Where AI Fits in a Small Business

You open the laptop at 9pm to "catch up." There are four customer emails to answer, a menu board that needs new wording, a supplier who wants a polite no, and a pile of notes from a staff meeting that nobody has written up. None of it is hard. All of it is slow. And it is the reason you have not looked at your actual numbers in three weeks.

This is where AI earns its keep in a small business. Not running the company, not replacing staff, not making decisions. Taking the first pass at the writing and organizing work that sits between you and the work only you can do.

The trick is knowing which jobs to hand over and which ones to keep. Get that wrong and AI either wastes your time or, worse, puts something in front of a customer that should never have left the building.

What AI does well

AI tools like ChatGPT, Claude, and Gemini are language tools. They are strong at anything where the raw material is words and the job is to reshape them.

Drafting. A first version of an email, a product description, a job posting, a policy for your staff handbook. You still edit it, but editing a decent draft is faster and less painful than staring at a blank page.

Rewriting. Turning a blunt reply into a friendly one. Shortening a long announcement for social media. Making a technical explanation readable for a customer who has never heard the term.

Summarizing. Pasting in a long email thread, a supplier contract, or customer reviews and asking for the main points. It gives you a map, so you know where to read closely.

Brainstorming. Twenty names for a new product, ten ideas for a slow-Tuesday promotion, five ways to explain your pricing. Most will be ordinary. You only need one good one, and a long list gets you there faster.

Organizing messy notes. Voice memos, scribbled lists, bullet points from a meeting. AI is good at turning chaos into a clean list, a checklist, or a table.

Notice the pattern: in every case, you supply the substance and AI supplies the shape.

What AI does badly

Facts about your business it cannot know. AI does not know your hours, your prices, your return policy, or that you stopped offering delivery last spring. If you do not tell it, it will often fill the gap with something that sounds right. That is one of the easiest ways for AI to embarrass a small business owner.

Math it does not check. AI can produce numbers that look tidy and are simply wrong. Totals, percentages, quotes, and payroll figures need to be done or checked in a calculator or spreadsheet, not trusted from a chat window.

Legal and tax advice. It can explain general concepts and help you write down questions to ask. It cannot tell you what applies to your business, in your state, this year. That is a job for an accountant or attorney.

Having the final say on anything a customer sees. AI makes mistakes, and it makes them in the same confident tone it uses when it is right. Anything that goes to a customer, gets posted publicly, or carries your name needs a human to read it first. Every time.

A quick sorting table

Worked example: a neighborhood bakery

Maria runs a bakery with two part-time staff. Here is what her week of non-baking work looks like, and how she sorted it.

Good fits for AI:

• Writing the weekly "what's fresh" post for her social accounts. She gives it the list of specials; it drafts three versions; she picks one and fixes the wording.
• Replying to custom cake inquiries. She has a template answer, but each one needs adjusting. AI drafts the reply from her notes on flavors, sizes, and lead time.
• Turning her scribbled staff meeting notes into a clean list of who is doing what.

Poor fits for AI:

• Pricing a wedding cake order. The ingredient costs and her labor are numbers she has to work out herself.
• Checking whether her new outdoor seating needs a permit. That is a call to her city office, not a chat window.
• Telling a customer about allergens. AI does not know what goes into her recipes or what shares a kitchen with what. That answer comes from Maria.

Notice that the allergen question is the most dangerous one on the list. It sounds like a writing task, but it is really a facts task, and the facts only live in Maria's kitchen.

Try this

This is your task audit. Set a timer for 15 minutes.

Step 1. Write down every recurring non-core task from your last week. Not your craft or your service, but the admin, writing, and communication work around it. Aim for at least ten items.

Step 2. For each task, mark it:

• Words in, words out? (Yes or no)
• Does it need facts only I know? (If yes, can I paste them in?)
• Does a customer see the result? (If yes, I review before it goes out.)
• Does it involve math, legal, or tax judgment? (If yes, AI is not the one deciding.)

Step 3. Circle the three tasks that are words-in-words-out, repeat every week, and do not depend on legal, tax, or math judgment. Those are your first three.

If you want to get a head start, paste your list into an AI tool with this prompt:

I run [TYPE OF BUSINESS] in [CITY OR AREA]. Below is a list of the
non-core tasks I did last week.

Sort them into three groups:
1. Good fits for AI drafting (text-based, repeated, low risk)
2. Possible fits if I supply the facts
3. Poor fits (math, legal, tax, or decisions I should make myself)

For each task, give one sentence explaining why.
Do not add tasks I did not list.

MY TASKS:
[PASTE YOUR LIST]

Treat its sorting as a second opinion, not the answer. You know which of these tasks carry real risk. Keep your three picks handy: you will use them in the next two lessons.

What to remember

• AI is strongest at shaping words you supply: drafting, rewriting, summarizing, brainstorming, and organizing.
• It does not know facts about your business unless you give them to it, and it may invent them if you do not.
• Do the math yourself, and take legal and tax questions to a qualified professional.
• A human reviews anything a customer will see, every time.
• Start with three repeated, low-risk, text-based tasks, not your whole business.