Anton Tomchenko
Build software on autopilot, not copilot
In this interview, Anton Tomchenko, co-founder and CEO of Emerence AI, explains why he left the staff augmentation model to build an AI-native services firm that owns the result for the client. He describes why vibe coding works for prototypes but falls short for the enterprise, how his firm prices work on fixed-price and outcome-based terms with tokens, platform and senior forward-deployed engineers built into the loaded cost, and why mid-size companies and large enterprises are opening the door to small AI-native challengers. He also covers forgetting how to code and coming back to building with AI, why he works more than ever, how his kids use ChatGPT as a tutor, and the operating system his team uses to build software on autopilot, not copilot.

Anton Tomchenko is co-founder and CEO of Emerence AI, an AI-native technology services firm built around AI-enabled delivery, senior engineering ownership, and owning the result for clients rather than supplying the people.
Anton founded his first company, WebKey Solutions Studio, while still at university in Minsk, Belarus. He then spent nearly 18 years at EPAM Systems, with a stint as a lead software engineer contracting at Google, founding EPAM's Engineering Excellence organization, running North American delivery and co-heading its West Coast business. He went on to serve as chief revenue and solutions officer for digital and software at Hexaware Technologies and, from October 2025 to September 2026, as CEO of FullStack.
“It's not an evolution. It's actually a lot about revolution in how we build things.”
Anton's core argument is that AI is a revolution, not an evolution, in how software gets built. Vibe coding works for prototypes, he says, but enterprises still need traceable, good-quality code, and that opens the door for small AI-native firms that price on outcomes and own the result.
"Honestly, I forgot how to code," Anton explains. "AI gave me an opportunity to start doing the same thing I used to do in my earlier career, building platforms, but in a very different way."
The conversation
In this conversation: Josh Rubin (Host, CTO Studio) and Anton Tomchenko (Co-founder and CEO, Emerence AI).
I've done technology services for most of my life. I was an entrepreneur early in my career and later joined the big enterprise. But I built software. This is what I do and what my company does.
So you've watched your life change a little bit, I imagine. You've been building software your entire life, before you recently started your own thing.
Correct, yes.
And before that you were CEO of FullStack, I believe.
I was, yes.
So what I want to kind of kick off a little bit here is around, frankly, coding, human labor, all of that stuff. You know how it is in that nearshoring space, FullStack, and you've operated in that staff augmentation world for a long time. But you're pivoting away, or you're deciding that the market is changing pretty dramatically. So let's start by just giving me some context. How did you start in that staff aug world, and what was that like? How long were you in it? What did you do in it, et cetera?
Absolutely. Well, staff aug was always a part of the type of work my companies were doing. And FullStack was a company that built a great reputation by providing staff augmentation services and hiring great talent. They built an amazing platform to interview people remotely. And this is where they built most of the reputation and most of the business. It boils down to the change of direction, end of the day. So the company decided to change direction and kind of stay close to the roots and continue building the platform. My background was primarily owning the delivery end to end and providing value to the clients and controlling the solution, consulting and advisory work. And this is what I want to build. This is what I intentionally want to do. And we mutually decided that it's better for me to continue doing this on my own. Generally speaking, my intention is to build more of the value-added services and own the results for the clients, rather than building the talent platform. And this, I think, was a little bit of a mismatch between what I wanted to do and what the reality of the company was.
Which is more of an agency model, as opposed to staffing.
Yes, and don't get me wrong. Staffing is a great model. And one of the questions I think you wanted to ask is, is staffing going to survive? Is staff augmentation still going to be needed? I think it will. And I think it's reasonable to assume that people will still need good engineers, good talent, but the nature of work will be a bit different, and the talent also will be different. So what I wanted to do is quite different from pure engineering work in the talent space.
You want to work on different projects. It sounds like ultimately this new technology comes out and that opens the world of what you can work on, what you can develop. If you spent most of your career doing product work and development work, is this almost allowing you to get back to your roots?
