How to use AI in your small businessStart with one job. Then the next.
No tech team needed. Here is the path we'd follow with any small business: find the jobs worth handing over, pick an assistant, connect it to your apps, set the rules and bring your team along. Plus where AI doesn't belong.
In short
To use AI in a small business, start with one job your team does every week that follows clear rules, like replying to order questions or prepping for sales calls. Give everyone the same AI assistant on a business plan, connect it to the apps that job lives in and write down your rules. Decide what always needs a person's OK, try it with two or three people for a week, then train the rest of the team on their own work. Check how it's used each month and add the next job.
You don't need to be early. As of October 2026, about a quarter of US businesses (24.6%) told the Census Bureau they had used AI in the past two weeks, and about half of small firms that use it are still experimenting, according to the Federal Reserve Banks. Our small business AI statistics page has the rest of the numbers. What matters more than speed is picking the right first job and getting your team to use it.
1. Find the jobs worth handing over
Don't start with the technology. Start with your week. The best first jobs for AI usually sound like one of these:
- "Let me pull that from QuickBooks and paste it into the sheet." Copying information between apps by hand.
- "Hang on, I'll check what we told the last customer." The same questions answered again and again.
- "The Monday report takes me most of the morning." Reports built by hand from the same sources.
- "Only Sam knows how we handle that." Know-how that lives in one person's head.
Then check each candidate against four questions:
- Does it happen every week, or every day?
- Does it follow rules you could write down?
- Does the information it needs already live in an app: your inbox, CRM, help desk or books?
- Can a person check the result in a minute or two?
Four yeses make a good first job. Common ones include replying to customer emails, prepping for sales calls, chasing unpaid invoices and the weekly marketing report. There are more in our list of everyday jobs AI can take on. Pick one. Starting with five at once is the most common way to end up with none.
If you'd like a second opinion, the free AI audit walks through a normal week with you and sorts each job into "AI handles it", "AI drafts, you approve" or "keep it human".
2. Choose one assistant for the team
For most small businesses the choice is between ChatGPT, Claude, Microsoft Copilot and Gemini. All four handle everyday writing, summarizing and questions well. The differences that matter for a small team are practical:
- Where your team already works. A business that lives in Outlook, Word and Teams will find Copilot close at hand. One that lives in Gmail and Google Docs may lean to Gemini.
- Which of your apps it can reach. Check that it can be connected to the apps your first job needs.
- A business plan, not personal accounts. The company should own the account, and everyone should sign in with their own work login.
- One assistant for everyone. Training, rules and support are far easier with one.
Our comparison of ChatGPT vs Claude vs Copilot goes deeper, and there are separate guides on Claude for Small Business and whether ChatGPT is safe for business.
3. Connect it to your everyday apps
On its own, an AI assistant only knows what someone pastes into it. That's why so many teams try it, get generic answers and drift back to the old way. The change comes when the assistant can look things up in the apps you already run on: the order in Shopify, the deal in HubSpot, the invoice in QuickBooks, the thread in Gmail.
That link between the assistant and an app is often called a connector or an integration. In plain terms, it's a secure sign-in that lets the assistant read from that app, and only change something there when a person says yes. Most assistants can now open some apps this way. Our plain-English guide to how AI assistants connect to your apps explains how it works, and the apps your AI assistant can work in page lists the common ones.
Three rules for this step:
- Connect only the apps your first job needs.
- Everyone signs in as themselves, so each person only sees what their role allows.
- Start with looking, not changing. Let it read before you let it update anything.
4. Write down your rules
A connected assistant still doesn't know how your business does things. Write that down, once, in plain English:
- Your tone: how you greet customers, sign off and the words you never use.
- Your policies: return windows, refund rules, payment terms, who handles what.
- Your facts: prices only from your price list, claims only from your approved list.
- What never goes in: passwords, card numbers, anything your clients asked you to keep private.
- When to stop and hand off, and to whom.
Some of this teaches the assistant. Some of it is a policy for your people. Our free AI usage policy template covers the second part, and you can copy it.
5. Decide what needs a person's OK
This is the step that keeps you in charge. Sort every task the assistant will touch into one of three groups before anyone uses it:
| Group | What it means | Examples |
|---|---|---|
| AI handles it | Looking things up and summing them up. Nothing changes anywhere. | Call prep, finding the right policy, the weekly cash check |
| AI drafts, you approve | It prepares the work. A person reviews it and presses the button. | Customer replies, payment reminders, updating a deal, a refund |
| Keep it human | The assistant can gather facts. A person decides and does it. | Pricing a big deal, hiring, calming an upset customer |
Write the list down and name who approves what. Our guide to human in the loop explains how to set this up so it doesn't slow the team down.
