last year, I posted 61
2:52full-length videos and 33 of them were
2:53about N8N. This year, I posted close to
2:56150 and over 70 of them were just
2:58Claude-related videos. And to that, I do
3:00want to say that N8N didn't fall off
3:02because in that same stretch, the
3:03company more than doubled in value from
3:05two and a half billion dollars to over
3:06five billion. The tool, it keeps
3:08winning. It just wasn't the thing to
3:10actually be teaching anymore. And
3:11honestly, I don't think this stops with
3:12Claude, either. The CEO of Airbnb told a
3:15room of investors in September that they
3:17are going to be more than 50 companies
3:19just like Anthropic, maybe 100. And I
3:21can't tell you if that number is exactly
3:23right, but if operating one tool is your
3:26whole skill, you start over every time
3:28the tool changes. I did it with Make and
3:31then with N8N and then with Claude. The
3:33only thing that came with me each time
3:35was what I knew about how a business
3:37actually runs. By the way, we actually
3:39made a completely free course on how to
3:40grow your own AI business inside of the
3:43link down below in the description. It's
3:44completely free. You can sign up there
3:46to learn how to actually start and grow
3:48a profitable AI business just like we
3:50have done at Reprise. And if you're
3:51business owner watching this and you
3:52want to actually start using AI in your
3:54business, you can book in a call with
3:55our team or at the very least learn
3:57more. Link will be down below in the
3:59description. So, to kind of tie all
4:00these things together, if it's not the
4:01tool, what is a business actually going
4:03to be paying for? Well, it's broken down
4:05into one of four different things. One,
4:08it's going to be time back. And just to
4:09pull an example out of a hat, home
4:10service company, they told us they were
4:11spending about 80 hours a month just on
4:13quotes and job preparation. The second
4:15one is going to be lower cost. 10-person
4:18accounting firm, they were spending
4:19about 25 to 35 hours a week just typing
4:22numbers into QuickBooks. Number three is
4:24more revenue. Dermatology group that we
4:26actually worked with in the past, they
4:27had about 40 locations. They were just
4:29missing calls every single day. That's
4:31revenue being lost. And then number
4:32four, having an edge. So, something that
4:34it can offer that your competitor down
4:36the street cannot. And we're all seeing
4:38that the motes nowadays and what is
4:40actually non-replaceable is being
4:42dropped to the floor. Now, the owner
4:44themselves typically won't use those
4:45words. They'll say like, "Can you
4:46automate our customer service?" And what
4:48they mean is like, "We answer the same
4:50kind of questions all day, every day.
4:52That's time." Or maybe they'll be saying
4:54like, "We need better lead tracking."
4:55Which means that leads are falling
4:56through somewhere. That, in that case,
4:58is going to be revenue. So, part of the
5:00job is just working out which of the
5:01four it really is. And what we've
5:03actually seen is most people are selling
5:05time because it's the easiest one to be
5:06seen. But a business will pay a lot more
5:08to stop losing customers than to save an
5:10hour on a simple report. And just to
5:12show you, I'm not guessing at those
5:14numbers. My company, Reprise, we went
5:15back through every single request that
5:17we've ever received, about 1,800 of
5:19them. And when you narrow that down to
5:20business owners describing what they
5:22actually wanted, it's about 360. And the
5:24four most common things were chasing
5:27follow-ups, 152 times. Weekly reports,
5:30127.
5:32The inbox and phones, 92. And onboarding
5:34new clients, this is about 91 inquiries.
5:36And there's actually one more number in
5:38there. We had 192 teams filled out their
5:41goals list on our growth plan form. You
5:43guys can see that in the link in the
5:44description, but 68 of them, so about
5:47one in three, ticked they are not sure
5:49yet. Show me where AI helps the most.
5:51Those are teams that are already wanting
5:53AI and they just couldn't say where it
5:55should be going and that is a majority
5:57of businesses from what we have found.