Absolutely. This is where it was an aha moment for me, because as an engineer in the past, I liked to play with the technology. Most of my career lately, I spent working with clients and businesses, but honestly, I forgot how to code. I didn't want to kind of learn the language again. I didn't want to go and start coding in Python or Java. And AI gave me an opportunity to start doing the same thing I used to do in my early career, building platforms, but in very different ways.
You had a 100x moment. What was that for you?
For me, there were several. Several years ago, of course, I was one of the first adopters of platforms like ChatGPT and later Gemini. And that was quite impressive. That alone helped me to understand what the new way of software, of interactions with the client, will be. That was so amazing. And it immediately brought the vision to me that software will be very different now. Everything that we built before will be different. But then actually trying to build something of my own, learning about LLMs, learning about how ChatGPT is working, getting to know more about AI and refreshing my knowledge about technology, and then playing with actual tools and trying to build something of my own, first vibe coding, but then working with actual systems, was the real aha moment for me. When I realized that it's not an evolution, it's actually a lot about revolution of how we build things. Then came the understanding that you can't just adjust what you do and make minor changes. You have to do it completely differently with this technology, from many different perspectives.
You were working in a field that was staff aug. Okay, this team has a product guy, knows what he needs, knows he needs people to execute on it. You provide the people that execute on it, and he's the one who tells them what to do. Now you've moved over to a career or a business model where you go in, you talk to a company, they tell you sort of what they need, and you full stack provide all of the... not FullStack the company, but like, you are the full-featured person: CPTO, CEO, product person, understanding and doing the engineering as well. Is that a defensible moat? What's to stop that company from just hiring that guy to do that internally? How does that work from an agency model?
It's not very different from the digital transformation 10, 15 years ago. There was always a question for enterprise: do we want to become a software engineering company, or do we want to hire professionals to do this work for us? And there is no right or wrong answer. Some of the companies would prefer to build these capabilities in-house. Everything will be different. Even staff augmentation will be different. We're talking about very senior hires. We're talking about forward-deployed engineers. We're talking about the systems that these engineers will be working with, provided by the enterprise: ecosystem, harnesses, all of that. But it's going to be very different from the traditional kind of staff augmentation and people placement, where we were working with big teams and just augmenting the existing teams. But it will still exist. Some of the enterprises will decide: no, we don't want to deal with that. We are not a software engineering company. We don't want to become one. We still need to understand what model is the best, what tools are the best, what ecosystem should be built around it, how many tokens we're going to burn, what the commercial model is, and we would not be willing to deal with that. So the moat here is the value creation for them: what to do, how to do it, how many tokens to burn. It's not going to be the enterprise's problem, for some of the enterprises. Buyers will still come to companies like Emerence AI, like other innovative AI-native agencies, to just give them the work to do and give everything else to them. So give them the full responsibility for the final product.
Are you accounting for ongoing maintenance, which is also the thing that I don't think a lot of companies are thinking about right now, because it's so new? You can vibe code anything. That's not the issue. The issue is what happens six months later when it breaks.
Yes, and vibe coding is not an answer for enterprise. Everyone knows it's good for prototyping, but to build the actual enterprise system, vibe coding is not enough. Absolutely not. It requires a lot of infrastructure behind it. It's a lot of hardware, a lot of compute to build things in the right way, to make sure it's traceable, that you understand what exactly was built. The code still needs to be good quality. You may argue that the final result is more important than what is under the hood, but to be honest, if you think about security, about traceability, about applicability of the new AI to the same system, good code still matters. And what quality of code is produced, what is under the hood, is still important for big enterprise, for sure.
What does your ideal client look like?
Well, we're still very small. We are very early. So our capability now is mostly in AI advisory, AI consulting, AI modernization and AI enablement. We also build with AI and build AI features. With growth, we will definitely expand to the full cycle of maintenance and longer projects, but today we're trying to solve very complex problems for the clients when they need advice. They need professionals to advise them what AI platforms to use, how to use them, to help them transform their organizations, to help them modernize very complex systems that nobody wants to touch. This would be our niche for now.
