What should a CPO report instead of agent interactions?

Answered by Falk Gottlob · Reviewed · 1 min read

THE SHORT ANSWER

Verified jobs done: the count of customer jobs completed to a written definition of done, on every path, your interface, your agents, outside agents through MCP or your API, and your own people, reported with the cost per job beside it. Agent interactions count activity, and an interaction looks the same whether the job got done or was abandoned. On its November 9, 2026 call, monday.com reported over 75 million MCP requests from outside agents and 6 million interactions with its own agents, and Roy Mann said customers care about measuring the output they get out of credit. The CPO writes the definition of done, because the definition is a product decision, and each seat reports one input to the number.

Report the job, not the activity. Agent interactions, sessions, and request counts tell a board the product is busy. They don't say whether the customer's work got done, and they mostly cover surfaces you own, while more of the use now comes through agents you didn't build. monday.com showed the gap on its November 2026 call: over 75 million MCP requests from outside agents, 6 million interactions with its own, and a co-founder saying customers care about the output they get.

The replacement is one number, verified jobs done, with a definition of done the CPO writes and the cost per job beside it. Product, design, product marketing, and the field each own one input to it, which is the practical reason all four sit under the CPO.

The full argument, including the four seat inputs and the predictions, is monday.com Counted 75M Agent Requests. The Board Needs One Outcome Number., part 8 of The CPO Transition. The whole system, all five seats, is The Product OS.

SOURCES

THE LONG VERSION

RELATED ANSWERS