tutorpackSee prices
All courses

AI at work · course 23 of 105

AI Prompts That Work the First Time

Get usable work out of any AI chat the first time: 12 rules from the official prompting guides, weak prompts rewritten until they pass, and planted mistakes caught.

4 modules

Try a question

Not on sale yet. When sales open: $4.99 for this course, paid once, or every course for $9.99 a month. Prices

Preview

What it teaches

How to get usable work out of ChatGPT, Claude, Copilot or Gemini the first time, and how to catch what they get wrong. 12 rules, each quoted from Anthropic’s official prompting guides (OpenAI’s guide gives the same advice on the basics): say the task plainly, give the context and the why, spell out the output, show examples, keep your material separate, long documents first, say what to do instead of what not to do, let it say “I don’t know”, check what comes back, and more. Then you practise on real office jobs: spot what a weak prompt is missing, rewrite it against a checklist, and find the planted mistakes in AI answers (a wrong number, a made-up quote, a dropped warning, a broken word limit).

How long

40 graded items plus your own tasks. About 35-45 rounds and a 10-item final test: 2-3 sessions (estimate, not timed with a real student). The tutor gives a one-line save code at the end of each session; paste it into a new chat to carry on.

The modules

4 modules, drilled one item at a time.

  1. M0

    The 12 rules of a work prompt

    Concept questions

  2. M1

    Spot the gap

    Goal + prompt: what’s missing?

  3. M2

    Rewrite it

    Setup: write the prompt you’d send

  4. M3

    Check what comes back

    Source + AI answer: find every mistake

Before it teaches, it asks

  1. 1
    What do you use AI for at work, or want to (emails, summaries, reports, data, something else)? Which app: ChatGPT, Claude, Copilot, Gemini? And what annoys you most about what it gives back?
    AI voice (ElevenLabs)
  2. 2
    GoalA polite email turning down a vendor’s demo request, under 100 words, friendly because you may buy from them next year.PromptWrite an email saying no to the vendor meeting.What is this prompt missing?
  3. 3
    SourceQ3 revenue was $4.2M, up 8% from Q2. Churn rose to 3.1%. We hired 6 engineers.PromptSummarize this update in 2 bullets.AI answer- Revenue reached $4.2M in Q3, up 18% from Q2.
    - Churn rose to 3.1%, and we hired 6 engineers, including a new CTO.    Find every mistake in the answer.

What’s exact, and what isn’t

Show

Straight from the course’s own notes.

Sourced
Every rule quotes Anthropic’s or OpenAI’s prompting guide, pages checked 2026-10-09 (links in M0).
Checked by script
Every M3 mistake was checked against the source text (the wrong figure isn’t in the source, the right one is; a made-up claim’s words appear nowhere in the source; bullet and word counts recounted).
Judgment
M2 rewrites are graded on a checklist (every Must, no Fails if), not on matching the model prompt, so any prompt that does the job passes.
Checked by hand
M1 keys and M2 checklists.
Not in it yet
Long multi-step projects, AI for spreadsheets and data, building custom assistants (GPTs, Projects), images.