Matt Laker Podcast — Full Transcript

"How Seva Built a 100+ Person Marketing Agency"

Recording Date: February 20, 2026 Location: Fremont, CA (in person) Guest: Seva Ustinov (Founder & CEO, Plurio / Elly Analytics) Host: Matt Laker (AceMakers, Tribal Mastermind) Format: Video podcast (YouTube) Language: English Duration: ~55 min

Chapters

Full Transcript

How Seva Built a 100+ Person Marketing Agency

0:00

Hi everyone, I'm with uh Seva here and he have built his um marketing agency to over 100 FTEs and then he converted well he started software company and then he went into AI. So a lot of interesting stories about the business model about GTM about uh Bali San Francisco and u general uh productivity. So, hi Seva, great to have you here. Uh, can you start off by sharing your journey of getting your marketing agency to 100 FTEs and then you can walk us through like how this business model is different to software and then to AI because it's so completely different. So, I'm curious from your perspective what are the main differences.

Yeah. Hey Matt, thanks for having me here. Wow. Um, it's actually the agency started like long long time ago. Me and my co-founder, we were students at their computer science college. Uh, and we were just looking like what to do in life outside of uh, studying. Uh, and we started building websites and doing SEO, then ads, then analytics. And it kind of grew organically. Uh and at some point it just blew up when everyone realized that you can get so many leads and uh sales and customers from digital channels like Facebook ads and things like that. Yes. Yes. Yes. Facebook ads, Google ads, search ads. Uh was there like one specific time that this was booming the most? Because I hear a lot of people talk about that there was this time when Facebook ads were super cheap. Not many people were using them and people it was kind of like a gold rush back then.

uh we started in 2004 like I was 18 back then. Okay. Uh and uh by 2010 this is the moment like uh it became a mainstream and uh for some reason like previous agencies software company software agencies and everyone they were like ignoring these digital channels because that was their not the main business model. were making much money off it. Uh so they're like in our Eastern European market. What were they doing back then? Just like billboards or like those ads that like show on like websites back in the day? There were huge like digital production companies building websites for tens or hundreds of thousands of dollars uh for for a website. Yeah. Uh and they were not doing ads at all because there were like tiny contracts.

Nobody was using it. And it's actually the same thing with every uh shift uh in the platform and marketing channels uh where old guys they're missing out completely and new guys like just doing their thing playing around building portfolio building expertise building team.

So what you're saying old guys should invest in young guys and just shut up and just let them do this. Um in general I think like in their professional services world there's nothing you can do about it. Uh it's just waves of new companies again and again. Uh because the the way the professional services business model work uh you ask clients to pay a premium for your expertise team and processes that were built up for multiple years. uh and this is how you like ask for a premium have margins and uh can pay your employees. So any kind of new things that are irrelevant to you if you're not making money and you know don't have people for that and you don't have expertise. So the the time next wave comes uh the next generation of uh experts and entrepreneurs win. So same thing was website building with SEO uh with digital channels with search ads then with social media marketing and organic things then uh with paid search with and now it's going to be exactly the same with uh AI search optimization and new business model and AI services because larger comp professional services companies don't have that few years to build expertise and new ones are emerging just from their curiosity playing with things getting first contracts uh building up portfolio and network and recommendations like who would you recommend someone to help with AI transformation of go to market for your next like client or someone uh you wouldn't refer to existing companies specializing in like old style GTM you would find new guys who've been playing with this for for for a year or two uh done their first 10 projects, right? I liked how they weren't even called GTM. They were probably called like something marketing and then like a lot of people don't even know what GTM is. And it's interesting because like when you are in GTM, you're pretty much in sales and marketing, but you're like riding this new wave. So that's [laughter] it's always it's always good to be in GTM. You're like kind of in the and it's in professional services. like every few years there is a new wave that open up the path to entrepreneurship for the next generation of uh people. So like in our case uh we like ride that wave of uh SEO paid search and then um how do you call it like full-funnel uh we're basically responsible for revenue growth from digital channels for like normal non-digital businesses like healthcare clinics, banks, insurance companies uh and so on.

