Studio
How to Write an AI Usage Policy for a Small Business
By the llcloud team 7 min read
Quick answer
An AI usage policy is a short document that tells your staff which AI tools they may use, what data may go into them, who checks the output, and when to tell customers AI was involved. A small company can write a useful one in a few pages and review it twice a year. llcloud offers fractional CAIO services and can write and roll out the policy with your team.
Your team is almost certainly using AI already. Someone drafts emails with a chatbot, someone summarises meeting notes, someone pastes a spreadsheet in to get a formula. None of that is a problem on its own. The problem is that each person is making their own rules about what data goes where and whether to check the answer. An AI usage policy replaces those private rules with shared ones. Here is how a small company writes one that people will actually follow.
This article is general guidance, not legal advice. For your specific legal duties, talk to a lawyer who knows your country and sector.
What is an AI usage policy?
An AI usage policy is a short internal document that says how your company uses AI tools: which ones are allowed, what data may go into them, who checks the output, and when you tell people AI was involved. It is a working document for staff, not a statement for your website.
For a small business it should fit on two to four pages. A policy nobody reads protects no one. Writing it is part of the governance work a Chief AI Officer owns, but in a small company the owner is often the founder, an operations lead or a fractional CAIO.
Which sections does a small business AI policy need?
It needs eight sections, each short. Use this table as your outline.
| Section | What it answers | Typical length |
|---|---|---|
| Purpose and scope | Why the policy exists, who it covers (staff, contractors) | A paragraph |
| Approved tools | Which AI tools are allowed, on which accounts | A short list |
| Data rules | What may and may not go into an AI tool | Half a page |
| Human review | Who checks AI output before it is used | A paragraph |
| Disclosure | When customers or readers are told AI was involved | A paragraph |
| Privacy and legal duties | Which laws and contracts apply to your AI use | Half a page |
| Training | What every person learns before using AI for work | A paragraph |
| Ownership and reporting | Who owns the policy, how to report a problem or request a tool | A paragraph |
Which AI tools should staff be allowed to use?
Name them. An approved list is clearer than a list of banned tools, because new tools appear faster than any ban list can keep up.
For each approved tool, write down:
- The tool and the plan. A business or team plan often has different data terms from a free personal account, so say which one is approved.
- What it is approved for. Drafting, summarising, coding help, translation, image creation.
- Who manages the account. Company accounts mean you can remove access when someone leaves.
Then add one rule: anything not on the list needs approval before it is used for work. Give people a simple way to ask, such as a message to the policy owner, so the rule does not push them into using tools quietly.
What data may go into an AI tool?
This is the most important section. Sort your data into three groups and say what each group may do.
| Data type | Examples | Rule |
|---|---|---|
| Public | Published marketing copy, public web pages | Fine in any approved tool |
| Internal | Plans, internal docs, non-personal figures | Approved company accounts only |
| Restricted | Customer personal data, health or financial data, passwords, client confidential material, source code under NDA | Never, unless a specific tool is approved for it in writing |
Keep the examples concrete and from your own business. "Do not paste customer emails into a chatbot" is easier to follow than "do not process personal data in unapproved systems".
A useful habit to teach: before pasting anything, remove names, email addresses and account numbers. Where possible, prefer tools that process data on the device or under a contract that excludes your data from training.
Who checks what the AI produces?
A person does, every time AI output reaches a customer, a decision or your codebase. Write that down plainly.
AI tools produce confident text that can be wrong. The European Commission's AI literacy guidance names hallucination as a specific risk that staff using general chatbots should be told about. Your policy should say:
- Facts, figures, quotes and legal or medical statements are checked against a real source before use
- Code written with AI is reviewed the same way as any other code
- AI is never the only decision maker on hiring, pay, credit, or anything else with a serious effect on a person
- The person who uses the output is responsible for it, not the tool
When should you tell people AI was involved?
Tell them whenever they would reasonably expect to know. In practice that means customers talking to a chatbot, content that could mislead if presented as fully human, and any decision about a person.
