Home/Knowledge Base/Knowledge Worker Playbooks/K-03

K-03K-Series · Foundation

The AI Decision Tree: Should I Use AI for This?

You’re looking at a task, wondering whether AI belongs anywhere near it. This tree answers that in five questions, asked in order. Stop at the first question that gives you a colour: green means go with verification, amber means AI assists while you lead, red means keep it fully human. It takes about thirty seconds once you know it — and it prevents the expensive mistakes.

Why this matters

Without a routine, people make the AI decision by mood: enthusiasts push everything through it, sceptics push nothing, and both get burnt — one by a leaked detail or invented fact, the other by watching colleagues hand back hours they’re still spending. A decision tree replaces mood with a repeatable check that takes less time than the doubt did.

The five questions

1. Does it involve confidential or personal information? Salary details, board papers, client data, health information, anything about an identifiable person. If yes: only continue if your organisation has approved a tool for exactly this data. No approved tool means red — keep it human.
2. Is the task mostly production? Drafting, summarising, analysing, comparing, brainstorming. If it’s actually a judgement about a person, a decision, or a difficult conversation, that’s red — AI can at most help you prepare.
3. Could you tell if the output was wrong? If you lack the knowledge or the source documents to verify it properly, that’s amber at best — don’t use AI alone here.
4. Does the outcome directly affect a person? Hiring, performance, health, legal standing, money. If yes: amber — you lead, AI assists under close supervision, and the final call is documented as yours.
5. Would an undetected error be costly or embarrassing? If yes: green with guardrails — verify line by line before it leaves you. If no: green — brief it, draft it, verify it, done.

Reading the colours

• Green — go. AI drafts, you verify. This is where the time savings live.
• Amber — assist only. You lead, AI supports, review is heavier and the accountability is explicitly yours.
• Red — human only. Judgement about people, confidential content without an approved home, decisions with legal effect.

Notice what the tree never asks: which tool is trendy, or what a vendor demo showed. The questions are about the task, the information and the consequences — the things that don’t change between products.

Putting it into practice

1. Run today’s task list through the tree — most people find more green than they expected and one red they’d been risking
2. Ask IT which tools are approved for which data before you need the answer
3. Put the five questions somewhere visible until they’re automatic
4. When a task lands amber, decide the supervision before you start, not after
5. Re-run the tree when a task’s stakes change — the same task can change colour