5:59And it's the same exact case outside of
6:01our own leads that came in. Goldman
6:02Sachs, they surveyed about 1,250
6:05small business owners from its own
6:06program this year. Three out of four
6:08already use AI, but only 14% say it's
6:10fully built into the core operations.
6:12So, the tool, it's already in the
6:14building. What's missing is somebody who
6:16can take it from, you know, like we use
6:18this sometimes to the business actually
6:20runs effectively on this. And there's no
6:22shortage of people who know the tool. I
6:25think we all know that at this point.
6:26Almost half of our inquiry forms, 637
6:30out of about 1,300, came from people who
6:32just wanted to build or resell AI
6:34themselves. So, what I did, I put
6:36everybody in AI right now, me included,
6:38into three different groups. We have the
6:40tourist, they just collect tools. They
6:42know every feature the day that it
6:44ships. And they haven't shipped really
6:46anything themselves that's really worth
6:48a damn. The builder, they get something
6:50working, but just for themselves. So,
6:52maybe this is a system that's going to
6:54be sorting their inbox perhaps or maybe
6:56it writes the reports. Nobody's paying
6:58for it. And then the operator, they have
7:00put a system inside of a real business.
7:02The business pays for the result and it
7:04just keeps running when the operator
7:05isn't in the room. And I get why people
7:07stay as a tourist, this first step right
7:09here. Nobody ever turns you down for
7:12watching one more tutorial and a
7:13business owner definitely can. And the
7:15builder's trap is thinking that the next
7:17build is the one that actually gets them
7:19paid. But AI has already made the build
7:21part incredibly cheap and easy to do. An
7:24operator, they actually get paid for
7:26knowing which build a business will
7:27actually be paying for and for just like
7:30standing behind this. Now, just a quick
7:32check, take your last seven days and
7:34just count the hours that you have spent
7:36learning or building and the messages
7:38that you have actually sent out to
7:40business owners. If that second number
7:42is close to zero, you know which one you
7:44are, and that's how you determine which
7:46path you actually fall under. So, that
7:48is what I would learn instead of Claude.
7:50How to get from this tourist into
7:52becoming this operator. And it shouldn't
7:53be to your surprise that it's really
7:55only one thing that you need to be
7:56doing. And before I get into what that
7:58actually is, I want to cover the five
7:59levels of work. So, what does an
8:01operator do that a tourist doesn't?
8:04Well, level one, this is just going to
8:06be operating the tool. That's the level
8:08that the N N assistant
8:10fundamentally took over. Level two, this
8:13is connecting different systems. So, the
8:15agents, they plug into most
8:17applications, APIs directly now, but not
8:20all of them, I will say. But, for
8:21example, we just built a phone agent for
8:23a radiology clinic whose patient record
8:25systems was locked with no way in. We
8:27couldn't access their PMS system. So,
8:29the bookings, they go into a calendar,
8:31and the staff copies them across every
8:33morning. Somebody still had to make that
8:35call. So, there is a little bit of a
8:36human in the loop. Anyways, level three
8:37is designing the fix once somebody hands
8:40you the problem. actually built a test
8:41for this, and it's just real business
8:43tasks across different applications with
8:45nobody having to help the AI. When that
8:47test actually came out this spring, the
8:48best model passed about one in 10. And
8:50right now, the best one passes about
8:52half of that. So, they've tightened the
8:54test since then, so it's not exactly
8:56like for like, but it's a big jump. And
8:58when the models failed, most of the
8:59time, they still reported the task as
9:01done. So, level three, it's going fast,
9:04but it isn't gone. Now, level four, this
9:06is finding the real problem, which
9:08usually isn't the one that the owner
9:09tells you about. five is changing how
9:11the business actually runs once you have
9:13found that real leak. Now, the closest
9:15test for that, it just hands the AI
9:16whole freelance projects that somebody
9:18really got paid to do. And the best
9:19model right now, it just turns in work
9:21that a client would accept at about one
9:23in five. And the demand, it's moving the
9:25same way. When on Upwork this June, AI
9:28applications and integration, this was
9:30the fastest growing category for new
9:31contracts inside of the states. So, the
9:33bottom of that pyramid is what the tools
9:36are actually taking, and then the top is
