Skip to content

Knowledge base

Product Management

When agents do the building, the scarce skill moves from building a feature to knowing which feature is worth building and saying precisely what it must do. Product management gets more important: the specification and the prototype become what the agents work from.

Written by

Updated:

Knowledge base 9

Our position. We give the product owner the coding agent too: a product owner with the agent can ask the codebase how a feature works today before writing a word about changing it.

When the coding barrier falls, the constraint moves upstream. A vague ticket used to be absorbed by an engineer who asked questions. An agent implements the ambiguity.

What does product work hand to the agents?

Product intent now arrives as something a machine can check. When a specification and a test disagree, the test wins, so acceptance criteria carry more weight than the prose around them (Specifications and context). Before launch, requirements can come from synthetic users, behavioral models of the target group, provided real user behavior replaces them once the product is live (Synthetic First, Back-Testing Later). Where an organization runs a software factory, a domain expert can commission a solution directly by describing it in a chat (Citizen Development as a Platform).

How do you know the product work paid off?

Judge it by delivery and business outcomes, not by how much the agents produced. Pair the AI metrics, such as adoption and AI-assisted share, with the delivery metrics that predate AI, such as throughput and change failure rate (Measuring AI impact). The operating model has to change with it: McKinsey describes smaller pods and spec-driven workflows among the top performers (Moving away from Agile).