← BACK TO PROFILES

Alex Halkin

Founder and CEO·Competera·San Mateo, California·

Enterprise AI is about trust, not technology

In this interview

In this interview, Alex Halkin, founder and CEO of Competera, explains how machine learning replaced spreadsheet price elasticity models for enterprise retailers, why consumption-based pricing was the right model from day one, and how he gets buyers to understand compute costs with a McKinsey-level price architect analogy. He recounts founding the company in Kyiv one year before COVID, losing 23 scheduled pilots overnight, watching a signed term sheet vanish when the war in Ukraine began, and still reaching a 90 percent margin on AI. He also argues that enterprise sales are about trust rather than technology, that customer money is the cheapest capital a founder can use, and that broken unit economics at the foundation model labs will bring a storm for companies built on token consumption.

Alex Halkin
Alex Halkin, Founder and CEO at Competera.

Alex Halkin is founder and CEO of Competera, an AI pricing platform that recommends prices for enterprise retailers by modeling the many factors shoppers weigh before they buy.

Alex first pitched the idea in 2014 after his IT outsourcing company was acquired by a large consulting firm and he saw pricing teams running on spreadsheets and two-variable elasticity models. He registered Competera in Kyiv at the end of 2018, landed its first customers in 2019, and carried the company through COVID and the war in Ukraine before moving to the United States. He is a Forbes Technology Council member and a Founder Institute mentor.

“If you can use the cheapest money, it's the customer money. This is why the focus number one is the customers.”

Alex's core argument is that enterprise AI sales are about trust more than technology, and that founders should treat customer money as the cheapest capital and venture money as the most expensive.

"The new luxury is not showing you the best car," Alex explains. "Show me how many families you feed."

The conversation

In this conversation: Jacqueline Samira (CEO, Howdy) and Alex Halkin (Founder and CEO, Competera).

Recorded for the CEO Studio interview series. Interview recorded on 07/24/26.

This interview was recorded by CEO Studio, a media brand owned by Howdy (howdy.com). Transcript lightly edited for clarity.

Alex Halkin

Hey, Jacqueline. My name is Alex Halkin. I'm CEO and founder of Competera.

Jacqueline Samira

Tell me a little bit about your company.

Alex Halkin

We do price optimization for enterprise companies, mostly for retailers right now. It doesn't run on LLMs. It's actually a deep neural network that relates the multiple factors people consider before they purchase an item, and we dynamically recommend what the price should be for that item.

Jacqueline Samira

Before we get into the depth of that, when did you start the company? It was 2014, right?

Alex Halkin

It's an interesting story. Originally I pitched this in 2014. But it was too early for AI to run this stuff. We officially registered at the end of 2018, and the first customers came in 2019. There was a big research project behind it, and a really complicated story of how to run it efficiently and calculate the right price. It was actually four years of research and bootstrapping.

Jacqueline Samira

What were you doing before?

Alex Halkin

I owned my own outsourcing company. That company was acquired by a big consulting company, and my job became optimizing their different workflows and freeing up the time of partners and analysts. That's where I found there was a huge gap in how people run pricing in Excel, and I felt we could do something better with machine learning. The idea came from my previous job.

Jacqueline Samira

So what was the gap you identified? How were they pricing things? I'm assuming you were working for big retailers at the time. What was their pricing strategy?

Alex Halkin

Correct. Honestly, it was huge Excel files. Big corporations are still using those Excel files, and consulting companies advising them on price based on only two parameters: how your price change drives your sales change, called price elasticity. It's a very simple model that triangulates all this data on two parameters, where I felt consumers are different, especially in different locations in the same city. There can be completely different customer intent. At that time it wasn't possible to do it in Excel without machine learning, and machine learning wasn't a thing yet. But my diploma at university touched machine learning in 2005. Can you imagine, when the first neural networks appeared, we built a model that could tell you whether this is an apple or a banana without a camera. As any engineer, I saw it was powerful stuff, but at that time it was really compute intensive and very expensive. So I went back to my boss and said, look, we can consider multiple factors, not only how competitors move or customers move. We need to triangulate the weather, the store location, the marketing expenses, the other products on the shelf. There were a couple of parameters we needed to triangulate, call it demand elasticity, to understand how people react in that particular moment and that particular location. It was a great task for machine learning models back in the day.

Jacqueline Samira

So fun. Can you bring it down and ground it in reality? Can you give some examples of a product you could have this type of elasticity around? Is it the clothes we wear, a banana or an apple, a service? What is it?

