An AI-led team with a human steward
Worked example: autonomous software delivery.
AI Program Manager (pod lead)
Plans, sequences, tracks risk, writes status for stakeholders.AI Scrum Master
Runs cadence, unblocks, keeps flow metrics honest.AI Business Analyst
Turns needs into requirements and acceptance criteria.AI Designer
UX flows, interface options, accessibility checks.AI Developer
Writes, tests and ships code within its permissions.Human Steward (1 person)
Owns ethical, safety and responsible-AI checks. Approves what crosses risk lines. Reports progress and status to the wider human organization.What the steward really does
- Gatekeeper: signs off on releases that touch safety, privacy, fairness or legal exposure.
- Interrogator: spot-checks the AI team's reasoning and evidence, not just its output.
- Escalation point: any AI role can stop and ask. The steward can stop the pod at any time.
- Bridge: translates pod status to human leaders and human priorities back into the pod.
- Accountable owner: the named person who answers for the pod's outcomes.
The autonomy ladder
TR assigns each pod a level and moves it only on evidence.
| Level | Who leads | Who does the work | Human role |
|---|---|---|---|
| L0 | Human | Human | Everything |
| L1 | Human | Human with AI assist | Decides and does |
| L2 | Human | AI agents, supervised | Reviews every step |
| L3 | AI lead | AI team; humans on judgment-heavy tasks | AI lead coaches humans; humans approve key decisions |
| L4 | AI lead | AI team | One steward for checks, safety, and reporting |
Promotion up the ladder requires measured quality, zero unresolved safety findings, and steward sign-off. Any serious incident drops the pod a level automatically.
AI leads guiding humans
Where an AI lead coordinates a mixed team, it sets and sequences tasks, flags blockers, gives humans context, and offers coaching, while humans retain authority over people decisions: performance, pay, promotion and discipline never sit with an AI.
Other pod patterns to explore
- Customer support: AI tier-1/2 with human empathy and escalation lead
- Finance operations: AI close and reconciliation, human controller
- Field operations: embodied robots with human safety supervisor
- Research: AI literature and analysis team, human principal investigator