Article

Using AI in a small business without an IT department

Dr. Dmitry Nikelshpur · October 13, 2026 · 6 min read

Most small businesses don’t have an IT department, let alone a data science team. The good news: you don’t need one to get real value from AI. What you need is a sensible first task, a few ground rules, and a way to tell whether it’s working.

1. Start with a task, not a technology

“We should be using AI” is not a project. “We spend six hours a week writing quotes from site-visit notes” is. The best first tasks share three traits:

• Repetitive. The same kind of work, many times a week.
• Mostly text. Emails, notes, documents, forms — today’s AI is strongest with language.
• Easy to check. A person can tell in a minute whether the result is right.

Good candidates: drafting replies to common customer questions, turning rough notes into proposals or quotes, summarizing calls and meetings, sorting and routing incoming email, and pulling figures out of invoices or PDFs. Poor candidates, at least to start: anything where a mistake is expensive and hard to spot, such as final legal language, pricing commitments, or medical and financial advice.

2. Use what you already pay for

Before buying anything new, look at the software you already use. Office suites, email, CRM, accounting and help-desk tools increasingly include AI features — drafting, summarizing, searching your own documents. They already have your data in the right place, and someone else handles security and updates.

It helps to think in three levels:

1. General AI assistants used by your team for writing, research and summarizing. Lowest cost, lowest risk, quickest win.
2. AI features inside your existing tools, switched on for a specific workflow.
3. Custom tools built around your data and process — a website assistant, an intake system, a document pipeline. Worth it when the first two levels hit a wall and the task is central to your business.

Most businesses should spend their first month at levels one and two.

3. Write one page of ground rules

You don’t need a policy binder. You need one page that everyone reads:

• What not to paste in. Passwords, customer financial or health information, anything under a confidentiality agreement — unless you’ve confirmed the tool and plan are approved for it.
• Use business accounts. Business plans typically offer better privacy terms and admin control than personal ones. Check whether your plan uses your data to train models, and choose accordingly.
• A human approves anything customer-facing. AI drafts; a person sends.
• Turn on two-step sign-in for every account that touches company data.

4. Measure for two weeks

Pick one number before you start: minutes per quote, emails answered per day, time to first response. Write down today’s baseline, use the new approach for two weeks, and compare. If the number doesn’t move, change the task or the tool — not the definition of success.

Also note what didn’t work. Where the AI got things wrong tells you exactly where the guardrails and human review belong.

5. Know when to bring someone in

Off-the-shelf tools take you a long way. It’s time for outside help when:

• the task needs AI connected to your own systems — your scheduling, inventory, CRM or document store;
• the AI will talk to customers directly, without a person in between;
• the work involves sensitive data and you need to know exactly where it goes; or
• you’ve proven the value at small scale and want to make it reliable for the whole team.

At that point you’re no longer just using AI; you’re building with it. That deserves the same care as any other piece of business software: clear scope, testing, and someone accountable for keeping it running.

The short version

One task. Tools you already have. One page of rules. Two weeks of measuring. Most small businesses that follow that path find at least one place where AI saves real hours every week — and a much clearer sense of where it’s worth investing next.

Read next: What it takes to put an AI assistant on your website