Smaller than the org chart

The early-stage teams I talk with are smaller than they would have been five years ago, and they are not smaller because they are earlier. Three people run what used to need a platform team, a data team and someone whose whole job was the deploy. Agents write the first draft of most things. The infrastructure is rented by the hour. The data is a query away rather than a project.

What has not shrunk is the number of decisions. If anything there are more of them, and they arrive sooner, because the cost of building the wrong thing has dropped to almost nothing and so the wrong thing gets built on Tuesday instead of next quarter. Agents lower the price of doing. They do nothing for the price of deciding.

So the scarce thing in these teams looks to me like judgment rather than hands: judgment about what not to build, which boundary to draw while it is still cheap to move, which capability to refuse permanently rather than police forever. That is the same question I keep arriving at from the other direction, building something on my own and letting agents do the delivery: the more the typing is handled, the more the whole job is knowing what to type.

I hold this as a working theory rather than a conclusion, because I am seeing a selected slice. The question I cannot yet answer is what a team of three needs that a team of thirty never had to think about.

In the index underproductconstraintsagents