Data on deck. Orders, Captain?

Where we stand on AI

The posture the whole guide takes: use AI, but with a method — because using it without one is worse than not using it at all. Sets the three rules — think before you prompt, record what you did, you decide — and the Plan–Do–Review cycle the rest runs on.

The digital team model

The organising idea, and not a metaphor. Treat your AI tools as a team of specialists you lead: you set the goal, assign the tasks, review the output, and stay accountable for all of it. Covers what makes the model work — goal before team, roles by stage, knowing when the human is a requirement.

Individual project or team project — what changes?

Read this early if you're in a group project. The team model assumes one lead; a group that hasn't agreed who leads has a coordination problem before it has an AI problem. Sets out what to settle between people first, so the virtual team can complement the human one.

Four zones of AI use

A plain map of where your use sits, from not using AI at all to designing a deliberate workflow. Names the line clearly: prompting and handing in the output is not acceptable — it isn't your work. Zone 4, intentional use, is the aim.

Before you build your virtual team — map the landscape

The planning most people skip — and the reason their output doesn't fit the question. You can't pick the team before you know the stage you're at. Works through five questions about your project's development pathway, the step that makes everything after it coherent.

Working with your virtual team

The day-to-day method: assign a role, give context, set a task, collect the output, review it hard, decide what to do with it — then repeat. Includes what these tools will and won't do — they follow instructions, they don't know what's right, they'll confirm a leading question — and the six-step task cycle.

Record keeping

No special system — just enough that someone else, or you three weeks later, can see exactly what you did and why. Lists the fields to capture for each substantive AI interaction: task, tool, prompt, what you used, what you discarded, how you verified, what you decided.

Tools — a starting point

What the current tools are good for, and where they fall short — with the honest note that the gaps between them have narrowed and keep shifting. The skill isn't knowing every tool; it's knowing what a given one is for, and being able to justify the choice.

What to include in your deliverables

How to make your method visible in the work itself — standard practice anywhere serious, not a rule invented for the project. A Methods section in two parts (the plan you made before any tool; the execution — what you asked, how you verified, what you decided), and how to carry the same into a presentation.

Three principles to work by

Three postures to keep in mind. DNA — don't trust anyone, never believe anything, always check everything — for high-stakes output. Trust, but validate, for everyday work. And the one beneath both: you are the captain. Your name is on it, not Data's.

The difference between nice reading and a report worth paying for

A missing methods section — and the teaching gap it revealed.

I researched after I wrote — the classic opposite of a research method

I checked whether any of it was original. Less than I'd hoped.

Do I give away what I told the university to "sell"?

Information is largely free now. Teaching isn't. So here's mine.