AI shrinks the team, not the manager's job

I want to be upfront: this is a bet, not a playbook I've run. I haven't seen an org do this end to end, and I'm not certain anyone will. But the more I watch what AI does to the size of a capable team, the more I think this is the shape companies will need if they want to keep driving products forward.

The lazy conclusion about AI and teams is "we'll need fewer managers." A small group with good tooling ships what a big group used to, so cut the overhead and let the agents fill the gap. I think that gets it exactly backwards. AI shrinks the team. It does not shrink the manager's job. It splits that job cleanly into two, and most orgs are bad at both because they keep mashing them into one role.

The two jobs are about and people, not tasks

There is a direction job and there is a people job, and they answer different questions.

The direction job is "what are we doing and why does it matter." That is the product owner. They drive the goals and the initiatives, they hold the why, and they make sure the thing the team is pointed at is a real thing the company needs. Not a backlog of tickets. A goal worth jumping on.

The people job is "how do you grow." That is the engineering manager, and AI does not touch it. Careers don't write themselves. Someone still has to know what a person wants to become, give them the work that stretches them, and clear the things blocking them. Think of an EM less as a task allocator and more as a gardener. You don't yank a cherry tree to make it bloom. You give it light, space, and time, and you protect it from the things that would stunt it. People are the same. That work is slow, human, and not automatable, which is exactly why it survives the AI cut while ticket-routing does not.

When one person tries to do both, one always loses. Usually it's the people job, because the direction job is louder and has a deadline.

Smaller teams, faster loops, their own ceremonies

Once AI lets a group of three or four do what eight used to, the team gets a property big teams can't buy: the whole context fits in everyone's head. Communication stops being a coordination tax. You don't need the heavy process that exists only to keep a large group in sync.

So let small teams run their own ceremonies. A standup for two people is theater. A planning ritual built for a twenty-person tribe is a cost a four-person team should refuse to pay. Give them the goal and the autonomy to decide how they coordinate, and they'll land on something that fits - often a five-minute conversation instead of a ceremony. The point of a ritual is the loop it creates, not the ritual. Smaller teams get tighter loops for free, so they need less of the ceremony and more of the loop.

The real motivation: drive something real, together

Here is what this structure actually buys you, and it isn't speed. It's motivation.

People are motivated by driving a real thing forward and by helping each other get there. Not by velocity charts. When the PO has named a goal that genuinely matters, and the team is small enough that everyone can see their hand on it, and the EM has made sure each person is growing while they do it, you get the thing every engineer actually wants: the sense that the work counts and that the people next to you have your back.

Never stranded at "I don't know what to do"

The failure state I want to design out completely is the engineer who opens their laptop and thinks "I don't know what to do." That is pure waste, and it's always a leadership failure, not an engineer failure. It means the direction job failed (no clear goal) or the people job failed (no clear growth path, or a blocker nobody cleared).

A small team with a PO holding a real goal and an EM holding the person should never reach that state. There is always a next thing that moves the goal, and there is always someone whose job is to make sure you can do it and grow while you're at it. AI gives you the leverage to keep the teams small. Two clear leadership jobs are what keep them pointed and whole.