Anton Malling on Scaling Kapa.ai From the Inside

Anton Malling joined Kapa.ai as a Founder’s Associate in late 2023, when the company was still much earlier in its journey. Today, he is Chief of Staff at one of Denmark’s fastest-growing AI startups, helping build a company that turns technical documentation into AI assistants customers can trust.

Kapa sits in one of the most important parts of the AI market. Every company wants an assistant in front of users, but most quickly run into the same problems: hallucinations, outdated answers, messy data sources, broken evaluations, and models that are not reliable enough for complex technical products. Kapa’s answer is to build AI assistants grounded in a company’s own technical content, from docs and GitHub to PDFs, Slack, Zendesk, Notion, APIs, and support knowledge.

What makes Anton’s perspective interesting is where he sits inside the company. As Chief of Staff, he works close to the founders and across the business, seeing how a fast-growing AI startup makes decisions, handles customers, builds systems, and keeps pace as the category changes around it. It is a role that requires switching between detail and context, urgency and structure, execution and judgment.

There is also the personal layer. Anton is doing this while studying International Business at Copenhagen Business School. Before Kapa, he had already spent years building entrepreneurial communities and projects, from Start-Ung and Bcubator to Fashionmarked-Danmark. That background shows up in the way he thinks about operating: close to people, close to execution, and comfortable building before the path is fully defined.

We sat down to talk about what Kapa looked like when he joined, why trustworthy AI assistants are harder than they sound, what a Chief of Staff actually does inside a fast-growing startup, and how he balances one of the most demanding roles in a startup with still being a student.

Q&A

1. You joined Kapa as a Founder’s Associate and are now Chief of Staff. What has changed most in the company since you first joined in 2023?

I think 2 things really stand out, (1) is the team scaling, and (2) is the AI market in general.

On the first point, back when I initially joined, the team was essentially just the 2 founders and an engineer on the technical side. Today we're +30 people across Europe and NYC, which is a big shift in how you operate as a company. Luckily we're fortunate enough that all 30+ people at kapa is insanely smart and ambitious so we've managed to keep the culture that really makes kapa, kapa.

On the second point, we got into AI when this was still pretty early, GPT was a thing but companies were still figuring out how to actually utilize it in a helpful way. For some of our early customers we were basically the first AI tool they bought. It's shifted a bit now where AI is definitely something these enterprises are thinking about, and the questions is not: "Should we do something with AI?", it way more becomes: "Between all these AI projects we could choose to spend time on, which ones actually move the need?". It's going to be interesting to see how this moves going forward for sure!

2. Kapa helps technical companies turn their documentation into AI assistants customers can actually trust. What do people misunderstand about how hard that is?

In my opinion the core thing people misunderstand is they confuse the demo with the product. You can hook an LLM up to your docs in a weekend and it'll answer okay, so people think they're done. But that's maybe 70% there, and the last 30% is where the actual hard engineering lives, and not all of it is the LLM, it's the also just as much the retrieval.

Naive RAG just grabs the nearest chunks and hopes the model stitches them together. With real technical docs that breaks, a correct answer often has to combine an API reference, a tutorial, and a changelog, so you need multi-source reasoning, re-ranking, and the hardest part, knowing when to say 'I don't know' instead of inventing a config flag that doesn't exist. We've seen a big dev-tool spend a year and a half on this and give up. And even if you do manage to get there, it's never finished, your docs and product keep changing under the index, and new models keep shifting what optimal RAG even looks like.

3. You are in a role that sits very close to the founders and across many parts of the company. What does Chief of Staff actually mean at a fast-growing startup like Kapa?

I think Chief of Staff doesn't really have a fixed definition, it means something different at every company, and probably honestly something different every few months at the same company. At a company like Kapa, the way I'd describe it is, that you find whatever the biggest bottleneck is right now, you try and figure out how to actually solve it yourself, and once you've got it some-what working, you go hire someone who's genuinely world-class at that one thing and hand it to them. Then you move on to the next problem. So in a way the job is to repeatedly work yourself out of a job. You're sitting close enough to the founders that you can take things off their plate before they become fires, and the role is defined by what the company needs next rather than by a fixed lane.

4. You have built communities and companies since you were young, from Start-Ung to Bcubator and Fashionmarked. How has that shaped the way you operate inside Kapa today?

I think I take a lot with me from the earlier projects I did. None of those came with a playbook or a job description, you just figure out what needs to exist and try to go build it. And that's the muscle I try to use at Kapa every day. When you go out and do stuff on your own, you also quickly learn that some things (for the lack of a better word) you just need to "hustle" into existence, which really is helpful when you're a part of a start-up, especially in the early days. It ties quite well into what the team learned doing YC back in 2023: Do things that don't scale.

5. You are doing this while studying at CBS. What has been the hardest part of balancing university with a demanding role at a fast-growing AI startup?

It's quite hard, and I'm not going to pretend I've found some perfect balance. The thing that makes it kind of work for me is that I decided from the start that Kapa was my number one priority, so it was never really a 50/50 split in my head. My bet is that the stuff I learn and get to do at kapa, is gonna help me in the long-run with what i'm personally trying to do. What makes it actually work in practice is that my course at CBS runs on a quarterly structure, so each course has a smaller curriculum than a normal semester. That means I don't have to spread studying thinly across months, I can go really intense and exam-focused in the run-up, late nights, full concentration for a compressed period, and still (hopefully) make it through.

6. If we sit down again in a year, how do you envision Kapa has evolved, and what do you think will look most different about the market?

I can obviously only speak to what we're seeing at Kapa, but I'd say agents is what comes up most. Not as a separate bucket from support, docs, and onboarding though. The real shift is from answering to doing, an onboarding assistant that checks a user's config and fixes the issue on the spot, a debugging copilot that pulls the real logs and diagnoses the problem. So support, documentation, and onboarding don't disappear, they collapse into one agent that actually has your product knowledge and can act on live data. We're already helping people build this on top of Kapa. They use the engine we've spent three years optimizing for answering technical questions, then hook their own company-specific tools onto it. We actually shipped an agent with 30 tools that hit our APIs and one tool that reads our docs, and the docs tool was the most-used by far, because the agent reads them to figure out which of its own tools to call. The knowledge layer is what makes the actions work.