On September 8, AI Builders Warsaw held its second edition, and this time the focus shifted from experimentation to something harder: what it takes to run AI inside a real organization.
Why Enterprise needed its own edition
The Launch Edition in August was about builders showing what they'd made – agents, gateways, voice assistants. Enterprise AI asks different questions. It's not about whether something works in a demo; it's about whether it survives contact with a regulated industry, a skeptical client, or a system where a mistake actually costs something. That's a specific enough problem that it deserved its own room, and its own speakers.
Inside the Enterprise Edition
The format stayed the same – no vendor pitches, nothing pretending to be a TED talk – but the subject matter got noticeably heavier. All three speakers talked about AI in contexts where getting it wrong isn't a minor inconvenience.
- Piotr Jabłoński opened the night with How to Avoid Fuckups When Deploying AI in a Regulated Sector, walking through what actually goes wrong when AI meets compliance requirements, and how to see it coming before it happens.
- Maciej Antoni Papiewski followed with You Don't Bully Your Agent Enough – Building Safe Agentic Systems at Scale, on the discipline it takes to keep agentic systems from doing something you didn't ask for, once they're running at scale rather than in a sandbox.
- Kacper Wikieł, from Fabryka AI, closed with The Pipe Is a Cylinder: Lessons from Shipping Safety-Critical Computer Vision – a look at deploying computer vision for pipeline inspections, where the model isn't just wrong or right, it's the thing standing between critical infrastructure and a failure nobody wants to explain later.
Three different problems, one common thread: none of them were talking about AI in theory. They were talking about what happens after it ships.
A full house, again
The room was at capacity again – enough that a waitlist opened before the event even started. New faces showed up alongside people who'd been at the Launch Edition in August, which is usually the sign that a community is starting to hold its own shape rather than depending on novelty to fill the room.
What this means, one edition in
What stood out this time was the shape of the conversation in it. Nobody asked how to get started with AI anymore. The questions were about what happens after: how do you explain a model's decision to a regulator, what do you do when a client doesn't trust an agent's output, how do you ship computer vision into a system where being wrong has real consequences. Those are harder questions than the ones August's room was asking, and a full house still showed up to sit through them. That says something about where AI adoption in Poland actually is right now – past the demo phase, into the part that's harder to talk about and harder to get right.
What's next
The next edition lands on October 13. If September was about the harder side of enterprise AI, expect the series to keep leaning into problems that don't have easy answers yet – that's where the format seems to work best.
Gallery
Join AI Builders Warsaw!
AI Builders Warsaw is a community of thousands of builders who organize over 100 meetups and hackathons each year in cities around the world, including Poland. Get involved with the Warsaw edition!
Author
Marta Miler
Connects technology with the communities and audiences it's meant to serve. Also the kind of person who notices when a sentence sounds smart but says nothing, then rewrites it until the answer is obvious.