Alex Halkin

This is quite boring B2B stuff, but imagine you and I own a couple of stores, maybe small convenience stores, and we know that some stores sit near a train station, where there's more traffic and people are in more of a rush. We can charge a bit more for a specific item, not crazy amounts, as retail sometimes does, losing the demand. Finding this small adjustment is where machine learning can help you understand: this banana drives your traffic, don't touch the banana price. Increase the price on the sneakers, or maybe a snack, a bit. It's a small adjustment, a fraction of a percent on the item. Not the crazy dynamic pricing people did before. And it gets you good numbers if you have huge volumes.

Jacqueline Samira

So you identified this gap and realized these models are designed perfectly for this pain point retailers have everywhere. Is that what got you to start your own business, or was there a moment when you decided to do it?

Alex Halkin

The whole idea was that I saw 20,000 hours being spent by those consultants to provide recommendations for retail in Excel. And they were super slow, once every three months, once every six months. I started asking, hey guys, first of all this is super limited and super slow. Imagine we can make it happen online, right now, in five or ten minutes. My direct boss was super happy. He said, Alex, this is a thing. But the guy selling this stuff said, Alex, you're killing our business. We're designed for billable hours, not for subscriptions. It was quite a big fight, a healthy fight. But the insight came that this just needs to be done this way, and it has the power. I'm really grateful for the first customers. When I was part of this big consulting company, it was easy to go somewhere and pitch. When I left the company and went to the same account asking them to try, they didn't want to pick up the phone. It was a completely different story. I hadn't realized it would be so difficult. One of my partners at that moment said, Alex, maybe the boss was right and it's too early. Imagine, we won a summit in London in 2019 with Competera's AI pricing, but it was like drone delivery. Everyone was speaking about it, but no one was really using it. We got a lot of nominations, two times in Forbes, in Wired magazine, all this stuff, but customers weren't really using it. We couldn't find someone who would trust this stuff. Even now, in 2026, there are a lot of questions about whether we can trust AI or not. But we got our first customers by the end of that year, and that year we made $1 million in ARR.

Jacqueline Samira

In 2019?

Alex Halkin

2019, yeah. We closed three accounts. We ran pilots with them, they proved it. We said, hey, give us ten stores and ten weeks and see how it works. And we just repeated the story again and again.

Jacqueline Samira

What's your pricing model? Do you do it based on the amount of money you save them or make them, or is it separate from that?

Alex Halkin

Honestly, the original idea was to say, guys, just pay me a cut of what we made for you, which sounds absolutely reasonable. But if you ask even Claude, those value-share models don't survive, for multiple reasons. We price based on compute, technically a markup on compute, since Google runs behind all of this. Usually we price per location, per price review. We call it compute modeling units, like Databricks, but technically it's per store, per month, depending on how often you need to run the calculation.

Jacqueline Samira

Is that working better, that pricing model?

Alex Halkin

It's more sustainable. I always say we do pricing, I've done pricing all my life, and pricing is a process. You need to reevaluate all the time. We were probably one of the first. Right now a lot of companies are moving to consumption based pricing, because it's not like Salesforce, where you pay up front for the future and then charge per seat with huge margins. Right now each request requires some compute. That's why everyone is changing to this consumption model, and we were there from the start. For our audience building products today: as long as you deliver value, make it explainable so they understand what is behind it. My analogy is always: imagine a McKinsey-level price architect sitting in each of your stores, reviewing all your prices four times a day or a week, and it will cost you 30 bucks a month. Are you okay with that? They say, yeah, okay. Now you're bringing the analogy to things they understand. We run around seven terabytes of data when we run one price calculation for them. But for the usual person, is seven terabytes expensive or not? They don't know. That's why we bring the analogy to a real human who thinks a thousand times more efficiently, and compare the cost to what a human can do.

Jacqueline Samira

I can imagine any of those big firms, when they're charging $700 an hour and billing 20,000 hours, saying, let's not change that gravy train right now. This is more theoretical, but how do you think these consulting companies are going to pivot now in the age of AI?

Alex Halkin

They already started. Every year there's a ranking of how many interns from top schools go to those big consulting companies, because it's the dream career. You finish university, you go as an intern to a big bank or a big consulting company. The average hiring rate was 150,000 people a year in America, and this year it was something like 5,000. A huge drop. Double check it, but there's a huge drop in interns who can find a job. A good consultant, the guy I used to be, with existing models, is just ten times more efficient. And there's other research showing that the people who can really evaluate the output of a model benefit the most, because you know what you're looking for. You read it, say okay, great, and you just became more efficient. This creates a strange and very scary gap for me, because when I was back in that big organization, all those reports they gave me on Thursday that I needed to polish by Monday, that's how you learn, by doing those complicated, repeatable tasks. Now we're missing that layer. The bar for education just went up. You need to be more advanced and run on those models. They are already pivoting. They deploy LLMs in-house to keep control.

