16 primary guides
Use AI to inspect, transform, explain, and validate structured information.
Start with a job
Main subject
These guides have Data & Analysis as their primary topic.
In one original short-video test, Qwen3.8-Omni-Flash found the saved review window and final status, refused an invented metric, and made one unsupported future-state inference.
Reproduce a bank-to-ledger blind test where AI discovers two matches, classifies a fee and outstanding check, and an independent calculator proves both adjusted balances.
Reproduce an eight-row duplicate-invoice test that finds one exact pair and one supplier-alias candidate while protecting a credit memo and legitimate same-amount bills.
Use AI to structure a three-way purchase match, then verify quantities, unit prices, unsupported amounts, tolerances, duplicates, tax, and the payment hold yourself.
Match the manual’s maintenance list to your service history, calculate dates and odometer limits, handle missing records, and keep the reminders current.
Identify recurring-charge candidates in a redacted CSV, reconcile every transaction, handle annual plans and variable bills, and record verified subscription decisions.
Turn receipt photos into a checked expense workbook. Follow practical extraction prompts, verify amounts, handle tips, refunds and mixed purchases, and keep unresolved claims visible.
Compare three exact products with checked facts, must-have requirements, explicit scores and weights, and a spreadsheet you can reproduce. Test missing inputs and changing priorities before deciding.
Audit an existing A/B test recommendation with a complete claim review, reproducible CSV preparation, numerical checks and worked experiment scenarios.
Analyze customer feedback with AI by calibrating a small labeled set, preserving evidence, reviewing the confusion matrix, and separating model quality from production prevalence.
Start with a copy-down formula, then learn to check ChatGPT's Excel lookups, date sums, status rules, and growing tables against known answers.
Turn PDF orders into JSON with a schema and a worked example, then adapt the checks for scans, continued tables, mixed documents, and larger batches.
Run a four-row CSV example, choose the smallest useful output, and see how to map and check the fields in your own file.
Six checked ChatGPT chart examples cover monthly totals, missing time-series data, raw categories, a histogram, a scatter plot, and expense shares.
Follow a six-row CSV from ChatGPT profiling to a reviewable candidate, quarantine, change log, and row-level verification.
Analyze an Excel workbook with ChatGPT, reproduce the key rates in Excel, and handle duplicates, blanks, outliers, multiple sheets, and incomplete reads.
Also relevant
These overlap with Data & Analysis, but their main subject belongs elsewhere.
Use a reproducible 24-second example to score WER, DER, and speaker-attributed words, then apply the same checks to silence, overlap, and your own recordings.
Why it fits here: Speaker attribution is checked against a labeled reference.
Measure transcript accuracy with a no-code or Python WER check, a worked example, and practical checks for missing references, numbers, names, captions, speakers, and silence.
Why it fits here: The guide calculates and interprets transcript error measures.