Udemy
    •  
    •  
    •  
    •  
    •  
    •  
    •  
    •  
Turn what you know into an opportunity and reach millions around the world.
Learn More
Your cart is empty.
Keep shopping
Critical Thinking with AI: Fact-Check Before You Send
New

Critical Thinking with AI: Fact-Check Before You Send

Catch AI hallucinations: check drafts against your sources for changed numbers, missing caveats and outdated facts.
Created byBeata Janeczko
Last updated 10/2026
English
English

What you'll learn

  • Check an AI-drafted email, report or brief against the documents it was built from, line by line.
  • Decide which lines to check first by asking who will act on each one and what happens if it is wrong.
  • Find out whether a cited source exists, whether it says what the draft claims, and whether it is current.
  • Test the numbers in a sentence: per cent and percentage points, averages and medians, units and periods.
  • Use a second assistant's objections as leads, and settle them against the source passages.
  • Repair a draft without losing its message: correct, qualify, remove, or hold a line for its owner.
  • Record each check in a claim ledger and write a handover note when the evidence is missing.

Course content

3 sections • 10 lectures • 43m total length

  • What This Course Is About1:28

    Julia's assistant drafts a client email that guarantees a reply within one working day, while the current terms only aim for two. You learn who the course is for, the four steps Catch, Trace, Test and Fix, and how the case and the practice work.

  • Before This Email Goes Out4:11

    Julia Hart at Northfield Partners has an AI-drafted email for her client, Tom Becker, that reads as if it is finished. You learn to check a polished draft against the four documents it came from, and to spot a promise taken from replaced terms.

  • Where AI Errors Hide4:25

    Julia's email hides more errors: a number in the wrong unit, a pilot average stretched into a promise about every query, and a refund rule built on a remark from a call. You learn to compare meaning with the sources and to leave fair paraphrases as they are.

  • Which Lines Could Hurt You?4:54

    Julia's draft has nine lines, a long report might have ninety, and checking takes time. You learn to rank lines by who will act on each one and what happens if it is wrong, start a claim ledger, and practise with Practice 1 in the resources.

  • Catch AI Hallucinations in a Draft

Requirements

  • You should be able to read short workplace documents in English. No previous training is needed.
  • You don't need an AI account or any special software: every exercise works on paper or in a document.

Description

This course contains the use of artificial intelligence.

In 2026, AI assistants draft emails, reports and briefs from the documents you give them, but polished wording can hide changed numbers, missing conditions and outdated versions. A clear sentence can turn a reply target into a guarantee or make a limited pilot result sound like a promise to every customer. This course gives you a practical way to check the draft against your sources and decide what can go out. It is fact-checking for AI writing at work: you verify each claim, number and caveat against its source, so accuracy does not depend on how polished the draft sounds.

Check the Draft Before Someone Acts on It

  • Catch the lines people will act on.

  • Trace each claim to its source.

  • Test the meaning, numbers, scope and currency.

  • Fix the line while keeping the useful message.

The method is Catch. Trace. Test. Fix. It answers two questions: does the draft represent the source accurately, and is that source suitable and current? A draft can repeat an old policy word for word and still give the wrong answer today.

You follow one fictional workplace case in which a small consultancy prepares a client email, a board brief and a management update. You see the source passages, the draft lines and the reasoning behind each decision. In three short sections you learn to spot the lines that matter, check sources, dates and numbers, and repair a draft without losing its message, with a simple claim ledger to record each check. You also learn when to hold an unresolved claim for its owner and why an AI-assisted rewrite needs checking again.

Every section includes practice packs with answer keys, and three quizzes and a final assessment let you apply the method to new situations. All required practice uses supplied drafts and source extracts, without an AI tool, account or upload. The course contains ten videos and about 44 minutes of video.

This course is for people who prepare or review AI-assisted emails, reports and briefs that others rely on, in customer service, finance, learning and development or any other team.

Before you send, keep the message and make each consequential claim match the evidence.

Who this course is for:

  • You prepare or review AI-assisted emails, reports or briefs that other people use to plan, spend or decide.
  • You work in customer service and need written replies to match current terms, conditions and approval records.
  • You work with finance summaries and need figures, units and reporting periods to match the source material.
  • You work in learning and development and need briefings to keep a study's findings, participants and limits.
  • You lead a team and want one shared routine for checking AI-assisted work before it goes out.

Report abuse