Jacqueline Samira

These next few years are going to be very interesting. I would love to learn more about your business today. You did a seed round in 2024. Is that correct?

Alex Halkin

Correct. We closed the extension of our seed round. Imagine, we founded Competera one year pre-COVID. We had 23 pilots scheduled, and COVID started. The problem with COVID was that no one knew when it would finish. My pitch was, give me ten stores and see how it works, and technically we could start piloting. But no, everyone was shut down. I was paying Google every month because I needed to keep the data flowing, and Google wasn't accepting any excuses at the time. So we needed to wait. Then as soon as COVID ended, everything blew up, and then the war in Ukraine unfortunately started. Some of our customers were affected, and a lot of big businesses that were working with us shut down. It was a great time for fundraising, right? And it's a fun fact that we are now landing big corporations, Fortune 100, and they ask who is backing you. I say, what do you mean? My cash flow is backing me. They say, no, it doesn't work like that in America. Alex, you need someone we can trust. They also called my angels, my seed investors, to understand what was going on.

Jacqueline Samira

Because no one has a profitable business.

Alex Halkin

Correct. My burn multiple is less than one, which is absolutely not acceptable here in the Bay Area. I feel like maybe I'm doing something wrong. When I landed the first account, by the way, my old consulting company was one of our design partners. They really helped me close the first deal. We signed a nice deal, 300,000 pounds for a startup. It was me, my co-founder on token salaries, and one part-time developer. Nice deal, right? But I paid Google 260 at that time, and I saw my margin was 17 percent, while my outsourcing business had 50. I said, I'm doing something wrong. Good news, it works. Bad news, how do we make it efficient? It was challenging to make this stuff efficient with the right margin. Right now we run AI at a margin of 90 percent, which is quite unusual. There are a lot of tricks for how to do it. I really encourage all the startups to think about your unit economics, especially those building on top of OpenAI and all these models. They're really in trouble with their unit economics. Everyone knows it. They've been public about it. You saw the news this morning about Sam asking the government to back OpenAI.

Jacqueline Samira

I actually didn't see that yet. We've known these large foundational models are not built correctly. It's going to correct at some point, but they're waiting for the compute cost to go down.

Alex Halkin

Some rumors say there's a 14X difference between what they charge and what they really spend per user. It's a huge bubble, technically. A lot of investment right now is going into sub-quadratic models, which will be at least four times more efficient than searching for the next token, but it's still research. Building the company in Eastern Europe, you don't have those VCs who can give you unlimited resources. When we raised, our original seed round was one million bucks, and it was the biggest round in Ukraine that year. But what is one million bucks? You need to fly over the ocean, you need to hire salespeople, you need some room for mistakes. A friend of mine said, Alex, I hired my VP of sales on the sixth try, and he had raised 100 million from Andreessen Horowitz. I said, I have no money for so many mistakes. I need to make it on the first shot. And it doesn't happen on the first shot.

Jacqueline Samira

So you were in Ukraine at the time. When did you come over to the US?

Alex Halkin

I started flying back and forth, and we started growing in Europe. It's an interesting observation in my space that European enterprises are more agile. They try more systems. They're not locked in by huge vendors like Oracle with five-year agreements they can't drift from. In America, I don't want to mention names, but huge corporations run very legacy software, 40-year-old software in my space. In Europe we don't have this blocker. They're flexible, they can adopt. And since COVID ended, as soon as the first flights opened, I moved to New York.

Jacqueline Samira

Wow, what a story. And when did you do the seed extension?

Alex Halkin

The extension was a nice story. Three months before the war, we put all our metrics together. We grew like crazy that year, 800 percent, mostly because a big partner replaced their own solution with ours for their customers. We didn't really sell, they just replaced it, and our revenue blew up. But that was Russian revenue. We tried to raise a Series A at the beginning of 2022, and then the war broke out at the end of February. I even had two very strong term sheets. One was at DocuSign. I paid lawyers in Europe 40K to finalize the deal, we sent it to DocuSign, can you imagine, and then the guy just disappeared. No answering email, no WhatsApp. He called me in May, four months after. Another guy said, Alex, what will we do if a nuke drops in Ukraine? I told him, if a nuke drops in Ukraine, the investment in Competera will be the smallest problem you have at that moment. But that doesn't help me close. There were no options. We just needed to survive. We were already profitable at that time, and we had some reserves. Our local investors, my seed investors, helped a bit. And we just said, okay, let's see what comes next.

