Role Blurring in Product Development under Agentic Automation: The Concept of Residual Judgement (Part 1 of 2)

In agent-assisted product development, agents now produce most of the artifacts: requirements, acceptance criteria, code, tests, reviews and releases. Checking that the result matches the intent and actually helps stays with humans, and we hypothesize that its cost has dropped far less. This practitioner's research essay draws on the three-month history of one open-source project, keryx (383 units of work, 3,175 acceptance criteria), and on anonymised data from a team with a conventional division of roles.

The essay formalizes six problems and proposes a working concept of residual judgement: for a given set of tools, budget and acceptable risk, a residual remains for humans, consisting of criteria whose verification the tools do not provide and decisions reserved for humans on normative grounds. The unit of observation becomes the claim record: what is claimed, how it is checked, what state the check is in, and who is responsible. Five mechanisms make the residual visible inside the process; four are implemented in keryx. Seven predictions, each with the conditions under which it would be refuted, are fixed in a protocol published before confirmatory data collection and will be tested in Part 2.

The paper claims only what the records show: how the documentation of decisions and checks changed. Conclusions about costs, behaviour and role boundaries remain hypotheses. The research and the text were produced jointly by the author and AI agents; contributions are described at the end of the paper. A Russian version is included.