AI Guides › Workbench
By Nigel Guy · 7 min read
Most people choose an AI course by its syllabus, its star rating or the badge at the end, then watch the videos at 1.5x and call it learning. It feels productive because the progress bar moves. Three weeks later you can describe what a context window is and still cannot get a usable first draft out of a tool on your own work.
The rule: judge a course by what you will have produced with your own material by the end of each lesson, not by what you will have watched.
This kit gives you four courses from official sources that are worth a look, a test to run on any course (including ones not listed here), and two prompts that turn passive lessons into practice.
Details were checked against each provider's public page on 2026-10-04. Pricing and course lists change, so treat the cost column as a starting point and check the page itself.
| Course | What it does | Cost | Best for | Catch |
|---|---|---|---|---|
| Claude Academy, "AI Fluency: Framework and Foundations" | Teaches a way of working with AI called the 4D framework; 14 lessons, one quiz, about 4 hours | Listed as free; no paid certificate mentioned | Anyone who uses AI at work and wants judgement, not tricks | Concept-heavy; the application is on you |
| Claude Academy, "AI Capabilities and Limitations" | Explains what language models can and cannot do (next-token prediction, context limits); 13 lessons, about 3.5 hours | Listed as free | Beginners who keep being surprised by wrong answers | Explains why, does not set you tasks |
| OpenAI Academy | Tracks for work ("Apply AI at Work"), builders, leaders and educators, plus livestreams and recordings | Cost not stated on the pages I could read; check before you commit | People who mainly use ChatGPT at work | Mix of events and courses; no fixed route, so you must pick one |
| DeepLearning.AI short courses | Short, hands-on lessons with video plus interactive notebooks; roughly one to two hours each for the short ones | Some free content, plus paid membership plans; I could not confirm current prices, so check the £ price at checkout | Anyone ready to write a little code or build a small agent | Assumes you are comfortable following code |
| Coursera, "AI for Everyone" (Andrew Ng) | Non-technical overview; about 7 hours over 4 weeks | Free trial for eligible learners; certificate costs extra. Check the £ price at checkout | Managers deciding where AI fits in a team | Strategy and vocabulary, not tool practice |
Claude Academy also lists a third, beta course, "Building Effective Human-Agent Teams" (about 45 minutes). It is new and labelled beta, so expect it to change.
Run this on any course page before you spend time or money. Four questions, each answered yes or no.
Score 3 or 4: take it. Score 2: take it only with the prompts below added. Score 0 or 1: it is a lecture, so treat it as reading.
Most courses explain well and apply poorly. Fix that yourself, lesson by lesson, with this routine:
The prompt below handles step 3 for you.
You are a practical coach helping me apply one lesson from an AI course to my real work.
Context
- Course and lesson: [COURSE_AND_LESSON_NAME]
- What the lesson taught, in my own words: [LESSON_SUMMARY]
- My role and a real task from this week: [ROLE_AND_TASK]
- Material I can use (paste it, with anything confidential removed): [MY_MATERIAL]
Goal
Design one 10-minute exercise that makes me use the lesson's idea on my task. A good result is something I can finish in one sitting and show to a colleague.
Steps
1. Restate the lesson's key idea in one sentence so I can check you understood it.
2. Propose the exercise as numbered steps I carry out myself.
3. Say what a good result looks like and what a poor one looks like.
4. Name the most likely way I will go wrong.
Output format
Use headings: Idea, Exercise, Good result, Likely mistake. Keep it under 250 words.
Rules
- If any bracketed input is missing or unclear, ask me for it before answering. Do not guess.
- Do not invent facts about the course. If you do not know what the lesson covered, say so and rely only on my summary.
- Do not do the exercise for me.
Before you answer, check that every step is something I can do with the material I gave you.
Fill in the five bracketed items. Remove names, client details and anything confidential before you paste.
A certificate shows you finished. A proof page shows what you can do. Build one from your lesson notes.
You are an editor helping me turn course notes into a short proof-of-skill page for [AUDIENCE, e.g. my manager or a hiring panel].
Inputs
- My notes from each lesson, including what I made and what failed: [NOTES]
- The role or outcome I want this to support: [TARGET_ROLE_OR_OUTCOME]
Goal
Produce a one-page summary that shows applied work: three to five things I made, what each was for, what went wrong, and what I changed.
Output format
For each item: Task, What I did, Result, What I would do differently. Then a three-line summary at the top.
Rules
- Use only what is in my notes. Do not add results, numbers or tools I did not mention.
- If an item has no clear result, mark it "result not recorded" and ask me what happened.
- Plain British English, no hype.
Before you answer, check that nothing on the page is missing from my notes.