Most companies treat stability as success. But when Erki Koldits returned to Kontaktikeskus — the Baltic call center he built from 25 to 250 people — he found a company that had grown comfortable, bureaucratic, and stagnant. His response wasn't to manage the decline. It was to break everything. Processes. Management layers. Outdated thinking. Because sometimes the only way to transform a company is to dismantle the systems holding it back.

One of Erki's most provocative observations: large companies are now moving as fast as startups. Amazon fires tens of thousands of people and rebuilds with AI. Mid-size companies with the right leadership are doing the same. The difference isn't size — it's mindset. And Erki is living proof: he spends roughly 50% of his time as CEO building internal prototypes using vibe coding tools, turning what used to be a 6-month IT project into a working software tool built in half an hour.

Erki Koldits grew up in the suburbs of Tallinn, Estonia — simple home, not much money, doing everything himself. His first computer was a hand-me-down from Finland with 8-inch discs and a screen the size of a modern mobile phone. He taught himself to program, built simple games, and developed a logical, systems-thinking mindset that would later define how he runs companies. He became a CEO at 22 after a management buyout — and never looked back.

Roland Siebelink (00:43)
Hello and welcome back to the Scaling Without Breaking podcast. Today, I'm very excited about our guest. He became a CEO at age 22 after a management buyout, then built his company from 25 to 250 people - that's a full scaling story - making it the largest call center in the Baltic States. Now, 10 years ago, he just stepped back.

He handed the reins to his team, moved on to angel investing, crypto ventures, and semi-professional endurance racing, including 24-hour races at the Nürburgring. But last summer, he got a call no founder wants to receive. His CEO had left. The company he had built two decades was adrift. He came back and what he found was eye-opening. We'll ask him about that. I could say what I think he found, but I'll ask it as an opening question.

Now he's on a mission to transform this old school call center into something new and he's making a counterintuitive bet that in a world drowning in AI, generated emails, and LinkedIn spam, the human voice is becoming premium. With that, everyone meet my guest, Erki Koldits, the founder of Kontaktikeskus. Welcome to Scaling Without Breaking, Erki.

Erki (01:58)
Hello, good to be here and good to see you again.

Roland Siebelink (02:02)
It's good to see you again. We met in Estonia in Tallinn when I did a lecture there back in September and it's February now. I think persistence always wins. Glad to finally have you on the podcast, Erki. Erki, you stepped away from your company for 10 years and then had to come back when things fell apart. What did you find when you walked back in the door?

Erki (02:25)
To be honest, I was not very, very surprised, but a little bit because I always kept in contact. We had a weekly call with the management. I more or less knew what's going to happen. But the thing is that when this is not my main responsibility, I just have to go for a little bit, think about, then 40 minutes later, I don't think about it anymore. It's not on my table. But I found a similar kind of company that most of the mid-sized companies are. They are going to do the same thing all over again, basically staying in more or less the same position, maybe growing 3%, 4%, 5% a year or staying at the same level. This is a very stable environment where I came in. Nothing terrible, but also not great. The stable situation for me was terrible, because in the meantime, I was working with startups. Starting startups where the growth is five times a year, not 5% per year.

Roland Siebelink (03:26)
What was telling you something was wrong as opposed to: "That's just the way it is. This is the way a company of 10, 15 years old operates."

Erki (03:40)
I think when the management team stays in one company for a very long time, it doesn't help because they get comfortable. The companies get fat, a little bit lazy. They don't feel very strong competition. They're not hungry enough. I think this is the main reason.

Roland Siebelink (03:45)
I've found when I was for a short stint in a more mature company after having been in so many startups myself, there's often a sense of denial there. Did you find that in Kontaktikeskus as well?

Erki (04:13)
Yes, I was pushing all the time. I understand that the daily management always have these responsibilities - I have to take this, I have to fill this report, I have to do these kinds of things. It's very difficult for them to get out of the routine and find the time to actually break the things that they have built before. It's very difficult. But when we move on a little bit now, I've been more closely operating this company for the last six months.

It's very easy for me to break things because I haven't built anything. I can just break people saying to me all the time that we have done always like this. I said, "I don't care." Right now, we're doing like this because this is much better and more efficient. This is the future. So I think in some sense when the companies are in very mature or very stable positions, the only way to make the change is actually breaking. Letting go of half of the management because I generally see that we need different kinds of thinking, new ideas, new approach. Sometimes I think the breaking is good.