9:38where the work is effectively going
9:40right now. So, to give you an example,
9:41here's level four on a past project. So,
9:43we had a property management company
9:44about 400 units. I use this example
9:46quite often, but there was 12 people on
9:48staff. They told us that their problem
9:50was in tenant messages when we were
9:51actually talking to leadership. So, we
9:53do the same questions over and over
9:55again about 15 hours a week for each
9:57manager from what we were finding. And a
9:59bot that answers tenant questions isn't
10:01easy build and what would seem like the
10:03next step for them. However, when we go
10:05down to the bottom of the ladder, going
10:06down the chain of command, and we're
10:08talking to the people who actually do
10:09this work, we're looking at how the
10:11place actually ran, and the bigger
10:12problem was new leads. So, people asking
10:14about renting a unit. Now, these were
10:16coming in from, you know, range of
10:18different sources, and nobody was
10:19qualifying them. And reply, it took them
10:22about 3 to 4 hours, and about 30% of
10:24them were just simply being lost. So,
10:26that was what got fixed first, and then
10:28after the build, a new lead got an
10:29answer in literally 18 seconds on
10:31average, and 97% they heard back in
10:34under 30. That's measured 8 months in.
10:36And then the tenant messages, these did
10:38get handled as well from about 15 hours
10:40a week for each manager down to under
10:42five. So, my point with all of this is
10:44it's rarely where the owner actually
10:46thinks it is. We audited a supplement
10:48brand that was sure that their ads were
10:49the problem, but the ads, they were
10:51actually completely fine, but only about
10:5222% of people were finishing the
10:54checkout when a normal store is
10:55somewhere about 40 and 60. Now, the
10:57express checkout buttons, they were just
10:58switched off in the cart with 98% of the
11:01traffic on their phones, and the
11:02checkout pages, they said that shipping
11:04takes 15 to 20 business days. So, it was
11:06just a toggle in a shipping label. That
11:08is it. Knowing every feature inside of
11:11Claude would not have found either of
11:12those. What found them was following
11:14just one lead all the way through the
11:16business, where it comes in, who's
11:18actually seeing this first, how long
11:20until somebody replies, and where it
11:21goes cold. So, just to tie that all
11:23together, the leaders, they think that
11:25they need to be automating something to
11:26go ahead and automate it or hire
11:27somebody to do that project, but it's
11:30not actually generating any returns.
11:31Nobody's using it. If we go back to
11:33those statistics of only about 14%
11:35actually have an AI strategy and that is
11:37actually working for them effectively
11:39inside of the operation. Okay, so once
11:41you have actually found a problem, where
11:43does the answer live? Well, it's not
11:45going to be inside of the model. Every
11:47business, they run the same models. I've
11:48got the same Claude that a
11:49billion-dollar company has and so do
11:51you. But what's different is what the
11:53business knows. So how they actually
11:55quote a job, which customers are getting
11:56called back first, what the owner does
11:58when a client is unhappy. And in most
12:00businesses, that's sitting in two or
12:02three people's heads. So the job is just
12:04to get this into one folder of plain
12:06files. Now the work itself, a rules file
12:09for how you want it done, skills for the
12:11repeatable jobs, and connections to the
12:13applications that it is allowed to be
12:15using. Now once this is written down
12:17like that, any agent can be working from
12:20this and it's one of the most
12:21fundamental steps. And to add to that,
12:23these big companies are already doing
12:24this. The CEO of Airbnb, he said in
12:26September that he built what he calls a
12:28brain. It's trained on about 50,000 of
12:31the company's documents and thousands of
12:32its slides and that he has cut his own
12:35meetings in half. One of the presidents
12:36of Salesforce, the one who runs their
12:38platform, said that the same month that
12:40the models are going to keep changing
12:41and intelligence is going to be
12:43available everywhere. Now what sets a
12:45company apart is the context that it
12:47actually brings to it. We actually built
12:48a small version of this for a surf camp
12:50in Costa Rica where they wanted a voice
12:51agent, but their guests, they were also
12:53asking on WhatsApp, on the website chat,
12:56and over email. So they had a bunch of
12:57different channels. So we built one
12:59knowledge base that all four read from.
