Task guide · comfortable · 15 minutes

Understand a file or spreadsheet

Inspect unfamiliar documents or data, surface quality problems and verify the calculation before using the finding.

Situation: You have a report, PDF, CSV or spreadsheet and need to know what it contains before drawing a conclusion.
Finished outcome: A checked summary, calculation or chart grounded in the material you supplied.

The complete workflow

Understand the hand-off before opening a product.

The reading order remains useful without the visual layout: person, AI, review, result.

01 · person
You provide
A permitted file, field context and a clear question.
02 · ai
AI helps
A data-quality inventory before conclusions.
03 · review
You check
Ambiguities, formulas, outliers and a manual sample.
04 · result
Finished result
A verified finding with its method visible.

AI helps with

A useful first pass

Inventories fields and likely meanings
Flags missing or inconsistent data
Drafts calculations, comparisons and charts

You bring

Context and responsibility

Ambiguous columns and source quality
A manual sample of the calculation
Whether the interpretation makes practical sense

Worked example

See where the draft changes.

The correction is the core teaching moment, not an optional warning after the result.

01 · Input

What the person provides

A customer-feedback CSV with date, channel, category, free-text comment and resolution-time columns.

02 · AI draft

The useful first pass

Initial finding: delivery complaints rose 18%. Quality note: 12% of category cells are blank and resolution time mixes hours and days.

03 · Human correction

Facts and judgement

The analyst normalises units, manually checks twenty rows and separates a channel migration from a genuine complaint increase.

04 · Final result

Human-approved outcome

A verified chart showing delivery complaints were stable after correcting the channel mix, plus a documented data-quality issue.

Try it

Use your own material.

Remove anything you cannot share, replace the bracketed details and keep the checking step.

Do not upload data you are not authorised to share.
Do not use a calculation you cannot independently sample or explain.

Before analysing this file, tell me what each field appears to mean, identify missing or inconsistent data and ask about ambiguous columns. Then answer: [question]. Show the calculation method, call out assumptions, keep source rows traceable and give me one chart only if it makes the result easier to verify.

Supporting product guides