Roland Siebelink (05:22)
You said the tactic is basically you have identified a person that's no longer a good fit. You let them go. You take over their role for a little bit. And then you define what you need in a successor, if any successor.

Erki (05:38)
Yes. I think of the management of the office level. Most of the people actually didn't need a successor. Before I let them go, they say that they're very important and doing very important things. But two months later, I see that I maybe answered two emails from them. It's not too much actually.

Erki (05:56)
Sometimes I think they didn't do anything. I think they actually did work, but it was not necessary. They didn't do the work that actually delivers value or is critical for the company. I just ignore these things.

Roland Siebelink (06:10)
What is that kind of work that you find less valuable? Is it report creation, admin? What did you find as examples?

Erki (06:20)
All this bureaucracy, I basically ignore it. I see that also many times that one email, one request is coming in, three persons are processing it. But actually the first person should respond that it's not my business; we don't deal with this. These small things take a lot of time. And when I identify things that actually need some process behind it, usually I take half an hour, one hour and just automate it.

Roland Siebelink (06:51)
I think you mentioned it 250 to 300 people at the peak and you brought it back down to somewhere between 105 - 130. Are those numbers right?

Erki (07:00)
Yeah, this is almost right. I think that the breaking of things is in some cases, I see that some of the companies are doing too good quality. For example, we have an internal IT department and they will develop some solutions, and they put a lot of effort in testing, making sure everything works like corporate-level software. But at the end of the day, there were 20 people using it. So it was like, why do you do it?

Roland Siebelink (07:33)
You're pointing here to a perfectionism mindset as a driver for all these, maybe lower-value activities or building up a lot of bureaucracy. Would you say that that's the mindset a lot of people come with?

Erki (07:49)
I think yes. In governments we see this a lot actually. When they take one process, they polish and make the laws and regulations for two, three years and then they see that, okay, life has moved on and there's no value. And I just try to do the opposite of this; just solve the problem in half an hour and see how it goes.

Roland Siebelink (08:12)
As a capacity constraint, you mentioned something that customers are begging for more. They're writing me every week asking when we can have more. Is that something in the call center that you see or more in the startups that you work with?

Erki (08:25)
It's in the call center and I see this actually in the telemarketing sector. I actually didn't realize this myself. I was talking with one of the candidates for the sales management and he told me that all the emails, all the LinkedIn stuff they're sending out doesn't work anymore because the AI automated it so much. It's super simple to build automation that sends all the prospects out some LinkedIn messages, emails, very personalized, very well.

But nobody's calling anymore. When you actually write a number, a call to this person, usually they pick up and usually they listen if you're just doing a proper call.

Roland Siebelink (09:05)
We're re-learning to do cold calling again, yeah.

Erki (09:07)
Yes, because one thing, you actually pick up and they actually pay attention. With social media, you don't see any ads, you just pay attention for half a second that scroll on. But when somebody calls, you actually pick it up. And if it's not a bad quality cold call, but actually this person knows about, "Hey, I'm calling from this company. You are operating this business. We like to have some kind of discussion around." Doing the proper call, proper sales, not just the aggressive marketing, I think this works in B2B sales and also in B2C sales.

Roland Siebelink (09:42)
That's been, I think, the avenue that you put Kontaktikeskus on. So does that mean in this age of AI that is actually the growth opportunity? It's less about support and more about telemarketing, reinventing telemarketing for this new age.

Erki (09:59)
I think we can move B2C telemarketing more to the B2B quality. Before we call any prior person, we know more data about where he's living, what time he wakes up, what time he likes to respond to calls. The simple things that when I call one guy at 10 o'clock on Tuesday morning, and he responds, "I'm at work, don't call me," then you're very stupid to call next week and go again at 10 o'clock on Tuesday morning. Usually telemarketing agents don't use this knowledge. They call next day at the same time and get the same response. Also, we can enrich some data, working with data to make the call very personalized. We can manage with the scripts with the actual agency on the screen, they're like really personalized scripts.

Roland Siebelink (10:49)
Erki, walk me through that cost per order model that you invented. How does that work?