The EU AI Act includes transparency rules along these lines: people should be made aware when they are interacting with a machine, and certain AI-generated content must be labelled. Even where those rules do not reach you, disclosure is good practice because trust is cheaper to keep than to rebuild. Your policy can be simple: "If a customer talks to a bot, say it is a bot. If we publish content that is substantially AI-generated, a person reviews it and we follow our disclosure rule."
Which privacy and legal duties does the policy need to cover?
List the ones that apply to you, in plain words, and point to where the detail lives. For most small companies that means three areas.
- Data protection law. If personal data goes into an AI tool, your data protection duties apply, including having a lawful basis and a contract with the vendor where needed. Check whether the vendor processes data in a country your law allows.
- AI-specific law. The EU AI Act sorts AI systems by risk, from banned practices through high-risk uses such as hiring and education to systems with no specific rules. It also asks companies that use AI systems to support AI literacy among their staff. The Commission's own Q&A says this covers a company whose employees use a general chatbot for tasks like writing or translation, and that no certificate is needed: an internal record of training is enough. The rules have been amended since they were adopted, so read the current text rather than an old summary.
- Contracts. Client contracts and NDAs often limit where their data may go. Your data rules have to respect them.
If you want a framework to structure your thinking, the NIST AI Risk Management Framework is free, voluntary and organised around four functions: govern, map, measure and manage. A small company does not need to adopt all of it, but its governance section is a good checklist for what your policy should cover.
How do you get people to follow the policy?
Train them, keep it short, and make the approved path the easy one. A policy that blocks useful work gets ignored.
- Train once, then refresh. A short session on the data rules and on checking output covers most of it. Keep a simple record of who attended, which also helps with AI literacy duties.
- Name an owner. One person keeps the approved tool list and answers questions.
- Make reporting safe. If someone pastes the wrong data, you want to hear about it within the hour, not discover it in six months. Say clearly that honest mistakes reported quickly are not punished.
- Review twice a year. Tools change their terms often. Put the review in the calendar.
How llcloud can help
llcloud is an independent studio that builds apps, and we also work with companies as a fractional Chief AI Officer. AI governance is part of that: we write your usage policy, pick and set up approved tools, define data rules that fit your business, and train your team so the policy is followed, part time and sized to your company. The rest of the role, from the first 90 days to how a CAIO differs from a CTO, is covered in this series, along with when to hire a Chief AI Officer.
Our approach is privacy first because that is how we build. We build and ship our own apps, with no ads, no account and data kept on the phone wherever we can.
To talk about an AI policy for your company, write to support@llcloud.app with a few lines about your team and the tools you use today.
Sources
Questions
Does a small business need an AI usage policy?
Yes, as soon as anyone on the team uses AI tools for work, which is most teams. Without a policy, each person decides alone what data to paste into which service and whether to check the result. A short written policy gives everyone the same rules, protects customer data, and helps you show regulators and clients that you take AI use seriously.
What should an AI usage policy include?
At minimum: a list of approved tools, clear rules on which data may and may not go into them, a requirement that a person reviews AI output before it is used, rules on telling customers when AI was involved, an owner for the policy, training for staff, and a simple way to report a mistake or ask for a new tool. Most small companies can cover this in two to four pages.
Is there a legal requirement to have an AI policy?
It depends on where you operate and how you use AI. The EU AI Act asks companies that use AI systems to support AI literacy among their staff, and data protection law applies whenever personal data goes into an AI tool. A written policy is often the simplest way to show you meet those duties. This article is general guidance, so check your own obligations with a lawyer.
How often should we update our AI policy?
Review it at least twice a year, and also whenever you adopt a new AI tool, a vendor changes its data terms, or a law that applies to you changes. AI tools change their features and terms often, so a policy written once and left alone goes stale quickly. Give one named person the job of keeping it current.
Can llcloud write our AI usage policy?
Yes. llcloud works with companies as a fractional Chief AI Officer, and AI governance is part of that work: writing the usage policy, choosing approved tools, setting data rules, and training your team so the policy is actually followed. Write to support@llcloud.app with a few lines about your company and how your team uses AI today.