The AI Delegation Cycle
What it is
The AI Delegation Cycle runs AI working sessions through the same five steps a competent leader already applies when delegating to a person. Define the work: state the task, the outcome wanted, the constraints, the standard of done. Choose the right worker: the model, the persona, the tools and the documents in working context are choices about which worker is being asked. Set the autonomy level: calibrate supervision to the familiarity of the task. Check in mid-stream: ask the model what it understood the brief to be, what assumptions it made, where it is least sure. Own the outcome: whatever leaves the leader’s hands is the leader’s.
What it is
The AI Delegation Cycle runs AI working sessions through the same five steps a competent leader already applies when delegating to a person. Define the work: state the task, the outcome wanted, the constraints, the standard of done. Choose the right worker: the model, the persona, the tools and the documents in working context are choices about which worker is being asked. Set the autonomy level: calibrate supervision to the familiarity of the task. Check in mid-stream: ask the model what it understood the brief to be, what assumptions it made, where it is least sure. Own the outcome: whatever leaves the leader’s hands is the leader’s.
Why it happens with AI
Three forces pull leaders out of the cycle when AI is the worker. Frictionlessness: the chat window invites a one-line prompt with no template and no colleague waiting for context. Speed misread as competence: a fast answer reads as confirmation that the brief was sufficient, when the speed is a property of the worker. Confusion about ownership: accountability slips into a gap when the worker is a machine, because the leader does not see AI as a worker with a defined relationship.
What working on it does, impact and benefits
Running the cycle converts AI sessions from vending-machine transactions back into delegated work. Briefs get written; check-ins get run; outputs get read at the leader’s standard before they leave the leader’s hands. The benefit is not slower AI use, it is AI use that produces work the leader is willing to put their name to. Leaders who run the cycle reclaim the quiet tax that thin briefs and unowned outputs were charging on every AI hour.
Canonical framework: virenlall.com/ai-delegation-cycle, the full ~600-word treatment of the five steps, the corrective moves and the three embedded habits.
“AI is a worker, not a vending machine; delegate to it the way you delegate to people.” — Viren Lall, Managing Director, ChangeSchool LDN (2026).