Responsibility
Choose a narrow job, define its limits and test when it must ask for help.
See how Day 2 moves from a narrow responsibility to explicit limits, tests and a useful escalation package.
Open your AI Work Delegation Blueprint. AI/GPT is not required just to use this day map; the linked sessions use guided simulations.
Review Sessions 4–6, then open each full session to define the job, bound its authority and test when it must escalate.
A bounded responsibility, control evidence and a tested escalation package with the human work still visible.
The question for this day
What responsibility can move without more chasing? By the end of the arc, you will have A tested escalation package and human-work map. Keep using the same task or fictional case so each session improves the previous work.
Session 4 · Delegate · 120 minutes
Question: What responsibility can move without more chasing?
You keep: One narrow responsibility with a human owner.
| Bite | Decision you practise | Evidence you keep |
|---|---|---|
| Describe the outcome, not the role title | What useful state should the work reach? | Outcome and remaining human work |
| Choose a first responsibility | Which job is useful and observable enough to try? | Candidate choice and reasons |
| Make the job small enough to explain | What does “AI salesperson” actually mean here? | Responsibility statement with a blocked state |
Open the full session and exercises.
Session 5 · Bound · 120 minutes
Question: How much authority is enough for this job?
You keep: A responsibility card with limits and evidence of control.
| Bite | Decision you practise | Evidence you keep |
|---|---|---|
| Make “done” observable | When is the responsibility complete? | Completion criteria and not-done conditions |
| Compare a written rule with a control | What stops an out-of-scope action? | Allowed, forbidden, approval-required; control evidence type |
| Check the whole definition of done | Which receipt supports each status? | Done criteria tested against a counterexample |
Open the full session and exercises.
Session 6 · Escalate · 120 minutes
Question: When should AI ask for help, and what should it bring?
You keep: A tested escalation package and human-work map.
| Bite | Decision you practise | Evidence you keep |
|---|---|---|
| Choose a task-specific mode | What may this task do now? | Mode, scope and evidence needed |
| Test normal, unclear and forbidden requests | Which failure blocks more authority? | Three-case test with a release blocker |
| Escalate a decision, not a vague problem | What can the next person act on? | Escalation package and retained human work |
Open the full session and exercises.
Show the day’s evidence
Open your AI Work Delegation Blueprint. Ask a partner to use it without your explanation. Mark one unclear field, revise it and preserve the supporting record. Do not substitute confidence for proof.
Tonight · Handoff rehearsal
A handoff draft and failure card ready for Session 7. Use the responsibility card and three-case test. Role-play the person receiving an escalation. Remove an owner acknowledgement; record what they need and what must wait.
Open the 60-minute studio and its card. Tomorrow starts from this evidence.