How to start using AI in a small business without an IT team

You don't need an IT department to start. You need one person who owns it, one business AI plan on company accounts, a short list of real tasks from the people who do them, and a few plain rules. Spend the first month on three to five tasks, train people on their actual work, and measure the time before and after. Expand only what clearly saves time. Bring in your outside IT provider for a small number of specific jobs, like single sign-on and file permissions, but don't wait on them to begin.

Here is how that first month can look for a company with 15 to 100 employees and a few people in the office.

Week 1: Pick an owner and a plan

Choose an owner who isn't you

The owner of the rollout should be someone who is organized, respected by the office staff, and curious about the tools. In many small businesses that is an office manager or operations lead. The business owner should sponsor it and use it, but usually shouldn't be the one chasing details. Give the rollout owner an hour or two a week for it.

Buy one business plan

Choose one of the business plans from Microsoft, OpenAI, or Anthropic and put it on company accounts, with the company as the owner and your rollout owner as an admin. If you already pay for Microsoft 365, Microsoft Copilot is the natural first test. If you run on Google Workspace, look at ChatGPT Business or Claude Team. We compare them in detail in ChatGPT Business vs Microsoft Copilot vs Claude Team.

Start with seats only for the people who will test it. You can add more later.

Move people off personal accounts

Your staff may already be using AI on personal accounts. That's a good sign; it means they want it. But it also means company information is going into accounts you don't control and won't keep when someone leaves. Say plainly that you are giving them a company account, and ask them to use it for work from now on. Treat it as a fresh start. Nobody is in trouble.

Week 2: Pick real tasks

The first tasks decide whether the rollout sticks, so they should come from the people who do the work, not from a vendor demo or a list of AI ideas. That means talking with each person in the office about where their week actually goes, then turning what you hear into specific, repeating tasks. "Reply to certificate requests" is a task. "Use AI for marketing" is too vague to test.

Start with only a few. Before anyone uses the tool on them, get a baseline for each one. Without it, you can't tell later whether anything changed, and the decision in week 4 turns into a matter of opinion.

Week 3: Set rules and train on real work

Put the rules on one page

Before people use the tool on real work, they need a one-page rules sheet that covers which account to use, what a person must check before anything goes out, what must never be pasted in, and who to ask when unsure. Keep it short enough that people will actually read it.

If you are in a regulated field, like healthcare, insurance, or accounting, check your industry's rules and your vendor's terms before using AI with client data. Some plans and features are designed for regulated data and some aren't.

Train on the actual tasks

Skip the generic tour of features. Train each person on their own task, using their own documents, with someone on hand to help. People learn faster on work they recognize, and they keep using a tool they have already seen save them time on a real job.

Week 4: Measure and decide

After a few weeks of real use, go back to the baseline and decide task by task. A lot of rollouts skip this step, and it's the one that tells you whether to spend more. It takes some care to do fairly. The time spent checking AI's output counts, and so do errors that only show up later. A tool people liked in week 1 and quietly stopped opening by week 4 hasn't passed.

Keep the tasks that pass. Drop the ones that don't, and don't feel bad about it; some work just isn't a good fit. Then pick the next few tasks and repeat.

Where your IT provider fits

Most small businesses without IT staff have an outside provider, often called an MSP, who handles devices, email, and security. You don't need them to start, but loop them in for:

• Single sign-on, so people log in to the AI tool with their work account.
• Reviewing file and folder permissions, especially before turning on Microsoft Copilot, which shows people anything they already have access to.
• Adding the AI tool to their list of approved apps and to your offboarding checklist.

What they usually won't do is figure out which of your tasks AI should help with, train your staff on them, and measure the results. That part is yours, or a consultant's.

What to skip at the start

• Customer-facing chatbots and custom-built apps. They take engineering work to run safely and are rarely the best first win.
• Automations that write straight into your accounting or management system without a review step.
• Buying several AI tools at once.
• A long AI policy document before anyone has used the tool. Start with the one-page rules and add to them as real questions come up.

Questions owners ask

How much does it cost to get started?

The software is per seat; standard business seats from the three main vendors list at roughly $20 to $25 per user per month. Testing with a few people keeps the first month small. The bigger cost is staff time, which is why picking real tasks matters.

What if my staff are worried AI will replace them?

Say what you are actually trying to do, which for most small businesses is to take the tedious parts off people's plates. Then involve them in picking the tasks. People are more likely to use a tool on work they chose.

How long until we know if it's working?

For a few well-chosen tasks, a month of real use is usually enough to see whether time is being saved. A broader rollout across the company takes longer.

Do we need a formal AI policy?

Eventually it helps, and some clients or insurers may ask for one. To start, the one-page rules above cover the most important risks.

Want someone to run the first month with you?

Armature helps small and midsize businesses in Washington pick the plan, set it up on company accounts, train their team on real tasks, and measure what changed. Email [email protected] for a free 30-minute call.