13:01So when a price changes, they change it
13:03in just simply one place. And this is
13:05actually how I run my company as well.
13:06We have one folder with a cloud.md file
13:08and an agents.md file sitting right next
13:11to each other. Claude reads the first
13:13one, Codex reads the second one, and
13:14they say the same thing. I'm not to
13:16scare you, but the big companies,
13:17they're much further along on all of
13:19this and you can actually read what
13:20they've been doing. McKinsey's survey
13:22this year, they found that four in 10
13:24companies over a billion dollars in
13:25revenue are actually scaling AI agents
13:27in at least one part of their business.
13:28Now, last year, that was only about a
13:30quarter and under that line, it's about
13:32two in 10 and it did not move. And they
13:34themselves, like they're not betting on
13:35one tool. In one survey this summer of
13:37about 100 mostly large companies, the
13:39average was three agent platforms
13:41running at the same time. On top of
13:42that, Salesforce, they put out a
13:44blueprint for running all of that and it
13:46has six different parts right here. So,
13:48we have the context, which is what the
13:49company actually knows. We then have
13:51agency, so how the agent plans the work,
13:54and then we have action, the
13:55applications that this is going to be
13:57connected to, governance, the policies
13:59and security that it actually follows,
14:01security, what it's allowed to see, what
14:03it's allowed to do, and then we have the
14:05model. So, the model is one part out of
14:08six different things. A small business,
14:11they have the model and almost nothing
14:13from these other five different pieces.
14:15And we actually built two different
14:16systems for an accounting firm whose
14:17contract said that the financial data
14:19simply just could not touch anybody
14:20else's cloud. So, the whole thing, it
14:22runs on their server. One line in a
14:24contract is just fundamentally deciding
14:27how it got built and no model was ever
14:29going to be knowing that unless we had
14:31that context. And that gap between
14:33having the tool and having the rest of
14:35it, this is the actual work. And closing
14:37it for one business is something that
14:39you can learn and something that you can
14:40be selling. So, how do you actually
14:42price this? How do we price this? Well,
14:44most people price their own hours, I
14:45will say, and then later down the line,
14:47the faster they build, the less that
14:49they are getting paid, which is
14:51inevitable. So, I worked it out in two
14:52different ways. The first is cost. Let's
14:55say a task takes somebody about 15 hours
14:57a week and that person costs $30 an
14:59hour. Now, over 52 weeks, that's $23,400
15:04a year. Add 15 to 12% for benefits and
15:06overhead and the manager's time and it's
15:08around 27 and a half thousand. And then
15:11add the revenue that they're actually
15:12losing while it stays broken. For the
15:14property manager, that was every lease
15:16that they were not signing. Now, the
15:17second one is actually capacity. It's
15:19usually the bigger number. To me you an
15:20example of what this looks like, another
15:22ad agency that we're working with, they
15:23had about 80 clients, they had seven
15:25pods of strategists, and no room for any
15:27other clients without hiring. So, we
15:29just simply built one workflow for the
15:30reporting, and it freed up about a
15:31quarter of every one of their
15:33strategists' times. And that is room for
15:35about two more clients per pod. So, if
15:38we look at seven pods, about 14 clients
15:40at $3,000 a month each, that's going to
15:43be $42,000 a month or $504,000 a year
15:46with the same people. And it's only
15:47worth that if they can actually fill the
15:49room, but that's the number that you are
15:50going to be pricing against. And the
15:52build fee usually isn't the whole deal.
