Create a structured register while preserving units, periods, targets, and attribution
6 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
Extract metrics, dates, and owners from the report.
Report:
[report]
Scope:
[scope]
Output and normalization rules:
[schema]
Use only explicit report content. Return separate Metrics, Dates, and Owners tables plus a Relationships table linking them. Metrics must retain name, value, unit, reporting period, actual/target/forecast status, comparison baseline, scope, and evidence. Dates must distinguish reporting dates, deadlines, forecasts, and historical events. Owners must distinguish accountable owner, contributor, approver, and contact. Use Not specified rather than inferring missing fields. Preserve ranges, approximately signs, currencies, and percentage versus percentage-point differences. Flag duplicate labels, conflicting values, ambiguous pronouns, and dates without a year.
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
Source report
[Q2 review, p.2] June active customers were 12,480, up 8% year over year; the Q2 target was 13,000. Retention was 91.2%, 1.5 percentage points below target. [p.4] Priya owns the retention recovery plan. Product and Support contribute. A revised plan is due 18 July and the COO will approve it. Finance forecasts annual recurring revenue of approximately £4.2m by 31 December 2026. The dashboard footnote lists active customers as 12,408.
Extraction scope
Extract performance metrics, targets, forecasts, deadlines, and named roles from the Q2 review.
Output and normalization rules
Use YYYY-MM-DD where the year is stated; otherwise keep the written date and flag it. Preserve original currency and units. Cite page labels; do not merge conflicting values.
Example output
Metrics: M1 Active customers | 12,480 | customers | June, year not stated | Actual | Q2 target 13,000; +8% YoY | p.2. Conflict M1b: dashboard footnote says 12,408; do not resolve. M2 Retention | 91.2% | percent | Q2 | Actual | 1.5 percentage points below target; implied target must not be calculated unless requested | p.2. M3 ARR | approximately £4.2m | GBP/year | by 2026-12-31 | Forecast | baseline not specified | p.4.
Dates: D1 18 July, year not stated | revised retention plan deadline | p.4. D2 2026-12-31 | forecast horizon | p.4. Owners: O1 Priya | accountable owner | retention recovery plan | p.4. O2 Product and Support | contributors | plan | p.4. O3 COO | approver | revised plan | p.4. O4 Finance | forecast author | ARR | p.4.
Relationships: M2→O1 and D1; plan→O2/O3; M3→O4/D2. Review the active-customer conflict and confirm the year for June and 18 July.
Why this works
1
Typed tables prevent a target, forecast, and actual from collapsing into one unlabeled number.
2
A relationship table retains which owner and date belong to which metric or action.
Check the result
Does every value retain its unit, period, status, scope, and locator?
Are owners' roles explicit rather than inferred from nearby names?
Are ambiguous and conflicting entries preserved for review?
ChatGPT