Start with problems, not tools
The most common mistake businesses make with AI is starting from the technology (“we should use AI for something”) rather than from the work. The best opportunities are usually ordinary tasks that are frequent, time-consuming, text-heavy and low-risk.
The task audit
Spend one week noting tasks that fit any of these patterns:
- Writing from scratch: emails, posts, product descriptions, proposals
- Reading a lot: long documents, reviews, reports, transcripts
- Repeating yourself: answering the same customer questions, onboarding new staff
- Reformatting: turning notes into documents, data into summaries
- Research: comparing suppliers, understanding a regulation, scanning competitors
Then score each task:
| Task | Hours per month | Risk if wrong (1–5) | AI fit (1–5) |
|---|---|---|---|
| Writing product descriptions | 10 | 2 | 5 |
| Replying to common customer emails | 15 | 3 | 4 |
| Monthly sales summary for the team | 4 | 3 | 4 |
| Negotiating supplier contracts | 6 | 5 | 2 |
Prioritize tasks with high hours, high AI fit and low-to-moderate risk. Leave high-risk, judgment-heavy work for later, with careful human review.
Let AI help with the audit
I run a [TYPE OF BUSINESS] with [NUMBER] staff. Here are the tasks
that take most of our time each week:
[LIST TASKS WITH APPROXIMATE HOURS]
For each task, rate how well current AI assistants could help
(1–5), explain how, and note the main risk. Then recommend the
three tasks we should start with and a simple first step for each.
A possible excerpt: “Product descriptions (fit: 5). AI can draft descriptions from your spec sheets in your brand voice. Main risk: incorrect specifications, so always check measurements and materials. First step: create a description template with three of your best existing examples.”
Quick wins vs. bigger projects
Quick wins need only an AI assistant and a good prompt: drafting, summarizing, brainstorming, rewriting. Start here. You will see results in days.
Bigger projects connect AI to your systems: automatic email triage, support chatbots or reports generated from your sales data. These need more setup and testing. Our Automating Work with AI track covers them.
Measure something
Pick one measure per use case so you know whether AI is helping:
- Time saved per task (estimate before and after)
- Turnaround time for customer replies
- Output volume, such as posts per week
- Quality signals: fewer revisions, customer satisfaction, error rates
Without a baseline, it is easy to feel busy with AI without actually being better off.
Keep humans accountable
AI drafts; people decide. For every use case, name who reviews the output before it reaches customers, and what they check. This single habit prevents most AI embarrassments in business.
A 30-day starter plan
- Week 1: task audit and scoring.
- Week 2: pick two quick wins and write prompt templates for them.
- Week 3: use them daily and track time saved.
- Week 4: review results, refine the prompts, and choose the next use case.
Practice task
Do a task audit for your business or team: list at least eight recurring tasks with estimated hours per month, then score each for risk and AI fit. Use the audit prompt to get an assistant’s view, compare it with your own scores, and choose two quick wins. For each, note who will review the output and one thing you will measure.