15:55Now, on most of our projects going live,
15:56it starts out at a retainer of about
15:58anywhere from like $500 to $1,500 a
16:00month. Often, it can be much, much more
16:02because just somebody has to keep this
16:04running. Now, if you are wondering how
16:05you actually get clients and get a
16:07business to say yes, a stranger isn't
16:09just going to be handing you their whole
16:10business. So, with this, you start with
16:13just one small fix that they can
16:15actually see is working. Now, in the
16:16past, we quoted a firm with over 35
16:18million records in an old database, and
16:20their owner, as he should be, was
16:22nervous about anybody touching that
16:23server. So, phase one, it ran on a test
16:26copy of his data at about $7,000 instead
16:29of 22. And you can actually find that
16:31fix before you ever are talking to them.
16:33So, you fill out the contact form on
16:35their website like a normal customer,
16:37and just start a timer. If it takes them
16:385 hours to get back to you, well, okay,
16:41well, you found the problem. And you can
16:42tell them exactly how long it took, and
16:44that can be your way in, one of many
16:46different ways in. Again, we cover so
16:48many other tips and tricks and
16:49strategies on how to actually grow and
16:51start a profitable AI business. Look me
16:54down below in the description inside of
16:55our school community. Number three, if
16:56you do not have results to show yet, you
16:58can just take the risk off of them with
17:00some sort of guarantee. You needs to be
17:02doing some sort of risk reversal. So, if
17:05the result that you promised doesn't
17:06show up, you carry the cost. Or maybe if
17:08you're doing some sort of lead
17:09generation offer, well, in this case,
17:11they only pay for the meetings that you
17:12bring them. This is going to allow them
17:14to have zero risk, where you take all of
17:16the risk, and it shows them that you
17:18have the confidence in delivering the
17:20result for them. Even still, if they
17:22work with you and nothing happens,
17:24there's nothing that they actually lose,
17:25of course, except for the time. Number
17:27four is promising less than you think
17:28that you will deliver. So, on the
17:30property project example, we told them
17:32that we had a 73% return in the first
17:34year. So, that payback period and it
17:36came in at 123. So, obviously a better
17:39result that we were hoping for. Number
17:40five, sometimes you just have to be
17:42saying no. We actually won a recruiting
17:44platform against two other bidders and
17:46we were just this most expensive of the
17:47three. They wanted us to scrape from
17:49LinkedIn and we refused to do that
17:51because it would have gotten their
17:52account banned. Saying no to that
17:54actually builds trust. And it's not to
17:55say that we didn't end up scraping
17:57LinkedIn, but there is workarounds.
17:58However, their strategy on what they
18:00want to do was what we were saying no
18:02to. Now, pretty fair objection to all of
18:04this and pretty common objection, I will
18:05say, was one of the top comments under
18:07last year's video. It's all great until
18:09something goes wrong and nobody knows
18:11how to be fixing this. And honestly,
18:13that's completely correct and I have
18:15this multiple times over. We had a
18:16clinic where the agent just started
18:18double booking appointments in front of
18:19their staff and their patients. We
18:21nearly lost the client as you would
18:22expect and it came down to just one
18:24setting that was skipping the
18:25availability check. So, just a simple
18:27overlook from my team and we just put a
18:29real check in front of every booking and
18:31gave the staff a switch to take over any
18:33chat and that fix is what allowed us to
18:35keep them. So, the technical side,
18:37obviously it still matters. Learn the
18:39tool until you can judge its work and
18:41find out why it's actually wrong and
18:43then you stop studying it. I still still
18:45use Claude every single day. I just
18:46don't stop at Claude. So, if I was
18:48starting over completely from zero
18:50today, this is what I would do in this
18:52exact order. One is pick one kind of
18:55business and just stay inside of it. So,
18:57maybe property managers, accounting
18:59firms, clinics, whichever one you can
19:01actually get in front of. It takes a
19:03year or two to really learn an industry
19:05if you don't have any of that industry
19:06experience yourself and nobody can be
19:08copying that out of a tutorial and
19:09that's where you can [clears throat] get
19:10your edge. Number two is finding one
19:13problem that you can actually measure
19:14before you ever are speaking to your
19:16prospects. Whether anybody picks up the
19:18after hours on their phones, how long a
19:20quote actually takes to come back, and
19:22then number three, writing down how the
19:24business works before you are building
19:26anything. That's usually on a
19:27conversation with each department head,
19:29and it all just goes in one folder. And
19:30then number four, is just building one
19:32fix for one business and measuring it.
