What is a human deliberator?
A human deliberator is what the human collaborator becomes once generative agents do the steps of the work: the person who owns the problem, the frame and the final judgment, and shapes the system trusted to turn intent into reliable outcomes.
Today's human collaborator works alongside AI on the output itself, prompting, editing, checking and fixing. A human deliberator never touches the steps. The agents that carry them out are usually contracted, and each vendor's agents run on other vendors' agents, several specialised layers down. The deliberator owns three things instead. The problem is the brief the contracted agents work to. The frame is the internal context that turns delivered work into value for the organisation's own customers. The judgment is the final call, and standing behind it. Deliberators, with the agents and tools they run themselves, make up an organisation's deliberation layer, which is defined by what the people keep rather than by what the AI carries. Seen from outside, the organisation becomes one value-adding step in a chain: work comes in, passes through the system its deliberators control, and leaves as the product or service only that organisation offers, and the vendors run the same setup one step down. Agents can prove the steps followed the specifications and cleared the regulatory checks. Whether the result deserves the organisation's name is the deliberator's call, and that call is the work.
As in
"Our project has four human deliberators and a couple of hundred agents, and agreeing the brief is the hardest part of the job now."
The closest thing today
The nearest things today are human oversight rules in AI governance, which keep certain decisions with a person, and the client side of outsourcing, where the brief goes out and the judgment on what comes back stays in. Neither has handed every step of the work to agents yet.
