The Practitioner’s Road
You already chat with AI now and then. This track turns that habit into a dependable working skill — still without a line of code. Here is what changes, how to think about it, and where each page takes you.
Two colleagues are handed the same job: turn forty pages of meeting notes into a one-page briefing by Friday. The first types “summarise this” into an AI assistant, gets something bland and slightly wrong, and decides AI is overrated. The second spends three minutes explaining who the briefing is for, what decisions it must support and what format the boss likes, asks the assistant to quote its evidence, checks the three numbers that matter, and finishes before lunch. Same tool. Very different results. The gap between them is not talent or technical knowledge — it is a small set of learnable habits. That is what this track teaches.
From Novice to Practitioner
The Novice track was about understanding: what a large language model is, why it sometimes invents things, and how to ask a clear question. The Practitioner track is about using that understanding to get real work done — reports, emails, study notes, spreadsheets, images, small automations — reliably enough that you would put your name to the result.
Three things change as you make the move:
- From questions to briefs. A Novice asks; a Practitioner briefs. A good prompt reads like instructions to a capable new colleague: the goal, the audience, the material, the constraints and what “done” looks like.
- From one tool to the right tool. General chat assistants are only one category. Search engines with AI answers, meeting transcribers, office-suite assistants and image tools each fit different jobs — and different workplace rules.
- From trust-or-reject to a checking routine. Instead of either believing everything or dismissing AI entirely, you learn which outputs need checking, how, and how much.
Habits of effective users
People who get consistent value from AI tend to share a handful of habits. None needs special software. Tap each to see what it looks like in practice.
Delegate, collaborate, verify
This guide’s framing A useful way to decide how to use AI on any task is to ask which of three modes it belongs in. Most real work mixes all three.
DELDelegateLow stakes, easy to check
Hand the whole task over: reformat a list into a table, draft a polite reminder email, suggest ten titles, tidy a messy paragraph. You glance at the result and use it. Good candidates are tasks where a mistake is cheap and obvious.
COLCollaborateYour judgement, its speed
Work back and forth: you supply the ideas, knowledge and decisions; the AI drafts, questions, restructures and offers alternatives. A strategy paper, a lesson plan or a difficult letter belongs here. The thinking stays yours; the AI is a fast, tireless sparring partner.
VERVerifyHigh stakes, hard to check
Anything involving facts that matter — figures in a report, legal or medical information, citations, claims about real people — needs a deliberate check against a reliable source before it leaves your hands. Here AI can still help (finding sources, spotting gaps), but it is never the final authority.
The rule of thumb: the higher the stakes and the harder an error is to spot, the further you move from delegate towards verify. A typo in a party invitation is fine to delegate; a wrong figure in a board paper is not.
The twelve pages of this track
Each page stands on its own, but they build naturally in this order.
- The Practitioner’s RoadThis page: the mindset and the map.
- Prompt CraftRoles, structure, examples, constraints, output formats and chaining.
- Working with ContextFiles, long documents, custom instructions, projects and memory.
- Choosing the Right ToolThe categories of AI tool and how to pick one safely.
- AI for WritingDrafting, editing and translating while keeping your own voice.
- AI for Research and LearningGrounded answers, citation checks and studying with AI.
- AI for Data and SpreadsheetsAsking questions of your numbers and checking the answers.
- Images, Audio and VideoGenerating and editing media responsibly.
- No-Code Automation and AgentsLetting AI take steps on your behalf — carefully.
- Checking AI’s WorkA systematic verification routine.
- AI in the WorkplacePolicies, confidentiality and bringing a team along.
- Practitioner CheckpointProve it with a mini-project.
How long it takes
Each page takes about 10–15 minutes to read. The real learning happens in the “Try this” exercise at the end of each one, which needs another 15–30 minutes with an AI assistant open beside you. A comfortable pace is two pages a week: the whole track in about six weeks, alongside a normal job. Free versions of the major assistants are enough for every exercise As of 2026.
Find out your organisation’s rules on AI tools. Many workplaces now have an approved assistant and a list of what may and may not be shared with it. Until you know, practise with public, made-up or personal non-sensitive material only.
Where are you starting from?
A quick self-check. There is no pass mark — it simply shows which pages to linger on.
An AI summary of a long report reads well. What should you do before sending it to your manager?
Models can misread or invent details even with the source in front of them, and asking the same model to grade itself is weak evidence. See Checking AI’s Work.
Which prompt is most likely to give a usable first draft?
Specific audience, purpose, length and tone beat adjectives. See Prompt Craft.
Your company has approved one AI assistant. A friend recommends a different free tool that writes better. For a client proposal you should…
Client material is confidential; deleting a chat does not undo how it was handled. See Choosing the Right Tool and AI in the Workplace.
A task is high-stakes and errors would be hard to spot. Which mode fits best?
The harder an error is to see and the more it costs, the more deliberate the check.
// Pause and reflect
- List three tasks from your last working week. For each, would you delegate, collaborate or verify — and why?
- Which of the seven habits do you already have? Which one would save you the most time if you started tomorrow?
- Do you know your workplace’s rules on AI tools? If not, who could you ask?
Your notes are saved privately in this browser.
Start a “prompt notebook” — a plain document or notes app page. Each time a prompt gives you a result you would happily reuse, paste it in with one line saying what it was for. By the end of the track you will have your own personal toolkit.
Key terms on this page
Sources & further reading
- Anthropic, Prompting best practices (living documentation, accessed 2026).
- OpenAI, Prompt engineering guide (living documentation, accessed 2026).
- Google, Prompt design strategies, Gemini API documentation (accessed 2026).
- Ethan Mollick, Co-Intelligence: Living and Working with AI (2024) — a readable book on working alongside AI day to day.