Jarek Kutylowski, CEO and founder at DeepL: Why DeepL wouldn’t settle for “good enough” AI.

Thank you for having me. Over here at Slush.

Yeah. I mean, like, we started in 2017, which was basically the point in time when deep learning, like the basics and the foundations for what we know as AI nowadays, back then nobody called it AI really, that was the moment when this has all been emerging. Maybe a little bit in the background, maybe a little bit in academia, maybe not everyone out there in the world has already been there. But we kind of really seen like this technology is going to be game-changing for kind of anything language related. We've seen like these models surpass anything that's out there on the market. And therefore we also seen like this moment, okay, this is the point in time when you can break in, like there is big offerings from big tech on this, but they're going to have to like throw everything away that they had until now and go with this flow. And this is the opening when you as a startup can come in and create something amazing. Yeah.

And then out of this, like really this very typical, I think PLG-ish kind of motion that we've seen, first a lot of self-serve, a lot of adoption of free usage, and then kind of doing our homework and walking our steps onto more and more sophisticated use cases, more and more sophisticated customers, up to enterprise basically.

Yeah, I think it was about the availability of this technology. Like, we've just seen this technology and, like, oh wow, this is going pretty damn cool, and you can do something about that. Like, yeah, we knew the language problem. I think that is part of that, and maybe I'm just kind of underestimating that a little bit. I'm more a technologist rather than a language person, I have to say. But still, I speak three languages and, like, you're kind of sitting in Germany with France across the border, Belgium across the border, like you know the problem. But maybe I just underestimate the importance of that. For me it was mostly about the technology and, hey, we can do cool research that is going to actually have an impact from day one on people's lives.

Yeah, I mean, like, the availability of general purpose AI, basically, everybody has AI at their fingertips somewhere, whether it's built in your phone or in your search, or like you have a GPT subscription, wherever, there's AI everywhere and companies are also embracing that of course. So the kind of availability of kind of good translation, that is now the reality for everyone. But that's been kind of planned into our business model from the beginning on. Like, we knew we're fighting against big platform players out there who once again, they own your operating system, they own your phone. So we knew we can't compete there really. So we very early decided also not only to focus on quality itself but also focus on the application of that quality in a business, in those areas where quality matters and where kind of each and every small piece of quality that you can add on top of what is maybe available generally makes a difference of whether you're going to have to make one correction in a whole contract or whether there's no corrections at all and it's just perfect out of the box. So like kind of getting the defined nuances ready still makes a huge difference.

Yeah, I mean, like, funny moment for you to ask because we've been always very focused and very concentrated on a smaller set of languages and it was kind of the common answer, like, DeepL is doing like thirty-something languages but they're doing them great. We have actually expanded our offering to over a hundred. Yeah, I think basically last week I saw it's like a beta or something? No, that's live now for everyone in production. And this has been very much requested functionality by our customers who, yes, the huge part of their content that they need to translate is still in those thirty languages that we've been supporting for a while, but here and there they do have other language requirements and they want to have just one single provider for all of it. So this is like a very clear reaction on this. Other than that, we're really focusing now much more on kind of enterprise customers and their needs. It is about kind of going deeper on the product side, but not only from an application layer, but also building the functionalities in the AI that's going to support this product. That's going to allow, I don't know, a customer to fully control their language. So it's FDA compliant if you're a healthcare company in the US or something like this. There's a lot of examples of how you can actually leverage having control of the technology and put that throughout the whole stack up into the product and into the application for particular customer use cases. But for that you got to really deeply understand what your customers are doing in language. And that's always a big question because translation is so broad, like you can use it for anything.

Talk to them, talk to them, and like, talk to them really across the whole company. Like, we have folks from our research teams who are sitting together with customers and trying to understand what they're doing. Like, of course you can't do it for each and everyone, but you have to do as much of this as possible or as reasonable in order to get nearer to them.

Yeah, I think our approach has been kind of bottoms-up and PLG. We've established this brand that DeepL is the specialist for anything language related. And I mean, okay, we can talk maybe a little bit about our new products which are actually going outside of the language space, but for the sake of this question, like, people do know that DeepL is the go-to place to go. And in the past that was basically it. I think right now we are then engaging in much deeper conversations, like, we're just leveraging this but then it's just like a door opener to our customers and then we can engage with them into conversations like, okay, what is your kind of highly critical translation needs, language needs that are most core to your business? If you're, for example, a professional services company that is working across the whole Europe, you're going to have your specialist on a specific topic sitting in France, but you have your customer in Germany who needs this expertise. How do you bring them together? Do you hire new people or grow them in Germany, or do you just take your French experts and use DeepL Voice to kind of have live translation in all of the meetings that you have with your clients? And that needs a little bit more focus on the company itself. And a lot of our go-to-market buildout was actually into this direction.