19:34That's it. So, you start with something
19:36that happens every day and actually
19:37touches money or their time, and you
19:40write down the number before and after.
19:42Now, our proposals, they don't say that
19:44the systems work well. They say that it
19:46has to sort at least 90% of a test set
19:49of emails correctly. So, having some
19:51sort of gold standard that you're going
19:53to be measuring against and defining
19:54that success criteria. And then number
19:56five, last but not least, is selling
19:58that same fix to the next business just
20:01like it and pricing it on the result.
20:03After five or 10 of these, what took you
20:05weeks started to take you days. Now, I'm
20:07not going to be sitting here telling you
20:08which of these tool wins next year
20:10because I simply don't know. It honestly
20:12doesn't matter. What I do know is that
20:14these tools that I teach on this
20:16channel, they change twice in a little
20:18over two years. Honestly, way more than
20:19twice, I I can say. Uh whether that's
20:22going from ChatGPT to Hermes, Grok Bot,
20:24whatever it may be. That whole time,
20:26businesses asked my company for the same
20:28things. The follow-ups, the reporting,
20:30the phones, and the onboarding, and
20:32automating all this stuff. And those
20:33problems, they were there long before
20:36any of these tools have came into
20:37existence. So, this week, you need to
20:40just simply be picking one of these
20:41businesses that I've covered or any
20:43other business, finding what actual
20:45business is going to be a little
20:47underserved, whatever it may be. There's
20:48a lot of criteria and a lot of nuances
20:50that you need to be determining for what
20:52niche, what offer you should be having,
20:54and that's exactly what we cover inside
20:55of the school community. But, anyways,
20:57just pick one business, run that contact
20:59form test on them if that's what you
21:01want to be going with. So, for example,
21:03going to a law firm, filling out their
21:04inquiry form, seeing how long it takes
21:06for them to fill out or get back to you
21:08for that point. And then you can pitch
21:10them saying, "Hey, it took you this long
21:11to get back to me. I can only imagine
21:13this is happening to a lot of other
21:14leads." Just one example of many
21:16different things that you could be
21:17offering and getting your foot in the
21:18door to just have these conversations.
21:20But anyways, for those of you guys who
21:22are interested in starting or growing
21:24your own AI business, there's so many
21:25nuanced problems and questions, and that
21:27is what our free school community aims
21:29to solve. So, make sure to check that
21:31out. Link will be down below in the
21:32description. And if you want to see what
21:33this actually looks like inside of real
21:35business, I just uploaded a video
21:36preceding this one covering all of our
21:39learnings, all of our lessons from
21:40implementing AI inside of real
21:42businesses. So, make sure to check that
21:44out if you want to see the before and
21:46afters and all the results and all that
21:47stuff. And if you're business owner just
21:49looking to implement AI this year, beat
21:51out your competitors before they start
21:53using AI, then make sure to book in a
21:55call. Link will be down below in the
21:56description to actually work with us.
21:58We'll hop on a call, see where your
22:00problems are, and how we can help you.
22:01But in any case, thank you guys for
22:03watching, and I'll see you in the next
22:04video.

About this transcript

This page contains the full transcript of no one wants to hear this, but the claude code era is coming to an end.. by Nick Puru | AI Automation, generated from the public captions YouTube serves with the video. The transcript has 5,105 words across 718 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.