Erki (10:49)
It's a combination of things. Usually, it's cost per order thing is quite standard in telemarketing. We get paid when our customer or partner gets paid. We sell one product for 30 Euros, and the customer buys it for 30 Euros, our partner pays us 10 euros, we make 10 Euros. If we make money, they make money. This is generally the idea how I like to make business because this makes long-term cooperation, everybody wins. But we have been thinking that how to implement this kind of value or the same logic in other environments. And for example, when we are using this AI analytics for call transcripts, we transcribe most of the calls and we have one agent who is listening to the recordings and try to understand if there is a possibility to sell for this person. For example, insurance case, person is calling in and saying that I have a car accident, how I solve it because I have to solve it really fast because I'm going on a holiday trip next week.

Some we helped to solve this problem, very good. But at the same time, from the transcript, we understand this person is going to a holiday trip next week. We should sell travel insurance to this person. And we give this lead back to the company and they will do sales activities towards this person, really personalized. We don't say that you are going holiday next week, but it's personalized marketing that is around of this travel insurance.

Roland Siebelink (12:40)
And so the insight you had, I believe, was that those companies are often not that good at working with those leads. Is that something that you don't want to bundle as let us also sell travel insurance to those customers?

Erki (12:41)
It depends. It's Europe and this is the GDPR. In some cases we can do things, some cases we cannot do things. It's a mix. When we can do things, of course we are calling this person the next day.

I also evaluate how strong the lead is. If the lead is very weak, so there's no point of making the call, we better send email. If the lead is really, really strong, then it's important to make the call two or three days later to just catch the moment.

Roland Siebelink (13:26)
Exactly. But you were mentioning before that, the more platforms start building AI spam, the better it is for your business. Help us understand that better.

Erki (13:40)
The logic is that it's so easy to do the sales and the lead collection and the marketing with AI over the platforms. The noise level is getting really, really high. In this sense, the personal and human touch is getting less. I think this helps the companies who are actually putting effort in there.

It's not the cheapest sales. Of course, it's much cheaper to make the AI agent make a call, send emails. It's very cheap. To differentiate and sell higher value and more quality products, I think this is the place where we still need humans right now and probably in the near future also.

Roland Siebelink (14:23)
Do you believe in AI taking over the BDR role, the outreach role, completely?

Erki (14:31)
I'm testing this every week basically. Right now, it's lucky that we are in Estonia and we have a very small language that nobody speaks. Right now, we are definitely in this depth connection, it works because people are in different mindsets. If somebody is calling you to pay your debt right now or we do this and this, then people are not thinking about the voice; is this AI?

Even in cold call, I think there is this emotional touch and sensing the other person's humor and knowing how to respond. It's very, very, very difficult.

Roland Siebelink (15:08)
In our pre-call, Erki, you also mentioned something that big companies are now moving as fast as startups. Can you explain what you're thinking there?

Erki (15:24)
Yes. So I consider Kontaktikeskus is also not a five-person startup. Also, I think that there are some bigger companies who are doing very fast moving changes. You see every week that Amazon's firing tens of thousands of persons. As I understand, they also want to understand how to use the AI, how to improve the processes, automate things. And this helps them to move faster and with less persons.

But I think there's startup culture persons who have managers and the pressure from the market to actually do these kind of things. And then there's managers and owners who don't care, who just hope that everything will continue as it has before. And even the managers of mid-size or bigger companies, if they know how to do the changes fast, they will do these things because they understand the potential.

Roland Siebelink (16:18)
That's very good. But it's interesting that you said your own value is actually enhanced by just building prototypes and coding. I think you said you do about 50% of your time building prototypes and those MVPs internally. Do you see that as the CEO role of the future?

Erki (16:46)
It depends on the size of the company. If it is a very big company I think the CEO role is more managing with each manager. But then it's still considered a small company in some sense, and I like to be hands-on, so it's exciting for me also.

But I think one of the innovations we did last couple of weeks is that when before I just made simple prototypes in replica to some other places; they were not far from production, far from real data, anything like this. We opened up the real APIs now for persons like me. It's a prototype project right now. We have read-only access only to real APIs. We host the things in local machines right now. But at the same time, we have the same logins, everything looks the same like it's production software.

