How to Identify the Best AI Automation Opportunities in Your Business

The best AI automation opportunity in a small business is usually a boring one: a task that happens every day, follows a pattern, involves reading or writing text, and annoys the people who do it. Find those, score them on value and risk, and start with the one that is valuable and safe. That's the method. The rest of this piece is how to do it properly.

Start with where the hours go, not with the tool

AI vendors will show you what their product can do. That's the wrong starting point. Start with a list of what your team actually spends its time on.

The quickest way: ask everyone to write down, for one week, each task that took more than 15 minutes and came back more than once. No timesheets, no software. A shared document is fine.

You'll end up with somewhere between 20 and 40 tasks. That is your long list.

Four signs a task is a good candidate

• It's frequent. Daily or weekly, not once a quarter. Frequency is where the savings multiply.
• It follows a pattern. Someone could write a one-page instruction for it, with a few examples.
• It's text-heavy. Reading emails, documents, notes or forms, or writing replies, summaries and drafts. This is where current AI models are strong.
• The input is digital. If the information sits in an inbox, a CRM or a shared drive, it can be reached. If it's on paper or in someone's head, fix that first.

A fifth sign is softer but useful: people complain about it. Tasks that everyone dislikes are tasks where a change gets welcomed instead of resisted.

Three signs to leave it alone for now

• A wrong answer would be expensive and hard to spot. Pricing decisions, legal commitments, anything that moves money without a second look.
• Nobody can explain how it's done today. Automating an undefined process gives you an undefined process that runs faster.
• It happens twice a year. The setup will cost more than it ever saves.

Score every candidate on four questions

Take the long list and give each task a score from 1 to 5 on four questions. Do it with the people who do the work. They know where the exceptions are.

• Value: how many hours a month, or how much revenue at risk? Five means a lot.
• Effort: how many systems are involved and how messy is the data? Five means easy.
• Risk: what happens if the output is wrong, and would anyone notice? Five means low risk.
• Readiness: is the process clear and the data reachable today? Five means ready.

Add the four numbers. Anything above 15 deserves a closer look. Anything with a risk score of 1 or 2 goes on a separate list, no matter how valuable it looks.

Put a number on the value

For your top three, do the rough arithmetic. It doesn't need to be precise. It needs to be honest.

A worked example with made-up numbers: two people each spend about six hours a week copying details from inquiry emails into the CRM and writing a first reply. That's 12 hours a week. At a loaded cost of $40 an hour over 46 working weeks, it's about $22,000 a year of capacity. If faster replies also win one extra client a quarter, add that.

Then ask the question most people skip: what will those 12 hours be used for? If the answer is nothing in particular, the saving is theoretical. If the answer is more sales calls, or not hiring for that role next year, it's real.

Typical candidates in service businesses

Every business is different, but in professional services, staffing, consulting and other sales-led companies, the same handful of processes tend to score well:

• Inbound inquiries: reading, enriching, routing and drafting the first reply.
• Meeting follow-up: summaries, action items and CRM updates from call notes.
• Proposals and quotes: first drafts from discovery notes and your templates.
• Document intake: CVs, purchase orders, invoices and forms into structured data.
• Weekly reporting: numbers pulled from three systems into one short summary.

Notice that none of these replaces a person. Each one removes the typing and the waiting around the person.

Pick one, define done, measure before you start

Start with one process. Not three. Write down what success looks like as a number: reply time from a day to under two hours, or admin time cut in half. Measure where you are now, before anything changes. Without a baseline you'll be arguing about feelings in two months.

Then decide where a person stays in the loop. For a first project the safe default is simple: AI drafts, and a person approves anything that leaves the building.

Automate the boring part. Keep the judgment human. Measure both.

The one-afternoon version

If you want to try this yourself before talking to anyone:

• Hour one: list every recurring task over 15 minutes. Ask two colleagues to add theirs.
• Hour two: score each one on value, effort, risk and readiness.
• Hour three: do the arithmetic on the top three and pick one.
• Last half hour: write down the baseline and who will review the output.

If you get stuck on the scoring or the arithmetic, that is exactly what a Business Strategy Session is for. Bring the long list and leave with a decision and a written plan.