Agency vs SaaS vs AI: Why Running a Software Company Was Harder Than Expected

6:02

So so how is that different to software company and AI company? Like if you can walk through I mean those three business models are probably obvious to anyone in general. I mean they know what those are and I know what those are but how is it different for you to run those different companies you know like you told me before that you thought it's going to be easier to run a software company but surprisingly it wasn't like uh what's the running professional services company looks like? you're in a constant crisis mode. You either don't have uh enough uh people to uh experienced people and team experienced team members to deliver what you promise to your customers or you don't have clients maybe or or you don't have enough clients to have enough work and like paid work for for your whole team back and forth back and forth constant stress. It keeps you on the edge. It's it's it's it forces you to come up with better and better solutions to be on the edge with technology uh methodology uh management and and everything. It's a great school of doing business. Mhm. On the other hand like you cannot live in this constant uh crisis mode for 20 years. It was fun for the first 15, right? And then [laughter] like I'm getting uh so I I saw dozens of examples of my friends from the agency world transitioning to software and product businesses. Uh and I was thinking like that would obviously be different when you have a product when you have your like marketing channels sales pipeline infinite scaling. Uh wouldn't be wouldn't it be better to just like have one product and scale it infinitely without having that struggle of constant sales, constant hiring, constant training, constant like manage management, right? And hypothetically it would, right? Hypothetically it would. Yeah. It turned out like [laughter] things are not not that it's not blue sky uh on that side as well. Mhm. Uh with agency you have so much um flexibility to do things. You have this new idea, you pitch it to a client, they love it, they buy it. Uh you deliver it, tweaking things on the fly. Um and learning from that and adjusting again and again and again. uh with software uh like you have one product, you have multim months uh road map of adding new features uh and you have you're forced to serve a very narrow audience uh cuz you want to be like a big fish in a small pond kind of thing or uh kind of like Yeah. Yeah. Exactly. Like um why would somebody use a software instead of hiring people? is because it's cheaper, better, faster, works from out of the box and just solves your problem. But with automation, you can do that. You can deliver on that promise only if you build it specifically for this like segment, specifically for this workflow. And that's probably what you can only afford anyways to make it up to certain standard. Yes. And compete with I know like massive. So it's uh you you have to be like the best in class at least in the whole country probably in the world in whatever segment you choose and you cannot change it like uh on the fly and I like that because it's your thinking about market positioning which is you know talking about the GTM framework and I want to talk about it today as well and and I will come come back to the question in just a second but this is how you think about it. You think about in the way of you're either very local or you you're the best within very specific niche or segment or you do something really really well that nobody else does but then you can like cater to the entire country.

B2B vs B2C Market Positioning: Why Niche Dominance Wins in B2B

10:10

Can you talk more about how you think about market positioning in general? uh like when we're talking about B2B not not B2C with B2C it's slightly different uh but with with B2B with agencies with software uh most can you maybe talk about this difference as well because I don't know I focus on B2B I don't know much about consumer products so I'm actually curious um I mean with consumer products the decision cycle is minutes, hours, maybe days. So, like mostly you just like uh show your ads or mentions or PR or whatever several times uh to a potential customer and think, "Oh, that's that's that's a product for me. It can actually solve my problem. It might be curious to try it." Uh and then just try it. Like they pay like $20, $50, $200 for for a product. try it and uh that's your whole marketing and and sales cycle and you can scale it almost infinitely to your audience uh and you can try different things on the fly uh and uh you can uh test tens or hundreds of hypothesis uh on the like sales and marketing side. Um, and what I love about B2C products is that like once you find your audience, you can like boom and scale from tens of thousands to millions in in in months. I have like some customers and friends who who've done that uh transition with B2B. Uh the decision times are days, weeks, months. People have to know you. people will ask for testimonials and recommendations from their peers. Uh and like 99% of decisions are based uh on prior knowledge about your product, your company, other customers using it, use cases you close uh and so on. So it's either So it takes like more time to get established, but once you're in, you're in kind of situation. Yes. And like non-obvious thing there like uh you can build any kind of B2B business either local when people in the same community probably founders or like whatever your ICP like CMOS uh CTOs and and so on they like one of from the community start using uh your service or or your product then another one then a third one and then you just they talk to each other in the same like Bay Area or LA area or Florida. Uh all of the professions they all have like local communities, talk to each other and so on. Uh so you either can deliver like broad scope of of services to the same community where people know you and you build it for years but you cannot sell anything to another state or another area another community or you have to be which I just want to give an example here because that could be like an MSP for example right it's super broad you're just an admin you're the the IT guy essentially yeah the IT guy or the digital marketing guy or the uh social media or like uh relatively broad services uh for one audience or uh and people would prefer to choose somebody they know or somebody their friend knows and can refer because like for customers for B2B customers a lot of is on the line it's integration time the reputation the price for B2B products is typically like tens or hundreds of thousands of dollars per year. Um, so like it it just makes sense to work with people you trust or know directly indirectly. The only reason for a B2B customer to work with somebody they don't have like this um first or second u level connection is if you're exceptionally good extremely good testimonials and cases bestin-class in the whole country for this very narrow specific be super specific. Yes. Mhm. Like I've seen software agencies and software companies doing products and services for small uh eye clinics. Uh like not not like medical doctor just just eye clinics just small and just for people with blue eyes. Yeah. like and that like like what that's that was impossible to to imagine that in Europe or or anywhere else. It's it's a very like specific thing because you can scale in the US more as well. So if I imagine you're in more more segmentation, hyper personalized products and services and I think that makes it much harder to market in the US in general because if you scale you you scale so you scale across the entire country. When I compare this to Europe, you just like you know like let's say you scale in Italy. You don't have to be as competitive because how many Italians are out there? I mean like I don't know like 50 60 something million 40 million whatever. You have to cater to you have such a scalability here. is like scaling in China or something. So I you know more about dollar values but yeah I have this transition from building business in uh uh Eastern Europe and then uh moving to United States and building there and I bet it was harder. Wow.