How does the money-making model of this work? Is it billable hours? Is it by contract? Is it telling them that this is how much value I'll generate, so this is how much you're going to pay me? How's it work?
Yeah, for a small company like ours, it's only outcome-based and fixed price. Most of the projects will be this type of pricing model. Time and materials will actually still play in this whole equation later, but it will be a different time and materials. We're not going to be charging only for the hours of the engineers. There are going to be a lot of components that add into the loaded cost, which is internal platform, consumption, tokens, all combined together with the hours of very senior forward-deployed engineers. That will be our expected T&M model.
Do you envision a future where you're all essentially in a token arb play, and you're fine-tuning your own internal open-weight model and charging them for your tokens, as opposed to giving the money to Anthropic or a frontier model?
Potentially, yes. It all depends on the specific tasks. So there is no ideal client. All the clients have different problems, different issues and different projects to work on. But I absolutely envision building, or renting, the hardware capability to do a lot of work internally instead of using the frontier models.
You say there's not an ideal client, but what about sector and space? You're in Silicon Valley. There's a technology company every 50 feet. However, those are arguably the most AI-enabled, have the technology teams, and they're pushing the most money into it. Are you going after endemic large-scale tech companies, or are you actually going after the rest of the economy that's out there?
The rest of the economy would be our choice, I would say. We're still going to work with digital natives, or ISVs if you will, high-tech companies. Happy to help them, and in a lot of cases the help is needed. But the primary focus will be other enterprises: healthcare and life sciences companies, financial services, CPG and retail, media and entertainment. So all the traditional enterprises, where we're going to work with mid-size companies and large enterprises.
Those are also companies, especially the larger ones, that traditionally would go to a PMC or a large-scale consultancy. So is the opportunity right now for small agencies to really take a bite out of those guys?
Oh, absolutely. I think the time is now. And I think enterprise companies and procurement organizations always think about ways to introduce smaller players. In today's reality, everyone realizes that sometimes the old companies with a lot of legacy will not really be able to innovate. And now innovation is very clear. What models do you use? What frontier are you partnering with? Do you have your own operating system? Do you actually own your own process, and do you have a very different approach? There's a very clear distinction between the old ways of doing things and the new ways. So enterprises are now opening doors for the smaller challengers. I've seen some examples where big enterprises are buying similar services at the same time from the big traditional players and the small players. And they're observing what is happening. They're trying to understand who can do the work better. But today I see a lot of enterprises opening the door for small companies to challenge the big traditional players.
How are you acquiring clients? Is this a beat-the-street, or is it submitting RFPs? Where does the inbound live for something like this?
All channels are good. We use our own connections. We use our own engine to actively develop connections. With enterprise, it's a lot about the personal touch and connections and people you know, and reputation, of course.
Yeah, well, it seems like AI has disrupted a lot of things, but trust is actually a big thing that it's disrupted. So personal relationships matter more than ever before.
More than ever before. I would say personal relationships are valued much higher now because of the productivity, the amount of work that can be produced, specifically for traditional work, for traditional tasks that are still in place and will be there for many, many years. The inertia of enterprises is still there. But it's very hard to impress with a good prototype now. It's very hard to impress with really good-quality delivery in terms of the technology itself. Reputation is still very important, and it will be.
You're obviously pursuing something that you've got passion around, and you're obviously all in on the capabilities of this new technology, but what anxieties do you have right now?
I would say I have more optimism than anxiety about the technology. I'm at the point where, to be honest, I'm so excited about the things that could be done in the future. I'm generally an optimistic person. That's what makes me an entrepreneur, and that's why I work a lot with the people on my team who balance me. But...
Does that mean you just have people that are anxious all the time?
