What Czech AI founders actually did this month
Findings · Jul 2026 · 6 min
We keep a standing watch on the Czech and Slovak AI founder scene inside Aim — the people, not the press releases. Skip the think-pieces: here are the founders who actually shipped, raised, or moved something worth noting this past month. A new fund with a hard thesis, a security researcher matching the field on zero-days with open models, a hundred-fold cut in agent costs, and a handful of companies that quietly grew up. Every link opens the full briefing in Aim.
Tomáš Čupr: AI as infrastructure, not a layer
On CzechCrunch's Money Maker podcast, Tomáš Čupr — founder of TCF Capital and co-founder of the AI-native company duvo.ai — argued AI stops being interesting as a layer on top of your processes and starts mattering when it becomes the infrastructure underneath them. His test for whether it's working: it changes the operating system of the business, not just the tooling.
He sees the same shift across TCF's portfolio, frames AI transformation as a leadership problem rather than a software purchase, and predicts people will soon talk to computers by voice more than keyboard. On where the region stands, he was direct about what Silicon Valley does better — and where Europe still holds an edge.
AI becomes valuable when it changes the operating system of the business.
Kaya VC draws an 'AI divide' — and puts CZK 2 billion behind it
Launching the firm's fifth fund at nearly CZK 2 billion, Kaya VC partner Tomáš Pačinda argued AI will amplify the best founders and derail the rest — it will "nakopne" some and "vykolejí" others — and that the real VC skill is still wiring money into projects before they turn "sexy a trendy."
Kaya has earned the right to the thesis. It was Rohlík's first institutional investor back in 2014, an early backer of DocPlanner (the first Polish unicorn) and Booksy, and counts Better Stack and E2B among its Czech bets. Its newest is Viktor, a virtual AI worker the fund says is growing very fast.
Stanislav Fort matches the field on zero-days — with open models
Stanislav Fort — founder and chief scientist at AISLE, with a Stanford AI PhD, Cambridge physics, and time at Anthropic and DeepMind — said his small European team matched Mythos on public zero-day vulnerabilities using widely available, open-source-derived models, and that the system can run air-gapped if needed. He cited a Berkeley study ranking the team #1 globally in three of eight categories.
It's a claim the region rarely makes out loud: frontier-grade AI security, built lean, out of Europe.
David Mlčoch cut one agent's bill from $10,000 to $100
Running browser agents at scale, David Mlčoch named cost and maintenance as the two things that actually bite. His fix is almost embarrassingly simple: use an expensive model to build the agent's declarations, then hand execution to a cheap one. On a single agent, that took the cost from $10,000 to $100.
The maintenance half is harder. Some of his agents have run more than 20,000 times, so any behavior change risks regressions — which means every update has to pull context from a representative sample of past runs. Even adding "memories," he argues, still needs genuinely capable models.
Vojtech Oravec sells inside the AI answer — where analytics can't see
On exec talks #83, Ranketta founder Vojtech Oravec made the case that AI already sways more than half of purchase decisions while Google Analytics attributes barely 1% of traffic to it. The channel is real; the measurement simply hasn't caught up.
Working with around a hundred brands to sell directly inside ChatGPT, Gemini, and Amazon Rufus answers, he says AI-referred traffic converts far better than Google's — up to 15× in some categories — and that Google and ChatGPT recommendations overlap only 13% of the time. Different channel, different rules.
Eva Kuttichová wants to add ten years to your active life
In a Czech-language interview, Eva Kuttichová traced her path from a Pardubice industrial school through STRV and Silicon Valley and back to Prague, where she's building a startup around a single mission: "help extend people's active life by ten years." A reminder that the region's AI ambition isn't all agents and infrastructure.
And the companies that grew up quietly
Beyond the founders, a cluster of the region's tracked AI companies made moves worth logging:
Rossum was acquired by Coupa, capping a two-and-a-half-year partnership between the document-processing company and the spend-management platform.
Ataccama joined the Databricks Marketplace, extending its data-trust controls from source systems into AI workloads.
Dataddo launched a free tier — it says 1,300+ companies already move 21 million rows a month through it, and pitches data reliability, not models, as the real AI bottleneck.
Born Digital shipped an AI sales trainer built on digital personas that let salespeople rehearse objections and closings before they ever pick up the phone.
Resistant.ai warned that AI receipt generators have turned document fraud into a scale problem, with hundreds of free tools able to produce convincing fakes by the thousand.
What it means
No single narrative ties these together, and that's the point. The Czech and Slovak scene isn't waiting for one big story — it's a fund drawing a line, a security team punching above its weight, an operator saving two orders of magnitude on agents, and a handful of companies compounding.
We'll keep watching this space at Aim. The interesting moves usually show up here first.