An AI policy template for a small business should fit on one page and answer four questions: which tools staff may use, what they may paste in, what they must never paste in, and who checks the work before it reaches a customer. A usable template is further down, ready to copy and adjust.
A policy does not have to be formal. Its job is to remove guesswork. Without one, each person decides alone whether it is fine to paste a customer’s email, a supplier contract or a patient’s note into a free tool, and some will decide wrongly without meaning to.
Why a small business needs an AI policy template
Staff are probably using AI already, with or without your say. They draft emails, summarise documents and rewrite quotes. Most of it is harmless, and some of it is not.
The risks are simple to describe:
- Information leaves the business. What someone types into a tool may be stored or handled under the provider’s terms rather than yours.
- Wrong answers go out. AI tools write fluent text that is sometimes incorrect, and a customer may act on it.
- Nobody is accountable. When a mistake reaches a customer, “the AI wrote it” is not an answer they will accept.
- Customers or partners may ask. Clients and partners may ask how you use AI. A short written policy is something to show them.
A policy gives people permission to use the tools sensibly, which is what most owners actually want.
What to decide before you write it
Make six decisions first. They become the template’s blanks.
- Which tools are approved? Name them. Someone should have read each tool’s privacy terms.
- What can go in? Public information, your own marketing text, and anonymised examples are usually fine.
- What can never go in? Personal, health and financial details of customers or staff, passwords, and confidential contracts.
- What must a person check? Anything that goes to a customer, anything with a figure, anything with a legal or safety angle.
- Who is the owner? One named person answers questions, approves new tools and handles problems.
- What happens when something goes wrong? Say who staff tell, and that they will not be punished for reporting a mistake quickly.
For an official view, the Office of the Australian Information Commissioner publishes guidance on privacy and commercial AI products at oaic.gov.au. The Privacy Act does not apply to every small business, so check whether it covers yours. Our guide on AI and the Privacy Act covers the Australian rules in more detail. Our guides on using ChatGPT safely and whether ChatGPT keeps business data confidential go through the tool side.
An AI policy template you can copy
Copy this into a document, replace the square brackets, and cut anything that does not fit your business. It is a starting point and is not legal advice.
[Business name] AI use policy
- Purpose. This policy explains how we use AI tools at [Business name] so that we save time without putting our customers, staff or business at risk. It applies to everyone who works for us, including contractors.
- Approved tools. You may use only these tools for work: [list, for example: tool name and version or plan]. To use a different tool, ask [Owner name] first. Do not use personal accounts or free versions of tools for work unless this list says so.
- What you may enter. Public information, our published marketing material, general questions, and examples that have had names and contact details removed.
- What you must never enter. Customer or staff names with contact details, health information, bank or card details, passwords, identity documents, contracts or quotes marked confidential, and anything we have promised to keep private. If you are unsure, do not enter it and ask.
- Check everything. You are responsible for anything you send or publish, whether or not an AI tool drafted it. Before it leaves the business, check every price, date, name, figure and promise against our records. A second person checks anything involving [for example: quotes over a set value, safety advice, legal wording, health information].
- Customer messages. Do not send AI-written messages to customers without reading them in full. If a customer asks whether they are dealing with a person or an AI tool, tell them the truth.
- Do not rely on AI for. Legal, tax, medical or safety decisions, hiring or firing decisions, or anything where a wrong answer could harm someone.
- Intellectual property. Do not paste in material we do not have the right to use, and check that images or text from AI tools do not copy someone else’s work.
- Keep a record. For [important documents, quotes, reports], note that AI helped and who checked it.
- Report problems. If you enter something you should not have, or send something wrong, tell [Owner name] straight away. We would rather know early. You will not be penalised for reporting a mistake promptly.
- Review. [Owner name] reviews this policy every [six months] and whenever we add a tool.
Ask each person to read it, ask questions, and confirm in writing or by email. A policy nobody has read protects nobody.
Making the policy work in practice
A document on its own changes little. These habits make it stick:
- Talk it through in ten minutes. Give real examples from your business: a customer email, a quote, a supplier invoice. Ask which parts could go in and which could not.
- Put the “never enter” list near the work, such as pinned in your team chat.
- Make the approved tool easy to use. If the approved option is awkward, staff will use a free one. Provide a work account where you can.
- Spot check. Look at a few AI-assisted messages each month and see whether the checks happened.
- Update after each incident, even a small one.
Rolling it out in a week
- Day 1: ask staff which AI tools they already use, and for what. Do this without blame, so you get real answers.
- Day 2: make the six decisions and fill in the template.
- Day 3: read the privacy terms for each tool you plan to approve. Remove any you cannot check.
- Day 4: hold the ten-minute talk with the team and take questions.
- Day 5: collect confirmations, store the policy where everyone can find it, and put the review date in the calendar.
Treat the first version as a draft. After a month of use you will know which rules were too tight and which were missing.
Examples from ordinary businesses
A building company might add: “Do not paste client plans, site addresses or contracts into any tool. Use AI to tidy your own site notes after removing the address.” A clinic might say: “No patient information goes into any AI tool unless the practice manager has approved that tool in writing for that use.” An accounting practice might add: “Never enter a client’s tax file number, bank details or financial statements into an unapproved tool, and a second person reviews any AI-assisted advice.”
The wording is yours. The test is that a new starter could read it on their first day and know what to do.
When a policy is not enough
A policy relies on people following it. If you want the work done in a controlled way, a system can do some of the enforcing for you: an assistant that answers only from your approved information, writes to your own CRM and hands over to a person when it should. CentriWeb builds that kind of AI system, and it is not always needed. For a team of a few people using a single approved tool, a one-page policy and a monthly check is often enough.
If your work involves sensitive information, regulated advice, or contracts with strict data terms, ask a lawyer or adviser to look over your policy before you roll it out.