By role

Researcher or analyst

You are the one who has to read everything before anybody else can decide.

This is you if your week ends in a document that has to be right, built from more sources than you had time to read.

What your week looks like

Somebody needs a briefing by Thursday. The evidence is 300 pages in three languages, half of it scanned, and you are the only person who will read it. Then the finding gets quoted in a report with your organisation’s name on it, and it has to hold. Most organisations here already use AI for research, so the question is not whether to use it. It is how to catch it when it is confidently wrong.

Start with these three tasks

  1. Research
  2. Making sense of large document sets
  3. Drafting reports and advocacy material

Tools you can actually afford

A free tier is usually enough for reading and drafting. It runs out on large document sets, so plan the work around that limit rather than around a subscription you cannot buy.

Never do this

  • Never repeat a fact, a number, or a citation you have not opened yourself. These tools invent sources that look exactly like real ones, complete with page numbers.
  • Never paste unpublished data, partner submissions, or draft findings into an online tool. Your draft is somebody else’s risk before it is your product.
  • Never let a summary replace reading the document when the finding matters. Use the summary to decide what to read first, not what to skip.

Your first hour

  1. 10 minutes. Choose one public document you already know well, a report you wrote or one you have read three times.
  2. 20 minutes. Ask an AI tool to summarise it and to list its five main findings.
  3. 20 minutes. Mark every single thing it got wrong: invented figures, findings that are not in the text, sources that do not exist. Keep that list.
  4. 10 minutes. Write one sentence at the top of your next research plan saying what you will check by hand every time. That sentence is your method.

Case studies

Recipes

Last checked: 2026-08-31