The short answer: don't start with "an AI strategy". Start with one boring task your team does by hand every week, and automate that. Boring tasks are easy to describe, easy to check, and you feel the time saved straight away.
AI agents are this month's big story. Microsoft is connecting agents to business data in Dynamics 365, and funding rounds for agent startups keep landing (AI Agents Directory's October brief has the roundup). For a small business, that news can feel like it's about someone else. It isn't. The same tools work at small scale. You just have to pick the right first problem.
What makes a good first problem?
Look for four things:
- It happens often. Daily or weekly, not once a year.
- Someone does it by hand, and it's the same steps every time.
- You can tell when it's done right. A certificate has the right name; a lead has an email; an invoice adds up.
- It annoys people. The person doing it will help you automate it.
A real example: certificates
An Islamic education team I worked with as COO issued a certificate to every student who finished a course. Each one was made by hand, and each took 15 minutes or more.
We built a system that takes the student's name and the course and produces the certificate in seconds. It has issued more than 3,400 certificates and saved roughly 850 hours. Not one part of it is flashy. That's the point. The full story is in case file 001.
Other boring problems worth automating
These are all problems I've seen businesses describe in almost the same words:
- Leads. "Where are our next customers coming from? Right now, nobody knows." → case file 002
- Attendance and payroll. The fingerprint machine knows who came late; payroll doesn't. → case file 003
- Customer feedback. Feedback arrives, but the angry customers never reach the staff who could fix it. → case file 007
- Captions. Captions take longer than the video itself. → case file 011
- Website visitors. People arrive with a question and leave before anyone knows they were there. → case file 013
Each one is a single, clear job. None of them needs a big AI platform to start.
Agent, automation or chatbot?
Don't worry about the label. An automation follows fixed steps. An agent can decide some of its steps and use tools. A chatbot talks to people. Most good first projects are mostly automation with a little AI where judgement is needed, such as reading a messy message or writing a reply. Pick whatever solves the problem with the fewest moving parts.
How to start this week
- Write down every task your team does by hand more than once a week.
- Next to each one, write who does it and how long it takes.
- Pick the one that's most frequent and easiest to check.
- Describe it in three sentences: what goes in, what comes out, how you'd know it worked.
That three-sentence description is most of the work. If you send it to me, I can usually tell you within a day whether it's worth automating. Open a case file. I reply within 24 hours.