AI adoption growing pains

AI-first company growing pains

We seem to have become an AI-first company

Our 18 developers now merge 3x more PRs than a year ago, and the bottleneck has shifted from building features to choosing the right ones. This led to structural changes:
- teams shrank to 1–2 people
- one engineer usually builds a whole feature
- human code review is optional
- engineers report on feature adoption and client feedback

Our engineers have owned their features and talked to clients for years. Now they do it far more often. Not everyone enjoys frequent conversations with clients, and hiring people who are strong in system design, communication and ownership is hard.

I will be happy to share our journey and hear how your teams deal with this.

AI adoption across 80K+ teams: patterns and pitfalls

We watch around 80,000 teams adopt an AI employee, and it gives a rare view of how AI adoption really happens inside organizations, whatever the team does.

- Teams end up treating AI more like a colleague than like software
- teams that give it full context, data and tools multiply their productivity; teams that delegate halfway get friction instead
- most often, AI is used to unblock dependencies on other teams rather than for the team’s own tasks
- teams rebuild internal tools from scratch with code nobody ever reviews, and over time it can turn into a mess that someone has to untangle
- even with all the right information, roughly one task in nine still fails on judgment

What nobody knows yet: if every company gets the same boost at the same cost, where is the advantage?

AI is not a tool, is it?

This talk is less about practice and more about a way of thinking.

Most people still see AI as a tool: an IDE on steroids that helps a specialist work faster. I think that view is already outdated. A tool makes one specialist do the work of ten. But a system that solves optimization problems, which AI is today, is something bigger than a tool.

I don’t have a comfortable answer to where this leads us. So I’d rather bring the question to the room: how do you treat AI today at your organization?