Yeah. I mean, you've asked me earlier like how did DeepL come to be, how did we build it and why did we build it? For me, it was all about the excitement about AI and how that can change the way that we're working. I think translation was an amazing place to kind of start because this was like the first area where AI really made this difference that was not only on paper, not only a proof of concept, it really kind of changed this industry. And we're kind of like, while we're known for this language aspect and for translation, for us it was always about AI and making that applicable, making research work in products, in real-life use cases. And out of this, at some point in time we decided, okay, we want to go also a little bit out of language, like there's bigger fish to fry actually, like AI can do even more. And we're still incredibly excited about that. And we think with the DeepL Agent, which is just a pretty horizontal agent that can help organizations just to be more efficient in anything that is kind of administrative, in anything that is kind of sitting maybe in finance, maybe in legal ops, maybe in people ops, abstract away all of those workflows that are kind of still very manual nowadays. We've moved quite far along in terms of automation as a civilization, but not really that far. We still all do copy-paste all the time. And kind of one of my mantras in this is like, no copy-paste anymore. This needs to be gone. And we're kind of following a similar approach that we did for language, where we kind of bottoms-up created a product which was simple to use and for everyone, so that everyone has democratized access to translation and can do it on their own. And we are following in the same tracks with our agentic approaches, like just give it to ten, twenty, fifty, a hundred, all of your employees and let them figure it out, because people are actually pretty amazing at taking AI and building workflows with that. And we've seen such incredible ingenuity with our customers there.

I think the good thing for us was the understanding of the product that we started with, translation. It's just like, if you're sitting in the Bay Area and the only language that you've ever learned is English, you don't get the problem space. And like, we always talk about the fact that you have to understand your customers, you have to understand your users, you have to understand your problem. And that's just hard if you're not building out of Europe for this particular case. Obviously, kind of access to talent, I think we haven't run into that much competition especially at the beginning when we've been building the company, comparing maybe to where you would be if you were in one of the big centers and big hubs. So that's definitely good. And kind of also access to the markets, like if you're building out of Europe, you are being recognized for being a European company. I do think that matters. I think that helps you in adoption in those markets and there was a strong correlation between who our customers were and where we've been sitting. So I think it all kind of fitted pretty well.

But then at the same time, you got to be also cautious. And I was always very, very obsessed about the fact that we do not limit ourselves to Europe. Like, we're sitting in Europe and we're maybe built out of Europe and there's a lot of this kind of diversity and European DNA in all of this, but like, if you're in tech, you have to target the global markets. I think there might be niches where you really want to go very specific onto Europe or even specific countries, but in general, look at the biggest companies that have built out of Europe. They've been built out of Europe, but at some point in time they said we want to serve global markets and that's incredibly important.

Super, super early. I think for us, maybe different to other companies, Asia was the market that we've expanded first. That's once again because of the product and the traction there. Anybody else would kind of start with the US, I guess, that's the typical one. For us, it was Asia first and US next then. But yeah, pretty early because of this.

Yeah. I think honestly, like, really the best moments come always from kind of new things that we can offer and us starting to see that these are coming through. And I do see a lot of that for the DeepL Agent and kind of the early things coming through. But it's still going to go back a little bit more for DeepL Voice. And I think the first meeting that I had with a Japanese customer where we had DeepL Voice running in the room, in the meeting, and how much of a real game changer that was for me. Like, for once I can really understand this customer. I know what they want and they can understand me and it's like, oh my god, like, the world has changed right now.

Yeah, I think, honestly, I think there was a moment when in this whole kind of rise of AI, and not really that much kind of competition coming up, but like uncertainty coming up and this market being just so dynamic and there's so much noise coming up and there's basically a bubble that is forming around us and we're in the middle of it and we've been around for quite a while and haven't seen it. And there was a moment when I kind of lost maybe my faith or where things have been hard for me personally and I had to kind of sit down and focus a little bit and close my eyes mentally maybe and kind of realign to what the world looks like right now. And that was probably like one of the hardest emotionally moments for me to kind of go through. But I think that's just like probably one of those typical things, like, you're a founder, you're kind of growing with the company, you're growing with the market and you just got to absorb that. Everybody kind of breaks at some point in time in that and you just have to figure out, are you able to get out of there or are you not able to get out of there? And how do you do that for yourself?

A lot of work with a coach, like an amazing coach that I had at this point in time, but also talking to my board, talking to my investors. They've been like amazing companions on this journey, like amazing people. Getting like actually talking about those problems with them and doing a little bit of a therapy session with your board, maybe.

It's a pleasure. Thank you very much for having me.