Translate a report into a decision narrative while retaining source traceability and material limits
8 min setupTested with:ChatGPTReviewed: 2026-08-28
1
Add your context
Your text stays in this browser. AILesson Prompts does not send it to a model or server.
2
Your prompt
Unfilled fields remain visible as placeholders, so you can still copy and edit the prompt
Condense the report into an exactly ten-slide presentation outline.
Report:
[report]
Audience, decision, knowledge, objections, and setting:
[decision]
Duration, ten-slide rule, required findings, citation, appendix, and exclusions:
[constraints]
Use only report-supported claims. Build a decision narrative rather than copying the table of contents. For each of ten slides provide assertion title, one message, 1–3 supporting facts, source locators, recommended visual, interpretation boundary, and speaker transition. Preserve denominators, periods, definitions, uncertainty, subgroup differences, dissent, and recommendations' status. Do not average incompatible measures or turn association into causation. Return a narrative spine, slides 1–10, appendix/backup list if allowed, material omissions with rationale, claim-to-source matrix, and a fidelity audit. If ten slides cannot support the requested breadth honestly, recommend scope cuts rather than shrinking every point into unreadable text.
Private by defaultPrompt assembly happens locally in your browser. Avoid placing confidential information into any AI service unless your organization allows it.
From input to outcome
A worked example
See how concrete context turns this recipe into a usable result
Actual input
Complete report
[Executive summary] A 12-week flexible-shift pilot covered 4 of 11 service teams. [p.6 Table 2] Scheduled overtime fell from 410 to 352 hours (14.1%) in pilot teams; non-pilot teams fell 3.0% over the same period. Teams were not randomly selected. [p.9] Customer response time was 4.8 hours before and 4.9 after; target ≤5.0. [p.12] 96 of 143 pilot-team employees answered the survey; 71% preferred continuing. Night-shift response was 18 of 39 and preference 50%. [p.15] Two managers reported planning complexity. [p.18] Finance estimates £21k annual savings if the overtime difference persists; estimate excludes training and scheduling software. [Section: Recommendation] Extend six months to all service teams with monthly review. Union consultation is incomplete.
Audience and decision
Executive committee decides whether to extend, modify, or stop. They care about service, cost, workforce fairness, and implementation risk; 15-minute presentation.
Slide constraints
Exactly ten slides, no appendix in the meeting but a backup note list is allowed. Cite page/table. Keep selection limitation, survey denominator/night subgroup, excluded costs, consultation status, and recommendation as recommendation—not decision.
Example output
Spine: observed pilot results are promising but not causal; service stayed within target; workforce and implementation evidence is uneven; choose a controlled next step.
1 “Decision today: extend, modify, or stop the shift pilot”—three options and criteria [Executive summary/Recommendation]. 2 “The pilot covered 4 of 11 non-randomly selected teams”—scope visual; boundary: limited generalizability [Executive summary, p.6]. 3 “Pilot-team overtime fell 14.1%, versus 3.0% elsewhere”—410→352 comparison; boundary: association, not causal estimate [p.6 Table 2]. 4 “Response time remained within the ≤5.0-hour target”—4.8→4.9; do not call improvement [p.9]. 5 “71% of 96/143 respondents preferred continuing”—denominator and non-response [p.12]. 6 “Night-shift evidence is weaker and less positive”—18/39, 50%; fairness question [p.12]. 7 “Planning complexity is an identified implementation cost”—two manager reports; prevalence unknown [p.15]. 8 “£21k savings estimate excludes material costs”—training/software exclusions and persistence assumption [p.18]. 9 “Union consultation is an unresolved launch dependency”—status and decision gate [Recommendation]. 10 “Recommendation: controlled six-month extension with conditions”—monthly service/overtime/fairness review, cost baseline, consultation, stop criteria; label as proposed.
Omitted: unprovided methods/details, not invented. Backup notes: exact definitions, team selection, survey instrument, cost model. Fidelity matrix maps every claim to listed locators.
Why this works
1
Decision narrative prioritizes report evidence by audience use rather than document order.
2
Claim-to-source mapping guards against meaning drift during aggressive compression.
Check the result
Are there exactly ten slides with one core message each?
Can every number and material claim be traced to the report?
Are omissions, limitations, subgroup differences, and recommendation status visible?
ChatGPT