How does your team actually collaborate now that everyone works with an AI agent?

AI made each of us individually much faster. The handoffs between us didn't change at all — and I think that's where the time is going now.

What a normal week looks like on our team:

• A teammate's agent is halfway through a fix. The only way to show me is a screenshot or a read-only link I can't type into. So I start over in my own session.

• A long task stalls overnight waiting on one approval that only the person who started it can give.

• Reviewing AI-written code means reading the diff without any of the reasoning that produced it. The why lives in a transcript nobody else opened.

Individually we're all 3x faster. As a team it feels roughly the same speed, just with more context to reconstruct.

A few things I'd genuinely like to hear from other teams:

1. When a teammate's agent session is stuck or heading the wrong way, what do you actually do? Jump into their machine, or restart it on your own?

2. Has AI changed how you review each other's code — do you look at the prompts/reasoning, or only the diff?

3. Has anyone landed on a handoff ritual that works? Shared prompt libraries, committed session logs, pairing on one agent, something else?

Disclosure so it's on the table: we build Bottari, which puts multiple people in one Claude Code session, so I'm not a neutral party here. But I'd rather hear what's already working for you than assume we've got the shape of the problem right.