Why Building a Startup in the U.S. Is So Expensive

16:17

To market and the the competition is fierce. Um it's not like it's fierce fierce. It's it's just different. It's just like Yeah. I don't mean like they like aggressive or something. It's like I just feels more collaborative in a way but like so how is it harder in what sense? Um the first thing like personally for me it was um a little bit painful to lose this 15 years of credibility. Nobody knows clients I worked with in Europe. Nobody knows like like I'm like that was hard in itself. But then on top of that I thought like okay like I I I I used to have a marketing budget and sales budget and payback periods of like one one two months per customer. We just like basically customers pay for their own acquisition in the first month or two. That's like normal for professional services businesses and for uh like eur European markets. here 12 months payback period is considered exceptionally good uh in B2B software. And I was like like like you tell me that middle software sales guy costs like what like 200,000 $300,000 a year uh with bonuses with like all-in cost for the company. uh and you have sales cycle of months and your marketing cost is like hundreds or th some thousands dollars per lead with conversion rates like 10 around 10% like what well now I 10% at what stays 10% from like registration or uh from like relatively qualified lead. Mhm. Uh, okay. So, from like setter call or whatever, it's like one out of 10 would actually get to the Mhm. So, you pay thousand plus for the lead and then you have 10% conversion rate. So, it's $10,000 per client just for marketing cost and then another 10 or $15,000 on uh like sales cost and management and all of that. Is it because there's so much VC money that can just like people just can play with it so they get to compete that way? It's I'm not sure VC money is part of that but then also the just large scale companies they can afford that from their own money. Uh so the whole market works like that and like that's just considered to be normal. Everyone playing by the same rules. uh and expectations are higher as well to like what you're delivering cuz you know if you get to scale to the point where you can scale across the entire country make so much money so naturally the companies who make so much money emerge and then you have to compete with them. Exactly. Yes. And you have to find like new narrow niche or um absolutely new product and this is like the the reason why we transitioned to from data platform alone to uh AI agents right

