Доклад

A Year Running AI Agents in a Real Team: What We Delegated, What We Took Back

Спикеры

Кому полезно

Team leads, founders, and engineering managers deciding how far to hand real work to AI agents, and what it does to roles, workload, and cost. For anyone past the demo stage who now has to make delegation-and-supervision calls for a team. No agent-building background needed, as the talk is about the management decisions, not the models.

О чём доклад

The question everyone asks is whether agents replace engineers. The question a lead actually has to answer is narrower and harder: what can you hand to an agent, what do you have to take back, and what does it do to your people's roles and your costs?

We spent a year both building an agent product and running agents inside our own small team. Some things we delegated stuck. Some we delegated and had to pull back, sometimes after they had already caused a mess. The human work didn't disappear; it moved up, toward specification, judgment, and verification, and that shift is a real change in what the job is.

This talk is the honest account of what we handed over, the one thing we revoked and why, how roles changed, and where the ROI was real versus where supervision quietly ate the gain. This is not a demo, but rather a year of management decisions.

Видео