A
Define
Over Use Case
Owns the use case and the context that frames it. The loop starts from human intent, never from the machine alone.
A machine loop runs in the centre and a human layer sits above it, reaching down to control three of the stages. The model proposes and the model judges, but the human sets the goal, owns the standard, and authorises every change.
The Incitrix methodology is a closed loop of five stages. Humans own three of them: the use case, the rubric, and the feedback loop. The detail is laid out in the sections below.
The loop runs on its own most of the time. But at three points a human is in command. These are the moments where intent is set, the standard is owned, and change is authorised.
Owns the use case and the context that frames it. The loop starts from human intent, never from the machine alone.
Refines and signs off the model-proposed rubric. The standard of judgement stays human-owned, not delegated to the model.
Decides which Eval recommendations give effect to the upliftment. No change reaches the agent without sign-off.
Five stages run in a cycle. The machine sets up agents and evaluates them; the human frames the work, owns the rubric, and authorises what changes. Then it loops, so the agent is uplifted over its lifecycle rather than left to drift.
The workflow to make agentic, framed with full context.
Configure the agents, their roles and behaviour.
The criteria output is judged against. Model drafts, human refines.
Score behaviour against the rubric, with an LLM as judge.
Apply approved improvements (upliftment), then loop back.
A model drafts and a model judges, but a human sets the goal, owns the standard, and authorises every change.
This is what makes the loop trustworthy rather than a closed, self-reinforcing automation.
Bring a workflow you want to make agentic. We will show you the loop, the control points, and the assurance, mapped to your world.