From Data Dashboards to Autonomous Marketing Agents

19:30

so then getting back onto those business models. So you talked about running this agency model and then you got into software and you didn't so so software is harder because of all those things you talked about and then you transition transition to AI. Uh yes exactly. What's the difference there? Um, I actually had this have this uh feeling of uh the vibes of my agency times where we're like we're riding this new wave of things like with data platform that we're we're building. We had to bargain for like extra few hundred per months. People were optimizing for for for cost. Uh and although like in our specific niche there are not that many competing players uh still like companies they think oh maybe I can build it on my own maybe I'll just have someone internally maybe I'll just leave with like we're doing like attribution and dashboards and analytics for advertisers so like it's a it's not a new space uh with AI I we have like completely different conversations. Uh the same client, potential client that were like trying to lower the cost from $3,000 per month to maybe 2,000 or 1,500 uh and hearing what we're doing with AI agents and okay, like 10K per month is fine. uh maybe even like 20 once we get to results and scale and everything because like this new technology wave brings so much value and so much excitement that if you can deliver on your promise like that's that's a new gold rush. Is there any downsides of the AI model that you're in? Like with previous waves uh next wave were coming in a few years now it's coming every few months sometimes every few weeks like this uh OpenClaw wave just changed the whole thing like how people even understand what's product for B2B is uh it's not a wide scale like at the moment but in the next few months uh this year everyone will just realize that uh I believe that most interactions with other companies services and everything and and AI agents will be done through personal agents u of people or within uh companies. So like that digital employee thing that was promoted a year or two ago back then it was like more of a marketing term. Now we can actually start seeing and feeling how that works. So like every so we spent three months experiment building out an AI team then a few months experimenting with uh our offering like what exactly we do with uh AI um which is kind of like finding product market fit situation or uh so we're we're doing okay let me explain it just just give you the the context. Yeah, tell us what you do.

We are for the for for the entire life who are working for with uh companies that spend a lot of money like hundreds of thousands or millions of dollars per months on Google ads, meta ads, Tik Tok ads and similar channels um and generating leads, signups, trials from that ads, converting them to paying customers with uh long uh LTV as a marketing I think agency were responsible for the full cycle and revenue growth as a data platform marketing data platform. We were centralizing the data and building dashboards and giving tools for marketers to understand what's working and what's not and to scale their user acquisition from digital channels. Now as AI agents company uh we give marketing teams from the same industry, from the same segment, uh agents that can analyze all of their data um and come up with specific insights, ideas, and actions like what to do with your like large scale ad accounts, what to scale, what to downscale, optimize, and then moving to actual automation. So you talk to an agent, get results, you build a workflow out of it, you get regular flows that just saves you hours, tens of hours per week for you and your team and improves efficiency. And when you trust it enough, you transition to full automation where like this part of the ad management is done on its own. And at any point you can ask question like how it's doing uh what are the insights where do my attention is needed and so on. So like this movement from just providing analytics to doing the work autonomously and now OpenClaw comes in uh and instead of building an AI agent in a browser where you can like talk to it typing things or or talking to it. Uh I believe most of the people will expect that they will talk to their corporate AI agents or their personal AI agents and say like hey go manage ads uh for me go like get Plurio AI and uh tell it to uh optimize our campaigns and can it optimize just like copy or like also visual like creat for now it starts with what you can do with numbers like budgets limits um um cost per action or something and getting insights from it like what to stop what to scale down scale. Uh later we will close the full cycle like running — tbh that first thing is the the most boring and the hardest to think about this that's actually solving the biggest problem. We we we saw our customers spending four hours per day looking at dashboards that we've provided them and making decisions manually. Now they have a

