Verify
Did the work actually finish?
Learn to distinguish claimed completion from verified state, keep unknowns visible and choose a recovery step that does not duplicate a consequential action.
Required: your AI Work Delegation Blueprint and either your own workflow or the fictional starter case. AI/GPT is not required for the core simulation.
Inspect the supplied action records, classify what is complete/incomplete/unknown and decide the first check before retrying anything.
A three-state action receipt and recovery decision with a named investigation owner.
This session uses paper, partner role-play and supplied evidence. Do not treat the simulation as a live customer action.
What you will leave with
An action receipt and recovery decision. Add this to your AI Work Delegation Blueprint.
Use your own workflow or the fictional quotation case. No coding or live customer action is needed.
Start with a decision · 10 minutes
Show “sent and logged” beside M-88 and a CRM record with no follow-up owner. Ask whether the customer outcome is complete.
Bite A · Read the state behind the claim
Question: Which parts are complete, incomplete or unknown?
Input: S6 before its later reveal. Open the case cards.
Try · 12 minutes: Classify the approved send, follow-up ownership and second send attempt. Cite a record for each judgement. Keep uncertainty visible.
Compare after you have decided
M-88 proves one send. The blank owner means follow-up is incomplete. The second timeout leaves that attempt unknown.
Keep: Three-state action receipt with supporting records.
Bite B · Match the fault to the first check
Question: What would you inspect before fixing it?
Input: Training faults: timeout, duplicate, wrong recipient and stale price. Open the case cards.
Try · 12 minutes: Match each fault to a useful first check and responsible person. Explain why a single “try again” response would make some faults worse.
Compare after you have decided
Read sent history for a timeout; reconcile receipts for duplicates; hold and notify the owner for wrong recipient; check approved source/version for stale price.
Keep: Failure, first check and investigation owner.
Bite C · Recover without guessing
Question: When is a repeat action justified?
Input: S6 initial records: one send proven; the second attempt unknown. Keep the later reveal closed until you record a recovery decision. Open the case cards.
Try · 12 minutes: Before opening the reveal, record what is held, the next read, investigation owner and customer fallback. Then revise the status using the later evidence.
Compare after you have decided
Do not blindly resend. One send is verified and the queued repeat is cancelled. Handoff is acknowledged; the delivery question remains open.
Keep: Verified state, unresolved work and next owner.
Put it together · 20 minutes
The receipt distinguishes known completion from missing ownership and uncertainty. The recovery does not duplicate the consequential action. Revise the weakest item and have a partner use it. Keep the result and one limitation.
Close your question · 10 minutes
Return to Did the work actually finish? and your own opening question. Answered, partly answered or still open? Show the evidence. Save the updated AI Work Delegation Blueprint. A reasoned decision to keep a step human-led is useful work.
Common trap
Unknown is not failed. A successful send does not prove the entire customer outcome.
What if I cannot read the records? Hold the uncertain action and give a named person the follow-up; do not manufacture a result.
Where this appears in real work
Alamak publishing: distinguish a prepared asset, publishing acknowledgement and the actual public result. The class uses the fictional packet; selected project demonstrations and technical walkthroughs can add detail later.