Most owners know they should be doing something with AI. Few know where to start, and the noise doesn't help. Here's the approach we use with clients, regardless of industry or country.
Step 1: Map the repetitive work
Spend a week noting every task your team repeats. Replying to the same questions. Writing similar quotes. Summarising calls. Updating spreadsheets. Producing social posts. Put a rough number of hours next to each.
Step 2: Score each task
For each task, ask three questions:
- How many hours a month does it take?
- How bad is a mistake? Embarrassing, expensive or irreversible?
- Is the information it needs written down somewhere?
The best first projects are high in hours, low in risk and rich in existing information.
Step 3: Pick one, and define success
"Use AI for marketing" is not a project. "Cut the time to produce our weekly email and three social posts from six hours to one" is. A clear before-and-after keeps everyone honest.
Step 4: Keep a person in the loop
For a first project, AI should draft and a person should approve. You'll quickly learn where it's reliable and where it isn't, and your team stays in control.
Step 5: Run it side by side
For two to four weeks, run the new way alongside the old. Compare time spent and quality. Then decide: keep, improve or drop.
Common first projects that work
- Instant replies to new enquiries (see speed to lead).
- Quote and proposal drafting from a price list.
- Meeting and call summaries with action items.
- A searchable assistant for your product information and policies.
- Consistent content production in your brand voice.
What to avoid
Avoid starting with anything customer-facing that can make commitments on your behalf, anything involving payments, and any tool that sends your data somewhere you don't understand. Those can come later, with proper controls.
One small, measured success teaches your business more than any strategy deck. Then you do the next one.
Part of our guide: Practical AI for Business: Where It Pays and How to Start