How AI changes the tech lead's role: roundtable recording (Category #AI4SDLC)

If an agent writes the code, what is left for the tech lead to do — and how do you avoid becoming the person who reviews every change the team makes, all on your own? On September 30, Rafael Tonakanyan from Sber and I discussed this at Podlodka Techlead Crew. Igor Antonov and Andrey Kuleshov moderated the conversation. I’ve collected the materials from the session on my website.

We discussed several questions that come up when agents become part of a team: - How do you measure the value of AI while accounting for the time spent reviewing and accepting its output? Token consumption and the amount of generated code explain little on their own. - Where can a junior engineer gain experience if an agent prepares the first version of a solution for them? - What should you review: the code itself, the task definition, or the agent's work process? We debated this — we did not reach a consensus on dropping code review. - What do you do with a queue of completed changes and old tasks that can now be implemented faster, even though that has not made them any more useful?

You can watch the recording on YouTube or read the written recap. I also prepared a longread setting out my own position: on architecture, agent permissions, team development, and measuring the effect of AI. It was written before the roundtable, so it represents my starting point for the discussion.

#AI4SDLC #Agents #Engineering #Leadership #Management