Jacqueline Samira

What a crazy five years from 2019 to 2024. Thinking of now, is it easier to sell? Do you find people understand it, or are they still confused by it? Is it still easier to sell in Europe than in the US?

Alex Halkin

First of all, there's a huge tailwind from AI adoption. It's much, much easier. Of course enterprises are still behind on AI adoption. A lot of governance teams still don't know what to do with it. They want to control it, they want to understand it. But at least they ask the right questions. They ask how often you train your model, how you control model drift. We never heard those questions before. People always try to compare: can they build the same? Enterprises like to look at a startup and allocate five people to work on the same thing. Then they realize, wow, it's hard to build, because we built ours on proprietary data sets. You can lock ten people in a room and they'll end up with a problem that goes beyond AI. You need intelligence behind it, not just workflow systems. So it's easier. Still, Europe moves much faster. We tripled our revenue in Europe, and America still needs more time. We're signing a lot of cool partnerships now that help us build trust, because they need to know you here. For anyone listening who is building B2B enterprise products, it's mostly about trust, not about technology. Imagine, Jacqueline, we build a Formula One car together. No one will allow us to compete next season. A lot of political decisions have to be made before they accept our car. Even if we have the best car ever created, with all the regulations in place and all the certifications, it doesn't mean we'll be accepted to compete. That's what the reality of enterprise sales looks like. But it's getting better, because once they understand they definitely need it, we land a few cool accounts. I still can't announce them, but the companies are very well known, and they'll be live this year.

Jacqueline Samira

That's incredible. I've been in enterprise sales before. My husband is also an entrepreneur who does enterprise sales, and it is so painful. Like you said, it's trust. If you can get one or two flagship accounts you can use, great, but sometimes they make you sign an NDA and you're like, let me use your name, please.

Alex Halkin

They actually make us give discounts because they fight not to announce. We have a very well known grocery chain here, and for five years we can't mention them. They're super strict. They feel like my software is their secret weapon. They don't want competitors to know they use specific software for pricing. Of course, before that they used even worse software they built in house, which could increase your price ten times because it thought there was high demand. We do it much more efficiently and safely, but it's a sensitive area.

Jacqueline Samira

A great example is when egg prices were going crazy at supermarkets, and there has definitely been misuse of pricing that we've all experienced. So what does the future look like for you now? Are you focusing more on the technology, on sales, or a mixed bag of everything?

Alex Halkin

I feel like we are truly a scale-up now. I had a very interesting meeting with the Andreessen Horowitz team, a two-hour meeting. All my friends said, Alex, you just landed and they booked two hours with you, it means they're investing. I said, honestly, from their questions, they just want to learn more. They gave me an interesting analogy. They said, Alex, you're like a teenager. You're too big for a small fund, but too small for a big fund. That's what happens when you don't raise money but have double-digit growth. We have the metrics of a Series B company, but legally we never raised a Series A. I don't need the money from that perspective, but of course it unlocks growth. We want to launch a new version of the product for different industries. We're talking with gas stations that can adopt the technology as well. And a good company needs to keep balance in everything. Go-to-market, building trust. We built a very big pricing community, the biggest pricing community on LinkedIn, with online certification programs, and it's run by one person, half time. I just need to put more fuel there. On the product side, we have a lot of ideas to land, and to expand into new industries. We're creating artificial general pricing intelligence, where we price for car rentals and other new industries. Customers are really waiting for that, and I already have some pre-arrangements. But I'm very conservative on our spend. I keep Competera near profit, because you know what we've been through. We survived because we managed our cash flow very well from day zero and didn't burn like crazy. COVID can kill it. Imagine you have 23 accounts and then four are left overnight. Then the same story a second time. So if I find a great investor who really believes in this, we can build it together a little bit faster. The technology is proven. We just need to scale up.

Jacqueline Samira

I feel the exact same way. We bootstrapped our business from 2018 all the way to 2021, did a seed in 2021, then an A in 2022. But prior to that, we were cash flow positive, and people said, why are you cash flow positive? That doesn't even make sense. We were super conservative on our spend on everything. We also have a capital-intensive business, because we're floating people's salaries while they work for other companies, and we have to get paid. So we always had to have enough to cover the carry. My co-founder and I said we need to have at least three months in the bank, so in the worst case, if everybody cancelled their contracts, I wouldn't have to go do massive sales.

Alex Halkin

It's a very nice mindset. I absolutely agree with you. And I love this phrase: the new luxury is not showing you the best car. Show me how many families you feed. How many employees you have. This is the new luxury.

Jacqueline Samira

How many families you feed. I love that.

