Practical guide
AI fluency for chiefs of staff: build a decision log
Turn fictional multi-owner updates into a decision log that separates facts, status, requested decisions and revisit triggers.
Chiefs of staff can use AI to reconcile updates while preserving who owns each fact and decision. This fictional leadership exercise produces a decision log, not an executive instruction, a meeting transcript or proof that a project is on track.
Separate updates from decisions
A polished summary can hide whether leadership actually chose anything.
Three fictional owners report that a launch date is proposed, a vendor review is incomplete and training capacity is unconfirmed. Record each as an attributed status. None is a decision merely because it appears in a leadership update.
Ask AI to classify statements as fact, proposal, dependency or decision request. Verify every classification against the supplied note and retain the speaker. Do not convert silence or attendance into approval.
Name the decision and its boundary
A decision log should state exactly what becomes true if the decision is made.
The fictional decision is whether to reserve a provisional launch window, subject to vendor review and training capacity. It is not approval to launch, spend or communicate externally. Write that boundary beside the requested choice.
Create options to reserve, defer or request more evidence. Keep consequences conditional and do not invent cost, demand or operational outcomes. The accountable leader still owns the choice.
Reconcile conflicting owner updates
Different timestamps can explain apparent disagreement.
Operations says training capacity is open as of Monday; the learning owner says it became constrained on Tuesday. Preserve both dated statements and treat Tuesday as the newer status, while asking whether scope changed.
AI may align dates and owners, but it should not accuse either source of error. If two notes use different definitions of capacity, record the ambiguity rather than forcing a single number.
Add a revisit trigger
A provisional choice needs a condition that brings it back for review.
If leadership reserves the window, the fictional trigger is receipt of the vendor review or a change in available training places. Record the trigger, owner and latest useful review date without inventing reminders or sending messages.
A calendar date alone is weaker when the underlying evidence may arrive earlier. Ask what observable event changes the decision and what happens if it does not arrive.
Deliver a compact decision record
The final artifact should make authority and uncertainty visible.
Include decision requested, dated inputs, options, selected status, dependencies, owner and revisit trigger. Leave the selected option blank until supplied and retain superseded updates in the record.
Review traceability and decision boundaries in this exercise. The executive-assistant agenda guide allocates meeting attention; this page preserves the meaning and follow-through of a cross-functional leadership choice.
Audit the log after the meeting
A decision record is useful only if later updates do not erase its original basis.
Introduce a fictional post-meeting note that the vendor review will arrive later than expected. Append it as new evidence and activate the recorded revisit trigger. Do not rewrite the original decision as mistaken or imply leadership would have chosen differently. The log should show the choice that was made, the information available then and the condition that reopened it.
Ask AI to produce a concise change summary for owners, then verify that it names the affected decision and leaves unrelated workstreams unchanged. Record whether the decision remains active, is paused or awaits review only when the accountable owner supplies that status.
Sources and scope
NIST: AI risk management. Skills England: workplace AI foundations.
These references provide background, not validation or endorsement of this exercise. The case details, calculations and suggested review questions are original instructional material. Use them to discuss observable work, not to infer customer outcomes, professional credentials or performance in every setting. Before adapting the exercise, confirm the relevant facts, approved tools, data permissions and decision owners. If you change the case, revisit the expected answers and checks as well. These examples describe practice tasks, not a promise that a particular product includes the fictional features.
Sources: [1] [2]
Sources
- 1.AI RMF Core · NIST
- 2.AI foundation skills for work benchmark · Skills England