How to use ChatGPT Work to draw charts and graphs, using computer spreadsheets, to illustrate technical reports

Chapter 1

Turn human-supplied financial data and technical-report definitions into a reviewable chart packet without certifying data, selecting the decision framing, approving interpretation, or publishing a report.

Use ChatGPT Work to turn approved financial data and technical-report definitions into a chart-review packet: test units, periods, populations, calculations, scaling, labels, and caveats before a graph is used to support a conclusion. It can render alternative views and reveal inconsistencies; it cannot certify a dataset, choose the material comparison, validate a model, approve an interpretation, publish a report, or make an investment or business decision. The task-rating evidence for this role is unavailable, so this is a controlled preparation workflow—not a claim about current use or importance.

Inputs

• Approved report question, audience, decision context, reporting cut-off, metrics, units, population, comparison basis, and chart requirements
• Human-supplied source data, extracts, calculations, spreadsheet logic, prior charts, methodology, and caveats with source and date
• Authorized review, disclosure, accessibility, and publication requirements

Steps

1. Register each source table, extract date, population, metric, unit, denominator, transformation, calculation, and evidence state.
2. Check totals, joins, period alignment, missingness, duplicates, outliers, denominators, axis scale, ordering, labels, rounding, and whether the selected chart preserves the approved comparison question.
3. Produce chart candidates and a chart-spec ledger that separates rendered views from data certification, analytical interpretation, causal explanation, and decision use.
4. Stop before changing source data, certifying a calculation, selecting a materiality threshold, approving an interpretation, publishing a report, or using the graphic to make or communicate a decision.

Using only the approved report question, reporting cut-off, metric definitions, source data, calculations, and chart requirements I provide, prepare a chart-review packet. Register each source, extract date, population, unit, denominator, transformation, and evidence state; check totals, periods, missingness, duplicates, joins, axis scales, labels, ordering, rounding, and comparison basis; and label CONFIRMED, CONFLICT, MISSING, ASSUMPTION, or HUMAN TO VERIFY. Keep chart candidates separate from data certification and interpretation. Do not alter source data, certify calculations, select materiality, approve conclusions, publish, or make a decision.