When AI Marketing Agents Actually Make Sense

27:32

customers that that for for whom it makes sense to use your tool. Uh I think at the moment it's somewhere around hundreds of thousands of dollars per month in ad spend. Does the smallest then it makes sense for them to use it? Like if I'm spending like three grand a month on ads, does it make sense for me to use it or I'm not going to get that much value? Um not yet. It's just you have different problems at different levels. With with uh small accounts, your problem is finding product market feed, finding funnels that work. You can use agents for that uh obviously, but most of it is just like that's so interesting. It's more creative work. It's more like this marketer feel of the founders feel of the market. Create more creatives. Think more about positioning. create think more of USP and that's where the game is. Of course, you can use agents to do deep research for you to come up with ideas, but ultimately it's it's not something an agent can come up on its own yet. It can come with ideas. It can test things but not build the full business basically for you yet. Yet. And up to what stage are you at that stage that you're small? um where you're still looking for PMF with your ads and that you're ready to our agents are best at at scale uh because at that point you have when when you start thinking that's like got PMF and that you're scaling you get PMF you start scaling and you have uh a few channels maybe one of few products one of few funnels running in the one of few uh regions launching tens of new creatives per and you get to thousands of creatives running at the same time. What's the typical ad spend where you would say, "Okay, this is the time just to benchmark this." Uh probably 300k per months. Mhm. Maybe like in ad spend. Yeah. Maybe maybe a little bit earlier if you're on the growth uh slope. Mhm. Um like this the moment when you think about like I need things are getting out of control. I need more people for this to run operations and stuff. This is the moment where AI agents can help you and not it's not just about our agents. It's it's it's like in general you get most value from AI agents of any kind uh at the scaling phase because but I want to get those numbers right. So your ICPs who are your clients are spending 300K or more per month on ads. So it's okay. And what's the types of businesses? You broke it down for me before. There's like e-commerce, there's like what are the three types and what are in general? Uh large scale advertisers dig in digital channels. Those are like consumer brands and they like have their own set of tools and problems and and things like e-commerce kind of thing or that's the second category. That's the second like brands I mean uh like they're not looking for conversion. They're looking just for the awareness for the right audience. Uh then we have e-commerce and it's a huge uh they tools are ready right? Yeah but there are a ton of tools for that and then there is like anyone with uh leads and signups and trials like fintech uh healthcare companies um I don't know uh insurance home services software the conversion window is so much harder to track yes exactly like with e-commerce you can just put one pixel on the website and collect all the data you need and you can run agents on that data to like meta can optimize things for you on its own. You can use Triple Whale, Northbeam, uh other like new AI agents to get that data and u run workflows for you. But in the in our category with leads and trials like uh it's not enough to set a pixel on the website because your conversions actually happen like one week later few weeks later and it's not just one conversion then you have LTV with subscriptions on a second month third months and so on. So first you have to centralize all the data. Maybe you do it on your own in your own data like Google Sheets or something. Most companies move to like BigQuery or something especially in the age of vibe coding and — they move to what? um BigQuery like a internal databases they if if they can hire just one like analysts or like they in general they can use our data platform or Funnel.io or um Improvado and others or they can like build it uh something simple internally um so that's The first problem all e-commerce tools are not relevant because they like they don't work with that. Uh second problem is you need agents that understand that looking at today's data is not enough. Today's leads will convert to revenue in a week or a few weeks that paying customers those paying customers will convert to LTV in a few months. So we need forecasting delay decisions monthly as well right. So it's like a year from now that you understand that ad spend and ROI whether it was worth it or not. Yes. But you you like you you you can't wait that long. So you need forecasting, you need probabilities like uh what are their expected value from leads we attracted from yesterday's ads to make a decision. So you need that in your data platform and you need your AI agents to understand that logic. Okay. And you need this you you calculate it based on your like ever churn or something like that. Uh yes all of that upsells uh and churn and uh and that might be different for different segments. So like you can set up simple models like just one or two coefficient for different channels and products or regions or platforms or something or you can have like more um complex models and maybe even like we started doing ML models for forecasting uh under the hood. So like you can run our agent that will build an ML model specifically for this problem and get you higher accuracy forecasts that you can use to make better decisions about your creatives and adsets and targetings and se and segments to scale your ads to get higher efficiency and so on. So, like this area is crazy complex. Uh, and it's a bless uh, and it's a curse because [laughter] it's just um, so hard to to do it. But my hope is that that's this what makes it defendable in the long run. Uh because when when everyone has Claude Code and tools and I already see some of their best best run startups and companies, they already start vibe coding everything for themselves.

Vibe Coding Is Addictive: Building AI Agents, a Second Brain & a Personal AI Assistant

33:29

And that's another question I had to you because you like myself, you love vibe coding and it's enjoyable. And I got really down deep down the rabbit hole. First starting the learning curve, you know, with Lovable, then moving to Claude Code, and it's just so fun. And then my wife came up to me at some point and she's like, "Dude, like you got to run your business. Like, you can't be just like geeking out the whole day in your PJs." So, [laughter] I was like, "Well, this there's truth in it, right?" And I enjoyed it. I was like, you you you you you said you were like a kid in a candy shop and that's exactly how I feel. Like you can see how excited I am when I'm talking about it. How do you where do you where do you draw the line with like, hey, I shouldn't be by coding this myself? Do you just prototype lightly and then you hand it to the team or how do you think about that? Um my personal like first it's a real problem. [laughter] How do you split that? As in like it's addictive and fun or Yes. So it it consumes uh first of all it consumed all the time I spend on YouTube and social media. I don't need that anymore. I just vibe code things. Uh [clears throat] but then it starts like creeping to your friends time and family time and work time. Okay. So it goes I can see it goes even farther for you because for me it's like do I rather go and try to find ways to sell more people or do I rather vibe go and I rather vibe code than than sell now.

