AI for small business: where to actually start

Most small business owners we talk to feel two things at once: a quiet pressure that AI is going to matter, and a louder confusion about what to actually do about it. There is a new tool every week. Every consultant has a deck. Every LinkedIn post promises 10x.

This post is the conversation we have at the start of a Genesee AI engagement, stripped down. It is not a list of tools. It is a way to decide where to start.

Start with where your time goes, not what's trendy

Open your calendar from last week. Look at the meetings, the recurring tasks, the things you delegated and then redid. AI is genuinely good at a narrow set of jobs: reading and writing language, listening and speaking, looking at images, pulling structured data out of unstructured stuff, and following written instructions reliably.

If a task on your list involves one of those, AI is in scope. If it does not, AI is probably not your problem.

A good first filter: is this task repeated, predictable, and does it eat a real number of hours per week? Three yeses means it is worth looking at. One yes means come back to it later.

The four buckets that pay back fastest

Across the businesses we work with, these are the projects that consistently earn their keep within the first 90 days:

• Customer support that runs at night. An AI agent on your website, in your inbox, or on your phone line that handles the routine 60–80% of questions and hands off the rest. Customers get faster answers. Your team gets fewer interruptions.
• Sales call review. Every call recorded and analyzed against a simple rubric. Reps see what worked, what was missed, what to follow up on. The fix-it loop closes in days instead of quarters.
• Internal knowledge that answers itself. A chatbot trained on your SOPs, contracts, and product docs. Onboarding new hires goes from weeks to days. Existing staff stop pinging each other for files.
• Quoting and proposal drafts. From an intake form or a transcript to a draft estimate in minutes. Sales still review and send, but the blank page is gone.

These four are not glamorous. They are reliable. They show up in your week immediately.

What to skip in year one

Plenty of AI projects look exciting in a demo and then quietly fail in production. We tell clients to put these on the backlog, not the roadmap, until the basics are paid back:

• Fully autonomous "AI employees." The technology can mostly do it. The risk of letting it run unattended is usually too high for the upside.
• Anything that depends on data you don't have or can't access cleanly. If your CRM is a mess, the AI built on top of it will be a confident mess. Fix the data first or pick a project that doesn't need it.
• Custom-trained models from scratch. Almost no SMB needs this. The off-the-shelf models from Anthropic, OpenAI, and Google are good enough for nearly every business problem.

A reasonable 90-day plan

Here is what we typically propose to a small or mid-sized business in its first three months working with Genesee AI:

1. Weeks 1–2. A discovery call and a short audit of where time and money are going. Pick two projects. Not five. Not one.
2. Weeks 3–8. Build and deploy the first project. Integrate it with the tools the team already uses. Train one champion on the team. Measure baseline metrics.
3. Weeks 9–12. Build and deploy the second project. Review the first one's numbers. Decide what to expand and what to leave alone.

That cadence sounds slow until you compare it to teams that bought four AI tools at once, never integrated any of them, and quietly let the subscriptions renew.

The honest answer to "should we use AI?"

You should use AI when there is a specific, expensive, repeatable problem that AI is good at solving. You should not use it because everyone else seems to be. The biggest wins we see come from picking one well-defined problem, building the smallest thing that works, and letting it run.

If you want help finding your first project, book a free consultation and we will talk through your week. No pitch deck, no pressure.