And I think this is a really interesting thing because think about the process earlier. For example, for a mid manager or top manager, we need to have one new software or new reporting or whatever kind of internal tool or external tool that we want to sell to customers. Usually I have three weeks of discussions with mid managers, top managers, customers doing some kind of evaluations with the IT data protection, whatever, six months project. Right now, I just open up one coding tool, vibe coding tool. I have the environment which knows all my APIs, all my design language, everything like this. I enter prompt. Of course, I prepare the prompt a little bit more. Half an hour later, I have a software that actually works.

Roland Siebelink (18:37)
So, Erki, I'd like to talk a little bit about the person behind the company, of course, right? So, tell us where you grew up and how did you prepare yourself as a kid and a teenager already to become such a powerful founder CEO at age 22?

Erki (18:54)
I was born in Tallinn in the capital of Estonia, in small suburbs, simple home, not very much money. Everything was simple, so we have to do everything ourselves. I have two tools to play with last 10 years. First 10 years, we don't have anything. This was the time that we made our own toys to play with. That was the situation.

And I think the excitement of technology was very driven with one of my neighbors, who was a couple of years older than I was. And we had these really, really old computers that have eight inch discs, I think, the really big ones when the screen was like a current mobile phone and the computer was half of my room. Somehow, my father managed to get this old computer from Finland, I think. These were computers they didn't use anymore, but it was the main tool. This was my main tool. I have learned how to program, make some simple games. It was all exciting for me. And I think this really changed my mindset and understanding of how things logically work together.

Roland Siebelink (20:04)
How did you get into racing?

Erki (20:13)
I think it's always been my thrill to find some adrenaline in different locations. I've been doing some skydiving, car racing, these kinds of hobbies that are little bit more dangerous. And one of my friends said, "Okay, come to test drive on the racetrack." I went, he said that you're very good, you're very good, you're very good. I'm stupid enough to believe this. And accidentally bought half of his race car. Then he pushed me to next weekend, he pushed me to take part in one race. And after this, he said, "Okay, you're very good. You should buy all the car."

So it kind of like, it's his fault, but also I enjoyed this a lot. And then one step to another. Some local races, I was doing this rally cross at the beginning, but this was very aggressive and very short races and I was already 40 years old. It was understood that I can do this, but I don't enjoy this so much. And somehow we ended up in this long distance race, the endurance races. This was much more enjoyable. And maybe when these 40 year guys are not the fastest every time, the strategy and the consistency also works.

Roland Siebelink (21:36)
What has that taught you about business being in those semi-professional endurance races?

Erki (21:36)
I think this is a combination with the skydiving and with the races. In the Nurburgring, the statistics are terrible. Basically, in 24-hour race, 200 cars start and maybe 120 cars finish the race.

This means that there is a 30% chance you actually crash, you burn, car breaks down or something like this happens. So you get the mindset in the start line that I already don't have this car. Basically I don't have it. It's much simpler this way. And I think it's very similar with the startup investment. When you make a startup investment, in my balance book I just write zero. I write it off immediately.

Roland Siebelink (22:07)
You write it off right away. Yeah.

Erki (22:29)
If five years later I see something comes out of this, I just think of as a lottery win.

Roland Siebelink (22:34)
So, Erki, last question. You already mentioned sometimes advising startup entrepreneurs, younger founders. What's the one lesson you want to impart to all of them?

Erki (22:35)
I think, most of them are doing the same classical thing; they are building for six months and then try to find the first customer. I see constantly it's a basic rule, many of them are doing it. I find myself in many cases, in many occasions, the same location, but I usually start selling if I have something. If I have customers, I start thinking how to actually make it. Usually I have idea how to make it, but now with AI it's a little bit more easier to build. Maybe you would just invest one day, two days to build something, but not more, not six months.

Roland Siebelink (23:24)
Erki, where can people reach you? What should they look up to? Is there a special offer that they could find on Kontaktikeskus? Where should they find out more about you?

Erki (23:37)
I'm trying to be not as active, at least active as possible in social media. I have an X account and I have LinkedIn account with one picture, but I think I am more active on LinkedIn. This is the place where actually I accept friends and actually read when somebody's writing to me. I think this is professionally the best place.

Roland Siebelink (24:01)
And of course, for those people that know me and do not know Erki yet, always happy to provide an introduction as well.

Erki, this has been an amazing interview. I love the full story about the turnaround of the company. Your way of building new products and new processes, and of course, also the whole analogies with racing and endurance in business has been great.

Erki (24:21)
Thank you very much. And thank you for the interesting questions. It has been really enjoyable to be here.