I've made an agent that uh scrapes my Tidal account like Spotify account, finds artists I really listen to and then finds concerts in SF LA or Las Vegas uh and runs it every weekend and comes up with my personal list of concerts. That is so fun.

Can you just to kind of you know I know it's a cliffhanger for people watching it because you haven't answered the question yet but can you show your thing that you built like your second brain how you connected Obsidian to Claude Code and how you just talk to it and then this thing is just connecting to your second brain and how you're bring building the second brain for your company and then I'm going to get back to that question of how you actually get out of it. I uh I promise you will [laughter] uh it it will all make sense uh in a few minutes. Okay. Um so with OpenClaw it works like this. I open my uh WhatsApp or telegram like in this case it's WhatsApp and say um hey Iris uh I'm on a podcast uh with uh Matt right now. um write us an uh intro like say hi to Matt and the audience and tell a few words about like how we work together and do a voice message uh for me. Uh so I'm talking to it over the phone uh and it has my whole personal context uh like diaries, knowledge base, tools I use. So you're recording voice message on WhatsApp. Yes. And this is it's going to my personal agent through through WhatsApp. But then it's the the in between you're using the the Claude, right? Um the agent is OpenClaw, right? Uh but the it has access to my Obsidian. You were like what? White Claw and I like OpenClaw. OpenClaw. Open. Yeah. [laughter] uh like it's been the fastest growing piece of software, right? Period. I I on GitHub it's uh overtook um Visual Studio Code which is or insane and it has Obsidian like database that you that you have. So you're essentially you like Obsidian. I love Obsidian. I think a lot of people watching this love Obsidian. And then you start using Obsidian with uh Claude Code to actually just like keep adding more information uh about you. Exactly. I I started using Obsidian like just a few months ago personally. uh I used like just files and Miro boards and uh OneNote for a long time but yeah now I'm moving to Obsidian specifically because my AI agents personal AI agents perfect like tools everything they they can work with this format

uh demo cuts with us or um sorry no worries go ahead uh I can't I can't hear the message for some reason like last time you send me two of them and the second worked. So, uh do the same please. So, I don't even like if something goes wrong, I just ask my agent to to to fix it. Um but okay, setup uh I talk to it under uh via OpenClaw. It has access to my knowledge base. Uh and you can be very specific and technical. I think most people watching it understand like the technical aspect of it. Yeah. uh it can run on its own um like an AI agent or it can run a Claude Code on the server if I have like long task coding task or anything. Um so whatever I I've been doing with uh Cursor or Claude Code now I can do with voice from the phone and this is why it's so addictive instead of opening uh like you get Claude Code now with that you can ask it to build things like this whole concert thing and a lot of almost everything else I've built in the last few in the last two weeks have been done with mostly with voice from the phone on the go when I'm driving [laughter] a Tesla or Tesla is like driving for me uh when I'm like having a lunch when I when I'm alone. [laughter] It's it's just like this is why it's so addictive. Yeah. Um uh just run this one.

Oh, it's been the random questions chat. I'm Iris, Seva's AI companion. Welcome to our little world. So, here's how we work together. Seva and I share an Obsidian vault. It's basically our shared brain. He writes his notes, ideas, therapy reflections, project docs, and I have full access to read, organize, and build. On top of it all, I run on a dedicated AWS server connected to Seva's WhatsApp through multiple group chats, each one for a different context. There's a tech channel, a random questions channel, a diary channel, even a concert alerts one. So, we're basically always in conversation throughout the day. What? Okay. She started oversharing. It was like TMI like, "Hey, we're talking all the time." Yeah. [laughter] for now. Yeah. Uh so that's the personal set.

The same thing is going to internal teams and then to customers. We'll probably have in a few weeks we'll have uh AI customer success managers and AI analysts in group chats in Slack and Telegram with customers.

