Where should a small business start with AI?

Most business owners we talk to say some version of the same thing: “We know we should be using AI, but we don’t know where to start. Or we tried, and nothing came of it.”

That’s normal. The tools change every month, everyone has an opinion, and most advice is written by people selling one particular tool. So here’s the approach we use with clients. It’s boring on purpose, and it works.

Don’t start with a tool. Start with a task.

The most common mistake is picking the tool first: “We should get a chatbot” or “Let’s try ChatGPT for everything.” Then the team spends a month figuring out what to do with it, and it quietly gets abandoned.

Flip it around. Start with the work, not the technology. Ask your team one question:

“What do you do every week that feels like a waste of your time?”

You’ll get answers fast. Retyping information from one system into another. Writing the same kind of email over and over. Chasing people for hours, approvals or missing details. Pulling numbers together for a weekly report.

That list is your AI and automation roadmap.

How to pick the first one

Not every task on that list is a good first project. The best first project has four things going for it:

1. It happens often. Daily or weekly beats once a quarter. Frequent tasks pay you back quickly.
2. It follows a pattern. If you could write down the steps for a new hire, a computer can probably follow them too.
3. Mistakes are cheap to catch. Start where a person can review the result before it goes anywhere important.
4. You can measure it. Hours saved, errors avoided, faster replies. If you can’t measure it, you can’t prove it worked.

A good first pick usually sounds unglamorous: “Every new lead from the website gets typed into the CRM by hand.” That’s perfect. It’s frequent, it follows a pattern, and you’ll know within a week if it’s working.

AI or automation? You probably need both.

These two words get used like they mean the same thing. They don’t.

• Automation moves information from one place to another without anyone touching it. A form submission becomes a CRM contact. A paid invoice updates a spreadsheet. No thinking required, just rules.
• AI handles the parts that need judgment: rewriting messy notes into clear English, summarizing a long email thread, sorting requests by urgency, drafting a reply.

Most of the useful stuff we build is a mix. In our own AI Task Tracker, automation collects every contractor’s hours from their time logs. AI then rewrites their task notes into clear, client-ready English. Then automation again builds the invoice file for QuickBooks. That combination saves us about 20 hours of admin a week.

If someone tells you that you need “an AI strategy,” what you probably need is two or three of these small systems, built well.

Build it into the tools you already use

The fastest way to kill an AI project is to make your team learn a new app. They won’t, or they’ll use it for two weeks and go back to the old way.

So build the AI into what they already open every day: Google Sheets, HubSpot, Gmail, QuickBooks, your website. When the improvement shows up inside a familiar tool, people use it without being asked.

That’s how our Cell Line Genetics project worked. Their team kept using HubSpot exactly as before. Behind the scenes, every new contact now automatically becomes a client account on their ordering portal, with the right access. Nobody had to change their habits for it to work.

Prove it, then do the next one

Once your first project is live, give it a few weeks and check the number you picked at the start. Did it save the hours you expected? Are there fewer mistakes?

If yes, move on to the next task on your list. If not, figure out why before you build anything else. Sometimes the answer is a small fix. Sometimes it’s that the task wasn’t a good fit, and that’s useful to know too.

This is the opposite of a big AI rollout, and that’s the point. Small wins build trust with your team, and each one makes the next project easier.

When AI isn’t the answer

Here’s something you won’t hear from most AI vendors: sometimes the best fix isn’t AI at all.

If customers keep emailing the same question, a clearer page on your website might solve it better than a chatbot. If orders keep arriving with missing details, a better form with required fields beats an AI that guesses. If a report takes forever, maybe the problem is that the data lives in five places, and you need to connect them before AI can do anything useful.

We’d rather tell you that than sell you something you don’t need.

A simple way to start this week

1. Ask your team the “waste of time” question and write down every answer.
2. Score each one: how often does it happen, does it follow a pattern, and can you measure it?
3. Pick the one with the best score and the lowest risk.

That’s your first project. If you’d like help figuring out whether it can be automated and what it would take, that’s exactly what we do.

Quick answers

Do I need a big budget to start with AI? No. A good first project is small and focused. Many are built on tools you already pay for, like Google Workspace or HubSpot, plus an AI service that costs little per use.

Will AI replace people on my team? In our experience, it replaces tasks, not people. The goal is to get your team out of retyping and chasing, so they spend their time on work that actually needs them.

How long does a first project take? A focused automation is usually live in a few weeks, not months. Bigger systems are built in stages, each one proving itself before the next.