Home/Knowledge Base/Knowledge Worker Playbooks/K-05

K-05K-Series · Foundation

How to Spot AI Opportunities at Work

The best AI opportunity in your job is almost never the impressive one from the demo. It’s the unglamorous task you repeat every week, that starts from a blank page, that drains an hour you’d rather spend elsewhere. This card gives you a repeatable way to find those tasks — and a filter that stops you wasting effort forcing AI into places it doesn’t fit.

Why this matters

Most people’s AI adoption stalls the same way: they try it on whatever’s in front of them, get a mixed result, and quietly drift back to old habits. Opportunity-spotting fixes the order of operations. Instead of asking ‘what can this tool do?’, you ask ‘where does my week leak time on production work?’ — and start where the fit is natural and the win is measurable.

The signals of a good AI task

• It repeats — weekly reports, recurring emails, standard documents. Repetition is where habit-building pays back forever.
• It starts from a blank page — first drafts are AI’s strongest contribution; a starting point beats an empty screen.
• It’s production, not judgement — words, summaries, comparisons and structures, not decisions about people.
• You can verify it — you know the subject or hold the source documents, so wrong is detectable.
• The information is allowed — nothing confidential, or an approved tool exists for it.

Five signals, and the strongest opportunities light up all of them. Three or four is still worth a trial. One or two means leave it — for now.

The five-question sweep

1. What did I do last week that was mostly producing words or structure?
2. Which of those tasks will I do again this month?
3. For each: could AI genuinely assist, or would explaining it take longer than doing it?
4. What are the risks — confidentiality, errors, policy — and are they manageable?
5. How would I redesign the workflow so AI does the production and I keep the bookends?

Run the sweep once and you’ll typically find two or three genuine candidates. That’s the right number. Nobody transforms their whole week at once; the skill compounds one task at a time.

The anti-signals

Skip tasks where you couldn’t judge the output’s quality, where the explanation would outweigh the task, where the information has no approved home, or where the task is secretly a relationship (the ‘quick email’ that’s actually a delicate negotiation). Forcing AI into these produces the bad experiences that stall people for months.

Putting it into practice

1. Run the five-question sweep over last week’s calendar and sent items
2. Shortlist two or three tasks that hit at least four signals
3. Pick the single most repetitive one and trial it this week
4. Note the time saved and what you’d brief differently — keep the note
5. Add one new task a fortnight; retire experiments that don’t pay