6. Try it with a few people first
Pick two or three people who do the job every day, ideally one keen and one skeptical. Have them use it on real work for about a week. Ask them to note:
- Answers that were wrong, out of date or off-tone.
- Questions it couldn't answer, and where that information actually lives.
- Anything that felt risky, or any rule that didn't hold.
- Roughly how long the job took before, and how long it takes now.
Fix what they find before anyone else sees it. A team forgives a tool that's still being tuned in a pilot. It doesn't forgive one that embarrasses them in front of a customer on day one.
7. Train the team on their own work
Generic AI courses teach people to write clever prompts. What they need is to see the assistant do their own job, in their own apps. An hour, live, with real examples, works better than a day of slides. Give each role a one-page cheat sheet with the five or six requests they'll use most, what needs approval and who to ask when stuck.
Name that person. "Ask Sam" beats a help page. Our guide on getting your team to actually use AI covers why AI tools sit unused and includes a 30-day rollout plan. If you'd rather have it done for you, see AI training for your team.
8. Look at how it's actually used
A month in, look at what happened rather than what was planned:
- Who is using it, and who has stopped.
- Which requests come up most, and which never do.
- Where answers were wrong, and why: missing information, an unclear rule or the wrong app.
- What people asked that it couldn't answer. That's often your next job.
This only works if your setup keeps a history of requests, with who asked what and when. If it doesn't, ask people directly. Five minutes in a team meeting tells you a lot.
9. Keep it working
AI at work isn't a one-time install. Your apps update, your prices and policies change, people join and leave. Put a short monthly check on the calendar:
- Update your rules when a policy, price or product changes.
- Remove access the day someone leaves.
- Check the connected apps still work after their updates.
- Add the next job once the first one is part of the routine.
That last point is where most of the value builds up: one job, then the next. Our page on building an AI plan shows how that looks rolled out one win at a time.
What not to use AI for
Being honest about this is part of doing it well. Keep AI out of, or a person firmly in charge of:
- Judgment calls. Pricing a big deal, hiring, firing, whether to take on debt. The assistant can prepare the facts. A person decides.
- Relationships. The calls and conversations your customers stay for should sound like you, because they are you.
- Work with no information behind it. If it isn't written down anywhere, the assistant will guess. Write it down first, or leave it alone.
- Anything you can't check. If nobody on the team could spot a wrong answer, don't hand it over.
- Final legal, tax or medical advice. Use it to prepare questions for your accountant or lawyer, then have the conversation with them.
- Anything that sends, pays, refunds or deletes without a person. Drafts are fine. The button stays with your team.
Do it yourself or get help
A small team can do steps one, two, four and five on its own in an afternoon. Steps three, six and seven are where most businesses stall, because connecting apps, testing on real work and training people take time nobody has spare. In the Federal Reserve Banks' survey, 37% of firms planning to use AI named the time to implement it or train employees as a challenge.
If you're weighing outside help, our guide on what an AI consultant does covers the kinds of help on offer and when you don't need any. This is also the work we do: we connect your AI assistant to your team's apps, set it up around your rules, train your team and keep it working. Most teams start with one of our four packages, and how it works walks through every step.
Common questions
How do I start using AI in my small business?
Pick one job your team does every week that follows clear rules and lives in an app you already use. Give the team one AI assistant on a business plan, connect it to that app, write down your rules and test it with two or three people before rolling it out.
What's the best AI for a small business?
The one your team will use. ChatGPT, Claude, Microsoft Copilot and Gemini all handle everyday work well. Choose based on the apps your team already works in and which of them the assistant can connect to, then put everyone on the same one.
How much does it cost to use AI in a small business?
The main costs are the assistant's business plan, which you pay to the vendor, and the time it takes to set it up and train people. Check each vendor's pricing page for current plans. If you bring in help, ask for a fixed price in writing before work starts.
How long does it take?
One person can try an assistant on a single job in a day. Getting a whole team using it on real work, connected to your apps and with rules in place, usually takes a few weeks.
Do I need technical staff?
No. You need one person who knows how work flows through the business, and logins to the apps you want to connect. The rest is setup, testing and training.
Is it safe to put customer information into AI?
It can be, with a business plan, individual logins, clear rules on what never goes in and a person approving anything that leaves the building. Our guide on whether ChatGPT is safe for business has a checklist.