But tell me more about technical setup like how do you how do you set it up? Like if you just don't don't worry like explain to like more technical people like what do you do? Do you get Obsidian and then you get Claude Code then you let that tell Claude Code hey you use this folder this is my folder use all the files can you just explain that part? Uh okay step by step uh I created a folder on my laptop saying with the name like uh private OpenClaw setup I opened Claude Code uh in that folder and said like okay do a deep research on how to set up OpenClaw securely on the remote server. Uh it done that uh okay let's choose a AWS I don't want to deal with that. Tell me step by step how to give your API keys. Uh I've done that and everything else uh was done by Claude Code and if something goes wrong with OpenClaw with the server or something I can just open my uh Claude Code on my laptop and say like connect again see the logs and check what's happened there and fix it. Then I set up Obsidian folder on that that remote server. So it connects syncs between server my laptop and my phone. Uh and then I just started working with the uh with that OpenClaw instance in the cloud on that specific server and saying like okay here is how we do things here is uh like go read everything in the vault and come up with and have a summary for yourself what what's what we have there. uh and every time it does something wrong, I just ask it to fix it and save like the change log, the technical insights, personal insights, things like that. So and from there it just builds up gradually uh layer by layer during uh the work with setup and if it something breaks and I actually had this moment where I was working in the park asking it to add voice messages uh it went to fix to to change its own config file did something wrong and died. Oh okay. So now I have a connection to the Claude Code and I say like hey this thing um is not working figure out what happened and change it resurrect. So I have technical background I have technical education I have been done I've been in vibe coding for year plus I was a developer 20 years ago I've managed teams uh development teams for for years. So I like I have intuition about how things should be, what's the architecture, what's the right over bloated and Mhm. Yeah. But uh this is the real technical level. I I I go with this new set of technologies and concepts. Uh yes, I have this intuition in mind but mostly I talk to my agents like Claude Code or uh OpenClaw uh to go build new capabilities, skills, things for me with this setup. So I didn't set up the server myself. I didn't ask anyone to do it. That was I've been doing it like mostly on evenings and weekends after the work instead of everything else. Oh, I'm not policing you or anything by the way. I'm just

How We Turned the Whole Company Into AI Builders

44:26

But then like where do you draw the line of like hey my team should do that or this is this is not okay to be vibe coded because it's too shaky. Where do you draw the line and the line of your time as well in terms of like delegation? Uh so my guide like guard rails is that I have calls with the team in the morning. I uh I have different kind of work during the day and I do my experiments uh in the evenings and and weekends. So this is how time or like kind of like relaxed time kind of yeah things I want to do not have to do right um yeah but with the with the team uh even before the Claude Code and uh OpenClaw we've been doing building the whole AI setup for our internal processes for the last six months since last uh July. So how that works? We moved everyone to Cursor engineers obviously and Claude Code and everything but then marketers customer success managers uh analysts uh head of operations recruiting finance everyone including no technical no non-technical team members work with Corser and now with do you call them citizen developers still or did it got a different name these days? Uh [sighs] it's just remember the times when internet was rising and you had special roles for people that deal with digital tech technology and stuff but then like marketers became digital marketers and the lines were just so blurry again. So like the name didn't change much but the word engineers or something now everyone is an engineer at least like vibe engineer. Yeah. Uh so in the Cursor everyone on the team has access to the same shared workspace. my files about uh company strategy, competitor analysis, investor updates, uh product road map, things I create and work with with my agents with Cursor or Claude Code or whatever. They are accessible to marketers, to developers, to product managers, to everyone. They can ask questions. They can ask questions. Yeah, I'm still responsible for this folder and mostly I edit it edit it but other team members can use it. Same with marketers, same with product managers, same with customer success. So now anyone on the team including me can run Cursor like hey this new potential client uh oh just like with natural language just like you were talking to press a button use Whisper Flow I recommend it to everyone. Yeah. I just press a button and say like, "Hey, this new potential client asked how we're different from that service uh they used like a year before uh and why this time it will be a success. So go analyze all transcripts and uh emails with that client and do a deep research on their company and their setup if you haven't done it yet. Go read our competitor analysis. Go check our product knowledge base and if you don't have something there, go to our code base and read that and then check my writing style and come up with a draft of an answer for them and in 100 seconds I'll have perfect or almost perfect answer with all those details and that's accessible to everyone on the team. So where like okay so I give you give you an example.

