Last updated By Sho
Run AI
Eight practical sessions, two evening studios and one AI Operating Charter.
Learn how to decide whether an AI capability is dependable enough to run, and how ownership, assurance, recovery, evidence, cost and change decisions fit together.
Required: computer or phone with AI/GPT for any activity marked Prompting exercise. Optional: one AI-assisted workflow or service you want to reason about. Otherwise use the supplied fictional starter case.
Open your AI Operating Charter, record your question, choose own work or the starter case, then begin Session 1.
An AI Operating Charter, a reasoned operating decision, supporting evidence, unresolved gaps and the next review or action.
Start this course
- Open the AI Operating Charter and fill in “My question”.
- Choose your route: your own AI-assisted workflow/service, or the fictional starter case.
- Begin Session 1: Trust. Most core exercises are evidence/decision work; AI/GPT is used only where explicitly marked.
This is a complete 3-day, 2-night programme: 16 daytime hours + two 60-minute evening studios = 18 guided hours. No production deployment is required for the core route.
One day. One clear step forward.
Define & bound
Decide what to rely on, who owns it and where authority stops.
Trust → Own → Bound
Open sessions 1–3 →Assure & observe
Test the promise, practise recovery and reconstruct what happened.
Prove → Recover → Trace
Open sessions 4–6 →Improve & decide
Judge value and decide whether a changed service may continue.
Measure → Release
Open sessions 7–8 →WHEN ONE RESPONSIBILITY BECOMES A SYSTEM.
Run AI is where Graph Engineering belongs. Once several loops, tools, people or AI responsibilities share state and authority, the relationships between them become something you must operate — not merely draw.
A graph can execute perfectly and the business can still lose. Run AI adds ownership, real-world anchors, recovery, traceability, measurement and the decision to increase, hold, reduce or stop autonomy.
Your three days and two nights
| When | Day 1 | Day 2 | Day 3 |
|---|---|---|---|
| 09:00–11:00 | Session 1 | Session 4 | Session 7 |
| 11:15–13:15 | Session 2 | Session 5 | Session 8 and final proof |
| 14:15–16:15 | Session 3 | Session 6 | Programme completed |
| 19:30–20:30 | Night 1: Trust review clinic | Night 2: Incident and value clinic | — |
Breaks, meals and rest sit between these blocks. The host supplies dates, venue and any local start-time adjustments. The evening studios develop work you already started; no late-night technical build is needed.
Find a session
- Session 1: Trust — What evidence would make me rely on this?
- Session 2: Own — Who owns the outcome when something is wrong?
- Session 3: Bound — What may continue, and where must it ask?
- Session 4: Prove — What must this service handle correctly?
- Session 5: Recover — What happens while the service is uncertain?
- Session 6: Trace — Can I explain what happened when it is disputed?
- Session 7: Measure — Is this useful enough to continue?
- Session 8: Release — What changes when the source or behaviour changes?
What you take home
Your AI Operating Charter, an output or decision with evidence, one remaining limitation and a next step. You will return to your opening question and explain what changed.
If you miss an evening studio, use its card and the next morning’s reconnect. A printed or paired route remains available. Open the workbook to save, copy, download or print your work.
WHEN PEOPLE START RELYING ON THE SERVICE.
Commerce gives Run AI a concrete question: can customers and staff depend on an AI-assisted service on an ordinary working day, including when something is wrong? ACOM is one business context for ownership, assurance, recovery and decisions about whether AI has earned trust.