Sometimes. Sometimes they bring their perspective, and they kind of ground me and tell me to slow down a little bit. But I'm generally very optimistic about the technology. I think it's making our life better already in a lot of ways, and it will continue doing the same thing.
It's making your life better, or it's making...
Everyone's life better. I think this is what's happening now.
Can you argue that? That it's making everyone's life better?
Yeah. Absolutely.
I can say that for the number of people unemployed because of the new tools, it is not making their lives better. That's interesting. Objectively, they are worse off with this than they were before, in the same way that a steelworker was worse off when the steelworking jobs left Pittsburgh and went to China. There are sociocultural ramifications of this level of disruption. I'm an optimist when it comes to the long arc of human history. I actually think it all ends okay. But that doesn't mean that there isn't disruption along the way, and I think we're in the very beginning of the midst of that. There are so many things, I think, that fundamentally need to be rethought, the more people I talk to around that, whether that is how we educate, from primary through secondary through college and beyond, or what work means. It is a lie to say that AI saves time, because we fill up time. You've got all these tools. I'm sure you've optimized a lot of stuff. I'm sure you're using the heck out of AI. Do you find that you're working less, or are you working more?
I actually work more, and I think this is because of the fact that I actually want to work more. This is my desire to work more, because I can produce so much more, and sometimes I just cannot stop and leave it alone. Many people I know actually produce more and start working more. And I've heard about no-AI days in some of the enterprises, where people just have talks and don't touch the tools. It's not going to be very different from previous technology. If you want to produce more results and you enjoy what you're doing, you will just naturally work more.
So that's a really, really interesting rabbit hole to go down, because if you think about the technology that's been rolled out in the last 10, 15 years, algorithmically driven content being a good example of that, TikTok, Facebook, Instagram, YouTube, it's passive consumption of stuff, the amount of time you waste on your phone, because what it's optimizing for is those little dopamine hits. I get the same or better dopamine hit when I'm in... it's about hitting the flow state. And so suddenly, when I'm using AI, I'm addicted to the productivity. It's still addiction. It is still hitting the same receptors in your brain for pleasure response. It's like going to the gym. It's a runner's high. Bottle that. That's the thing, I think, when I talk to CTOs and CEOs and the people that are quote-unquote AI-pilled: they are addicted to it. They're addicted to the productivity and the output and what they're able to accomplish, as opposed to the passive consumption. It's like getting addicted to going to the gym, which is not a bad thing.
It's not a bad thing, we think.
Because you can overexert yourself. You can hurt yourself by doing it. There are ways of going too far. We haven't built the frameworks and the healthy consumption patterns around it yet. We never did for social media.
GenAI and all the new modern tools, in a way, are kind of solving the social media problem that used to exist, as people get more engaged in building their own things and stop being just consumers. I think this could be a pill for passive consumption. I think the next generation will already figure it out: that doing something with your own hands, even if it's simple, even if you just talk to the environment and it does something for you, is much more engaging than just consuming things. So I think in this way, it's a wonderful alternative.
Does your 14-year-old have access to AI tools?
Yes.
Okay, tell me what and how, because I've been thinking about this constantly, and I don't know what to do here.
Well, I'm trying to engage my 14-year-old. I have four kids. The younger kids turn five and seven in a couple of weeks, and then 14 and 16 years old. They're all very different. But what I'm trying to do is show them tools where my teenagers can play and generate something, build something. Of course, everyone started with ChatGPT. That's the entry point. But even the way they interact with ChatGPT is now very different from a year ago. They understand that it actually can create different results based on the query, and that ChatGPT can do many wonderful things for you.
Are they using it for schoolwork? Are they using it to educate themselves, to create?
It's actually interesting, because ChatGPT can be a good tutor. This is a really good benefit of it. And of course, every school is opposed to it generating the work, and my kids, of course, don't do it and stay away from this. But using ChatGPT to explain things when the teacher is not there, to give step-by-step instructions, is very, very helpful. And I feel that it's an amazing way for kids to learn.