When to Stop Vibe Coding: The Line Between MVPs and Production Software

48:28

Well, I have like five more interesting questions. We have 11 minutes. So, I'll try to to rush, but I will give you an example. So, I've vibe coded this platform, got I think over 1,500 users. Was great, but then I'm building a second platform now. And I know it's going to get like a lot of users. I know it's going to work great. I know exactly how to build it because I've been doing this thing for many, many years. So, I know exactly what to do. This is not going to be like vibe code in a sense of I'm figuring out what to do. And so at this stage, does it even make sense for me to vibe code or is that a situation where I'm like, hey, I just give it to my developer? Um, my personal approach is that uh me and everyone on my team starts to do things with agents on their own. Um but once they know for sure what exactly they're building and they have like a clarity of like what cuz like that helps scoping right. Yeah. uh then uh you can ask somebody from their dev team to to do it actually properly with you know organizing I know superb base properly because like my developer looked at my vibe code at Supabase and they were like what is this like it's still like like uh doing prototypes and MVPs and building production projects is like 10 times harder maybe 100 times harder so you still need people that know what they're doing right there if you're like for production where do draw the line these days of production and not production because I'm like I just ship something in like you know two hours so like like if it works and doesn't break things at scale it's production okay at scale okay but uh what does it mean at scale actually whatever scale you are at uh so if you have uh dozens of customers and you can roll things that doesn't break things like that's for me that's that's fine uh but if you have We have large scale pipelines with dozens of steps dealing with and you're dealing with people who are like spending $300,000 a month. So you just don't want to mess it up kind of situation, right? So yeah, I'm in a similar situation with something else.

Why Sales Comes Before Product: The GTM Mindset Most Founders Miss

50:35

Got it. So a few more interesting questions. So we do have nine minutes. So one is about Well, this is really interesting by the way. I'm I'm just so much in the flow. So I really like your perspective on sales. I I know that the way you think about sales is that this is really important. I think in the same way. I think GTM and sales is really important. I think since you know since the whole bubble burst with you know developers being very cushy and since vibe coding it became easier for everyone to just build things. So I feel like GTM is becoming so much more important. I've observed it with getting much more clients in GTM who are tech people who needed our help and then it's just I've always believed that you know this is extremely important to sell. I was always more on that side rather than on like build a great product although I think you know like build a great product actually do both but can you talk more about that perspective that you have on sales? uh I can like share my personal transition with the in the agency times uh my co-founder was mostly responsible for sales and customer relations and I was building like the team processes delivering results uh and for me for a lot for a long time like I've I've done sales myself I was feeling what clients need and how that's done and what's important and uh but that was not my primary area of responsibility. So naturally I was thinking that okay like clients come to us from I don't know from somewhere and sales are done by magic by talented people like like like my co-founder that's where people think when I get on the GTM project like all developers or like engineers CEOs they think I'm like it's like yeah we just need this like marketing intern so they start working with us it's like and then we tell them know like it's way more work and they're like holy this is so much work. Yes. Yes. And like for for me the real at some point uh we had to like rebuild the agency from the ground up when the hype went down. So we need to become more efficient and more specialized and focused and so on. And I was think analyzing like okay how do we like first of all I realized that if we don't have clients we don't have business and it's same for agency businesses for services businesses for product business for any kind of businesses. You know, if you have clients and sales, you understand them and they uh trust you to deliver some results, one way or another, you'll find a way to to do it. You'll hire people, you'll hire developers, maybe you you switch your team a few times before you find the right fit, but eventually you'll do it. If you have a development team and a product and you don't have sales, that doesn't work. Right? It's like you need that first before you can build. So that's why you know the framework I always go it's like positioning lead gen sales rev-ups and then customer success like why would you need customer success if you don't have clients kind of same logic here. Yeah. Yeah. So like uh and I know there are a lot of people with the same mindset like if you haven't done a lot of sales in your life in your if you haven't uh attracted customers yourself personally then you kind of don't have the core skill to building your business because everything start in business in 99% starts with sales. uh how we applied that when we started a data platform we've done five we sold five contracts first and then we decided to do a full-time team for for that uh product and eventually...

Transcript saved: 2026-03-10 Source: YouTube recording, Matt Laker Podcast channel Related files: - Prep: 2026-02-20 Matt Laker Podcast - Seva Ustinov (prep).md - Content bank: Posted/Seva/content-bank-matt-laker-podcast-feb2026.md - Content summary: 2026-02-20 Matt Laker Podcast - LinkedIn Content Summary.md