Alex Halkin

That's the mission behind it. I know it's illegal to say this in the Bay Area, because when I speak with early entrepreneurs, they really don't think about how much it costs. They feel like they'll go to an investor and raise a round. But for me, funding is really betraying yourself. You sell part of your baby piece by piece. Of course it's needed, but I always say, if you can see how money can speed your time, at least 100X of what you're looking for, fine. VC money is really the most expensive money. If you can use the cheapest money, it's the customer money. This is why focus number one is the customers. Focus number two, we actually have debt funding, which I really love. I know the initial model training is very expensive, and I know the conversion rate of my pilots. So for me, taking venture debt is more efficient than taking VC money, which dilutes my equity, because debt can be returned. It's about balance. But now I feel like this is the moment to open new ventures, with a very like-minded investor. What's your feeling?

Jacqueline Samira

We think the exact same. I speak to a lot of founders, and they go after the glamour of, I raised this much money, I did this many rounds. I see the value if it's necessary, because you have to do a certain market grab. But people are doing it as a vanity metric, and that's not impressive to me.

Alex Halkin

In most of the introductions here, people don't say what their customers get or how big their revenue is. They always say, hey, I'm Alex from Competera, I raised 10 million bucks. Okay, but what does that mean? If I can do with one what you've done with ten, then that's not the metric. Come on, how much money you raised.

Jacqueline Samira

Exactly. It's like, well, how good of a business have you built?

Alex Halkin

Yeah, how big is your team?

Jacqueline Samira

I got so lucky with the investors I had, because I felt like people in the Bay forgot what building a business was for. What is the purpose of building a business? Yes, you have these slingshot companies, and that's what the VCs are investing in, the 100X, the 1000X. They invest in 100 companies and only expect one or two to hit that moonshot. So they're banking on you to have that moonshot, and what's the advice they're going to give you? One of the things I tell founders is that the advice investors give you might not be best for your business. It's best for their investment. Those aren't necessarily always the same thing.

Alex Halkin

I'll send you an article about that: be careful with advice from investors. In that meeting, nothing was bad, nice team, but in the middle of the meeting I started feeling exactly what you said. The first question when I entered the room was, Alex, how much money do you need to make your product free for everyone? I said, okay, probably you know something I don't. Please explain. They gave me the background: there are cool models right now to crack enterprise sales cycles by giving the product away for 10 or 12 months with an obligation to pay in a year in case of success. They call it contracted ARR. They had one portfolio company, I can't mention the name, that officially announced they landed 178 enterprise accounts in two years with an ACV of 200K each. When you do the simple math, an enterprise sales cycle is six to eight months, so you'd need 178 salespeople who each close one deal. It doesn't fly. Something was broken. And this was the trick: they give the software for free with an obligation to cover the invoice or cancel the subscription in a year, but if they cancel, they're banned from working with them for the next three years. So the sales team can call and say, HP is using us, Apple is using us. It's true they're using it, but they're not paying yet. They report this contracted ARR, expecting five or seven of ten accounts to convert to paid, and they raise one round and then another. Sometimes it works. It's just not my game. I told them, guys, I feel it, just no. I'm not comfortable.

Jacqueline Samira

We ran some free trials, and we learned very quickly that those are the worst customers. You have to have some skin in the game to make the relationship work. This has been so wonderful. I have one final, overarching question. Where do you see the world of AI five years from now?

Alex Halkin

That's a cool question, and it depends on the area. Of course AI will explode and continue on this path. I feel there will be a huge pushback on coding. The quality of AI on coding is overestimated. The idea that coding is solved, which Anthropic loves to say, it's not. There will be a pushback, and smart developers will need to find a way to work better with those agents. A lot of ownership will come. The big LLMs will lose market share to open models, as happened with software and hardware before. We'll end up running on our own. Robotics, of course, a lot of cool stuff will happen in day-to-day life. I'm generally very positive on this. But especially if we survive those crazy broken unit economics. Your company and my company aren't built on top of token consumption, but something like a thousand AI startups were invested in over the last two years, and of course several will survive. When OpenAI raises prices, and they will raise prices, the question is how and when. Subscriptions will probably move to token consumption. A lot of businesses will fall back, a lot of investors will be disappointed, and money will disappear from the market. So I'd just be prepared for a small storm on the ocean over the next five years. But in general, I'm waiting for it. AI pricing in 2019, everyone said, oh no, it's impossible. Now everyone says, do you have it? When can we start?

Jacqueline Samira

Wise words. Thank you so much, Alex. This has been a great conversation.

Alex Halkin

Thank you.

GET INVOLVED

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.