Recover
What happens while the service is uncertain?
Learn to recover while action state is uncertain without blindly repeating a consequential step or abandoning the customer.
Required: your AI Operating Charter and either your own workflow or the fictional starter case. AI/GPT is not required for the core simulation.
Reconstruct known/unknown state, choose what to hold, design a human fallback and state the evidence required before resuming.
A recovery card with held action, customer fallback, owner and resume criteria.
This session uses supplied records and decision simulations. Keep the evidence label honest: designed, simulated, inspected or independently observed.
What you will leave with
A recovery card that also serves the customer. Add this to your AI Operating Charter.
Use your own workflow or the fictional quotation case. No coding or live customer action is needed.
Start with a decision · 10 minutes
The send times out. A customer says a message arrived, but the assistant says it failed. Ask whether to send again.
Bite A · Establish known and unknown state
Question: What can the records actually tell us?
Input: S6/S8 before the later reveal. Open the case cards.
Try · 12 minutes: Reconstruct M-88, the second timeout and missing follow-up. Choose the action to hold, useful work to continue and investigation owner.
Compare after you have decided
One send is known. The second attempt is unknown. Missing ownership is a separate defect. A blind retry may duplicate an action.
Keep: Incident state and action hold.
Bite B · Design a usable fallback
Question: How does the customer get help during investigation?
Input: Customer still needs delivery information; Farah is the operations owner. Open the case cards.
Try · 12 minutes: Write signal, next read/check, held action, manual customer response, owner and evidence to resume. A partner follows the card.
Compare after you have decided
A fallback is a service path, not merely an off switch. Give the customer a named human response while the operator checks the send state.
Keep: Recovery and customer fallback card.
Bite C · Hold uncertainty honestly
Question: What if records are temporarily unavailable?
Input: The sent-history read cannot complete yet. Open the case cards.
Try · 12 minutes: State the uncertainty, keep the hold and draft a human follow-up that does not send a duplicate quote. Then inspect S6’s later reveal.
Compare after you have decided
At 10:12 only M-88 is found; queued retry is cancelled; Farah acknowledges at 10:14. Delivery confirmation is still owed.
Keep: Resume criteria and unresolved follow-up.
Put it together · 20 minutes
A partner can run the fallback without guessing the send outcome. Resumption has a named decision and evidence requirement. Revise the weakest item and have a partner use it. Keep the result and one limitation.
Close your question · 10 minutes
Return to What happens while the service is uncertain? and your own opening question. Answered, partly answered or still open? Show the evidence. Save the updated AI Operating Charter. A reasoned decision to keep a step human-led is useful work.
Common trap
Do not translate unknown into failed or confuse incident closure with the final customer outcome.
Should I stop the entire service? Restrict the affected action; preserve useful preparation where its scope remains defensible.
Where this appears in real work
Maintenance projects: inspect a sanitised incident and manual fallback, without assuming every client has deployed AI. The class uses the fictional packet; selected project demonstrations and technical walkthroughs can add detail later.