ChatGPT earns its keep in a small business as a fast first-draft writer and reader: replies to customers, summaries of long documents, quotes and checklists built from your notes, job postings, and spreadsheet fixes. Use a business plan for work instead of personal accounts, keep personal and confidential information out of it, and have a person check everything before it goes out.
That covers most of it. The rest of this post is the detail: prompts you can copy this week, where it goes wrong, what never to paste in, and what OpenAI’s own pages say about who trains on your conversations. When a business is ready for more than copy and paste, our AI integration work picks up from there. We’ll get to what that looks like at the end.
Five ways to start this week
Each of these takes a few minutes, and each one ends with you reading the result before anyone else sees it. Swap the bracketed parts for your own details. Leave out names, phone numbers and account numbers. The prompts work just as well without them.
Customer emails and replies
Paste the customer’s question with their name and contact details taken out, then ask for a reply in your voice.
Write a short, friendly reply to this customer question. We can come out next Tuesday or Thursday morning. Keep it under 120 words and don’t promise a price. Question: [paste question]
Summaries of long documents
A vendor’s terms of service, a long equipment manual, a city permit checklist. Ask for the parts you’d act on, and ask it to show its work.
Summarize this in plain English for a business owner. List anything I have to do, any deadlines, and anything that costs money. Quote the exact line each point comes from. [paste text]
Asking for the exact line makes checking fast. If it can’t point to a line, treat that point as a guess.
Notes into a quote outline or an SOP
Here are my rough notes from a site visit. Turn them into a clean quote outline with the scope of work, what’s not included, and the questions I still need answered. Don’t add prices. [paste notes]
Here’s how we open each morning, in my own words. Write it as a numbered checklist a new hire could follow, and flag any step that looks missing. [paste notes]
First drafts of job postings and policies
Draft a job posting for a part-time front-desk coordinator at a [type of business] in [town]. Plain language, no buzzwords. Duties: [list]. Hours: [hours]. Leave the pay as [PAY] for me to fill in.
Draft a simple appointment cancellation policy in plain English. We need 24 hours’ notice. Keep it under 150 words and friendly in tone.
Job posts and policies can carry legal weight, from pay rules to refund terms. Treat the draft as a starting point, and have someone who knows the rules where you operate look at anything that matters.
Cleaning up a spreadsheet
You often don’t need to paste the data at all. Describe the columns and ask for the formula.
My Google Sheet has job dates in column A written different ways, like 10/3, Oct 3 and 2026-10-03. Write a formula for column B that turns them all into one date format, and tell me how to fill it down the column.
Nothing in that prompt is customer data, and you get a fix you can reuse every month.
What it gets wrong
It writes well, which makes its mistakes easy to miss. Watch for three kinds:
- Made-up facts. It can produce a product spec, a code section, a court case or a statistic that doesn’t exist, in the same confident tone as everything true around it.
- Confident errors. Wrong math in a quote, a misread figure from a document, a step dropped from a procedure. It rarely flags its own doubt unless you ask, and sometimes not even then.
- Out-of-date information. What a model learned stops at a point in time, and when it does look things up online it can still land on an old page. Prices, tax rules, permit requirements and vendor policies all change.
The rule that covers all of it: a person checks everything before it goes out. Read every customer reply before you hit send. Check every number in a quote against your own price sheet. Open every link and source it gives you. If you can’t check a claim, cut it. ChatGPT drafts, a person decides.
What never to paste in
Keep these out of any AI chat, whatever plan you’re on, unless the system was built and set up for that kind of data:
- Customer personal information. Names with addresses, phone numbers, birthdays, account numbers, anything that points to a real person.
- Health or payment information. Patient details, insurance IDs, card numbers, bank details. That data comes with its own legal duties.
- Passwords, API keys and login codes. If one ends up in a chat by accident, change it.
- Contracts and documents under an NDA. If you promised to keep it confidential, pasting it into an outside service may break that promise.
- Employee records. Pay, reviews, medical notes, disciplinary files, Social Security numbers.
A simple test: if you’d be uneasy seeing it on a stranger’s screen, don’t paste it.
Personal accounts vs business plans: who trains on what
We checked OpenAI’s own pages on October 7, 2026. These policies change, so check them again before you rely on them.
Personal accounts (Free, Go, Plus and Pro). OpenAI’s help article on how your data is used to improve model performance says that for its services for individuals, “we may use your content to train our models.” You can opt out by turning off Improve the model for everyone under Settings > Data controls, and after that your new conversations aren’t used for training. Two details from OpenAI’s data controls article are worth knowing. Giving a thumbs up or thumbs down on a response can send that whole conversation for training even after you’ve opted out. And a Temporary Chat isn’t used to improve models, but may be kept for up to 30 days for safety.
Business plans (ChatGPT Business and Enterprise). OpenAI’s business data page says that by default it does not use data from these plans, inputs or outputs, to train or improve its models. Its enterprise privacy page says ChatGPT Business workspace admins can view, export and delete conversations in the workspace, and can control how long data is kept. The business data page lists two-step sign-in, Member, Admin and Owner roles, a switch for third-party GPTs, and single sign-on. Enterprise adds more, like automatic account setup and removal and custom roles.
| Personal account | Business workspace | |
|---|---|---|
| Used to train OpenAI’s models? | May be, unless you turn the setting off | Not by default |
| Who controls the settings | Each person, in their own account | The workspace admin |
| Can the business see and export work chats? | No, they sit in the employee’s own account | Yes, admins can view, export and delete them |
Training gets the headlines, but for an owner the last row matters just as much. When staff use personal accounts for work, your quotes, customer questions and procedures sit in accounts the business doesn’t control and can’t review, and they leave when the employee does. OpenAI’s article on managed accounts also notes that a personal account and a work account stay separate even when someone switches between them, so make sure work happens in the work account.
A business plan improves how your data is handled. It doesn’t cancel the list above. OpenAI’s business data page, for example, mentions a Business Associate Agreement only for its healthcare products and API healthcare customers.
When copy and paste stops being enough
Pasting into a chat window works for a single user and a handful of tasks. You’ve outgrown it when:
- staff paste the same price sheet or policy into a chat every morning
- the same customer questions get answered from the same documents, over and over
- nobody can tell whether an answer came from your policy or from the model’s imagination
- the information it needs is spread across email, PDFs and somebody’s head
That’s the point where we set up an assistant that answers from your own documents, price sheets and procedures, with the source attached so anyone can check it. It runs inside accounts you own, with role-based logins and settings that keep your data out of public AI model training. It answers only from material you approve, says so when it doesn’t know, and hands the question to a person. When it drafts something for a customer, a person reviews it before it goes out. We wrote up the rest of how we treat client information on our data handling page.
Sometimes a business workspace and the habits in this post are all you need. When that’s true, we’ll say so. And if the work that eats your day is answering the phone, that’s a different tool, covered in our guide to AI receptionists.
If you’d like a second opinion on where AI fits in your business and where it shouldn’t, that’s what the call below is for.