I'm ADHD, and I also have a 12-year-old who's super ADHD. And I've been toying with the idea, I was like, okay, let's do your math homework, because I'm terrible at math. Let's talk to ChatGPT, talk your way through it with it. And I haven't done it yet, but this might have just given me the confidence to try that when I get home.
No, absolutely. Absolutely. Even for me, without these tools, I would not be able to quickly ramp up my projects and start building. I actually use the GenAI tools as tutors for myself. It's much more efficient than reading a book or listening to podcasts. It's actually an engineering way of trying: trial and error. Our generation, we used to take the stuff out of the box and start building it without reading a lot of instructions and manuals. I hate instructions. Yes.
Well, that's what everyone asked me when I connected Claude to my video editor for the first time. Oh, did you watch a lot of YouTube tutorials? No, I did not do that. There you go, you just do it.
Do it wrong sometimes.
You do it wrong, and then you ask what's wrong. Yeah, and do it again. It has compressed time in some weird ways. Once you start compressing something that would have taken weeks into minutes, minutes suddenly become unending. Oh, it's been thinking for so long. Ten minutes go by.
Yes, yeah. I was actually thinking about one of the possibilities for the future of work, where people will actually have a lot of interactions at the same time and have a lot of spare time, because maybe in the future it will all be done momentarily, in the blink of an eye. But now it's not. It's spent in like 15, 20 minutes waiting for results. So I think it frees up a lot of time. It frees up a lot of actual human capabilities to do other things. And maybe the future of work will be very focused one-to-two-hour sessions, where operators will actually orchestrate things, use their judgment to define what's next, review work, but then the rest of the time they may enjoy life and do something else. And maybe that's the balance.
We fill up time, though. I mean, for me, the goal is always: can I get into a flow state? Can I lock in, and I'm just going? And I find that those little moments can interrupt the flow state that I'm in. But also, to your point about one to two hours, realistically that's all most people are working, have always... unless they're laboring out in the fields, knowledge work has always been in drips and drabs.
No, absolutely. I think, yeah, the balance should be there. You cannot do more and more and more just because you can. There should be some frameworks and some kind of common sense about how much time you actually spend there. But what enterprises need, what big companies need... I would say the actual time you spend on the machine doing some meaningful work will be reduced. You would not need 40 hours a week.
As you're working with clients and you're going out into the field, do you have a step-by-step framework that is your framework? Like, your design: this is how my company, how we operate, and how it applies to you?
Yes, absolutely, yes.
And is that...
What we bring to the clients is a framework of how to become an AI-native company, step by step, through different levels.
And how did you develop that framework?
Through our experience, by doing it on our own, first trying and building something for us. And I did that in the previous companies as well, and in the early days, but it was all focused around what the right ecosystem is, how to do things right.
So how do you do things right in this new world?
Well, it's a lot of questions, it's a lot of components, but primarily you have to have your own operating system where you actually execute with AI. You cannot just use the publicly available tools. You need to build the ecosystem that is used for building software properly, in the right way.
What does that mean?
That's a platform that covers a lot of aspects: the models that you use, the preference of the model, the transparency of what is produced, the quality gates where competing models test each other's results. But it's basically the system that you use to build software on autopilot.
So you essentially have a process for building bespoke autonomous software engineering systems.
Correct.
As on-prem solutions, cloud solutions, whatever is necessary to any of them.
Whatever is necessary. The modern technology services company is not a SaaS company. It's not a product company. Every company that will compete in this space will have to have this layer of technology, their IP. But it doesn't necessarily mean that this IP is going to be sold as SaaS, or that we're going to use licenses in our financial model.
Essentially, you're building a matrix that is individualized to the company.
Correct, yes. And we should have our own, you know, for when the company doesn't want to deal with that. And a lot of enterprises will not want to deal with it at all.
Be part of the
conversation.
Whether you’re a CEO or founder who wants to be featured or a company interested in sponsoring, there’s a place for you here.