How to start using AI in a small business

The safest way to bring AI into a shop or small team is to treat it like any other change to how work gets done. Pick one job that eats time every week, measure it, set some rules, try a tool on low-risk work and keep it only if the numbers say it helps. You do not need technical staff for that. You need a week of notes, a clear trial and the discipline to stop when a tool costs more attention than it saves.

Most advice on how to use AI in small business starts with a list of products. This guide starts with your own week instead. For the wider picture of where AI fits across stock, staffing and daily routines, read the guide to AI for business operations.

Where small businesses stand

You are not late. The Census Bureau's Business Trends and Outlook Survey found that between 17% and 20% of US businesses used AI from December 2025 to May 2026, with retail at around 14%. Most businesses in that survey were not yet using it, so a careful start costs you little.

Step 1: Pick one recurring time sink and measure it

Choose one task that comes back every week and that you would happily hand to someone else. Good candidates are writing product descriptions, replying to routine customer emails, drafting job ads, turning notes into a staff notice or summarizing supplier price lists.

For one normal week, write down each time you do that task: the date, the minutes it took and anything that made it slow. Keep the notes on paper or in a phone memo. At the end of the week you have a baseline, such as "customer replies: 14 emails, 2 hours 20 minutes in total."

Without that baseline, any tool will feel faster in its first week because it is new. With it, you can compare like with like.

What tasks can AI do for a small business?

General assistants are good at first drafts and reshaping text. They are weaker at anything that depends on facts they cannot see, such as your current stock, your prices or what a customer was promised last month.

Start in the top rows. Move down the table only after the earlier trial has worked and your records can support it.

Step 2: Write the rules before anyone uses a tool

A one-page set of rules prevents most early mistakes. It should say which tools staff may use, what information must never be pasted in (customer details, payment information, passwords, staff records), who checks output before it reaches a customer, and who to ask when something looks wrong.

You can draft this in an evening with the free AI policy template. The guide to writing an AI policy for staff covers each clause and how to introduce it to a small team. For a fuller framework on managing risk, NIST publishes a free AI Risk Management Framework; most owners will only need its ideas about naming risks and assigning owners, not the whole document.

Step 3: Try a general assistant on low-risk drafting

For a first trial, use a general AI assistant for small business drafting rather than a specialist product. Many owners begin with a general chat assistant from a large provider; Microsoft Copilot, Google Gemini and ChatGPT are common examples, listed in no particular order. The guide to ChatGPT for small business shows prompts for everyday shop writing.

Give the assistant your own facts and ask it to shape them. "Write a 60-word description for a hand-thrown ceramic mug, 350 ml, dishwasher safe, made in Ohio" works better than "write a mug description," because the second invites the tool to invent details. Read every draft as if a new employee wrote it. Correct errors and note how long the review took, because review time is part of the real cost.

Step 4: Tidy the records AI would need

Drafting tools need only what you type. Anything more ambitious, such as stock suggestions or task summaries, needs clean records underneath. Most so-called AI tools for small business are only as useful as the product, stock and task records they read.

Before any trial that touches operations, check three things:

• Products. Each item has one name, one code and a current price. The guide to SKU numbers explains how to set up codes that stay readable as the catalog grows.
• Stock. Quantities are counted recently and recorded by location. If your stock lives in a spreadsheet, the inventory spreadsheet guide lists the columns worth keeping.
• Tasks. Recurring work has an owner and a due date, written down rather than remembered.

This tidying pays off whether or not you adopt AI. headoffis, for example, keeps products, stock by location and assigned work in one hosted workspace for US$29 per business per month including five active people; it does not include an AI feature, and clean records are useful on their own.

Step 5: Run a 30-day trial with a before and after count

Treat the trial as a small AI pilot project with a start date, an end date and one measure. Use the same task you measured in Step 1.

1. Write down the baseline from your week of notes.
2. Decide what counts as success, such as "customer replies take at least 30% less time, with no reply sent containing a mistake."
3. Use the tool for that task only, for 30 days, and log minutes as before, including review and correction time.
4. Note every error the tool made and whether a person caught it.
5. At the end, compare the weekly average with the baseline.

Keep the trial narrow. Testing three tools on five tasks at once produces impressions, not evidence.

Is AI worth it for a small business?

Answer with a simple sum. Add the monthly cost of the tool to the value of the time spent reviewing its output. Compare that with the value of the hours saved.

A worked example with illustrative figures: a tool costs $30 a month. Your trial shows customer replies dropping from 2 hours 20 minutes a week to 1 hour 10 minutes, including review. That saves about 70 minutes a week, or roughly 5 hours a month. If you value your time at $25 an hour, the saving is about $125 a month against $30 of cost, so the tool earns its place. If the saving had been 20 minutes a week, it would be close to break-even, and the extra tool to manage might not be worth it.

The free ROI calculator runs this sum with your own figures, including review time and setup hours.

Questions to ask before buying an AI feature

Software you already use may start offering an "AI" add-on. Before paying for it, ask the vendor:

• What exactly does the feature do, and on which of my records?
• Is my data used to train models, and can I switch that off?
• Who sees the information I enter, and where is it stored?
• Can I review or undo what it does before it changes a record or contacts a customer?
• What does it cost per person or per month, and can I trial it first?
• What happens to my data if I stop paying for the add-on?

Be most careful with AI agents for small business, meaning tools that take actions on their own, such as sending emails or placing orders. Start with tools that suggest and let a person approve. Give a tool permission to act only once its suggestions have proved reliable over a trial.

Should you hire an AI business advisor?

Some consultants sell setup help, and some products present themselves as a virtual adviser for owners. For a shop with a handful of staff, the steps above usually cover what you need. If you want an outside view, the SBA's resource partners, including Small Business Development Centers and SCORE, offer free or low-cost counseling. Ask any paid adviser to work from your measured task and your trial results rather than a general presentation.

A quick readiness self-check

Before your first trial, ask yourself four questions:

• Can I name one task and how long it takes each week?
• Do I have written rules about what staff may paste into a tool?
• Will someone review every output before a customer sees it?
• Are my product, stock and task records accurate enough to rely on?

If any answer is no, start there. The free readiness checklist goes through each area in more detail and shows where to begin.

When not to bother

AI is a poor fit when the task is rare, when an error would be expensive and hard to spot, or when the work depends on judgment and relationships your customers value. It is the wrong first step if your records are a mess, your team is already stretched by a change in progress, or you would spend more time checking output than doing the work yourself.

AI for small business owners works best as a small, measured habit rather than a big project. One task, clear rules, a 30-day count and an honest sum will tell you more than any product demonstration. If the numbers do not add up, stop the trial and keep the time you measured; you have lost little and learned where your hours go.

← All posts