What AI quietly broke in my engineering team

3 things AI quietly eroded in my engineering team this year.

I spent the first half of the year chasing an AI speed-up that never came. Senior leadership expected development to get several times faster. Other teams spun up swarms of agents and built “AI factories”. We tried too. I believed in that speed-up as well, or at least I wanted to. When we couldn’t reach it, I got anxious, and then burned out.

Mid-year, while analysing why we couldn’t ship faster, I noticed something deeper: what AI had quietly taken from our team. These are my observations:

💸 The cost of building, which forced us to choose. When code was expensive, we couldn’t build everything. Now every idea arrives as a working demo, and every demo waits for a decision. The bottleneck moved from the people who build to the people who decide.

🪜 The ladder of experience, which gave people standing. AI makes less experienced people sound sure of themselves. Seniors still say “it depends”, because they know what usually breaks. Leaders like confidence. Guess who gets heard, and who fixes it later.

🙋 The habit of asking each other for help, which held the team together. The Python import error, the failing cloud deployment, the “why is this container out of memory?” now go to a chat window. It’s faster. But those small questions were how we got to know each other. You helped someone fix their deployment and ended up talking about the tennis racket behind their chair.

None of this shows up in a sprint report. These things used to do part of a lead’s job for free. Now they have to be built on purpose.

I’m still positive about AI, and we have to adapt. But adapting well means watching the costs as closely as the gains.

What stopped happening in your team this year that no metric would catch?

#AIEngineering #EngineeringManagement #Leadership