AI Readiness Assessment
How do I write an AI policy for my employees?
A good employee AI policy fits on one page and answers five things: which AI tools are approved, what data never goes into them, who reviews AI output before it leaves the building, who to ask when something is unclear, and how the policy gets updated. Write it in plain language, name an owner, and give people the approved option so they stop reaching for the unapproved one.
Most owners we talk to assume they do not need an AI policy yet because they have not "adopted AI." That is rarely true. Microsoft and LinkedIn's 2024 Work Trend Index found that 75% of knowledge workers use AI at work, and 78% of those bring their own AI tools. Your staff are probably using it right now, on personal accounts, with your customer data. A policy is how you move from hoping that goes well to deciding how it goes.
Why does a company past $10M need a written AI policy?
Three reasons, in the order owners tend to care about them.
- Confidential data leaves the building. Pricing, customer lists, contracts, employee records and financials can all end up pasted into a public tool.
- Wrong output gets treated as right. AI tools sound confident when they are wrong. Without a review rule, a made-up number can land in a customer quote or a lender package.
- Other people will ask. Customers, lenders, insurers and PE sponsors increasingly ask whether you have an AI policy. "We're working on it" is a weak answer. A one-page document is a strong one.
This is the Governance pillar in our AI readiness assessment, and it is usually the cheapest gap to close. A policy takes a few hours, not a few months.
What should an AI acceptable use policy include?
Five sections. If a section grows past a short paragraph or a short list, you are writing a handbook, not a policy. Keep the detail in a separate reference sheet that you can update without reissuing the policy.
- Approved tools. A named list. Be specific: product, plan and who has access. "Company-licensed Microsoft Copilot" is a tool. "Any reputable AI" is not.
- Data that never goes in. A concrete list of categories. People follow lists better than principles.
- Review of AI output. A person is responsible for anything AI helped produce. Define what must be checked and by whom.
- Who to ask. A named person and a way to reach them. If people cannot get a quick answer, they will guess.
- How it gets updated. An owner, a review date and how people suggest changes.
Not sure how exposed you are today? Find out where you stand across all six pillars, Governance included.
Get your AI Readiness ScoreWhat does a one-page AI policy template look like?
Here is a template you can copy, edit and use. Replace the bracketed items with your own. Have your attorney or HR advisor look it over before you issue it, especially if you are in a regulated field such as healthcare.
| [Company name] Acceptable Use of AI Tools | |
|---|---|
| Purpose | AI tools can save time. This policy tells you which ones to use, what to keep out of them, and who is accountable for the result. It applies to every employee, contractor and temp. |
| 1. Approved tools | You may use only these tools for company work: [Tool A, plan, who has access]; [Tool B, plan, who has access]. If a tool is not on this list, do not use it with company information. To request a new tool, contact [name]. |
| 2. Data that never goes in | Never enter into any AI tool, approved or not, unless the tool is specifically approved for that data: customer names and contact details; employee or applicant records; pricing, margins and cost data; contracts and legal documents; financial statements and bank information; passwords and access keys; patient, health or other regulated data; anything a customer has told us to keep confidential. When in doubt, leave it out and ask. |
| 3. Review of AI output | You are responsible for anything you send or act on, whether or not AI helped. Check every fact, number, name and quote before it leaves the company. Do not send AI-written content to a customer, lender or regulator without a human read-through. Anything that affects pay, hiring, pricing or legal commitments needs a second person's review. If AI materially shaped a deliverable, tell your manager. |
| 4. Who to ask | Questions, concerns or possible mistakes: contact [name, title, email, phone or chat channel]. Reporting a mistake quickly is always the right call. You will not be penalized for raising a concern in good faith. |
| 5. Updates | This policy is owned by [name, title]. It is reviewed every [quarter] and whenever a new tool is approved. Last reviewed: [date]. Suggest changes to [name]. |
| Consequences | Violations may lead to disciplinary action, up to and including termination, consistent with company policy. |
| Acknowledgment | Name, signature, date. |
How do you get people to actually follow it?
A policy that lives in a folder does nothing. Four habits make it real.
- Provide the approved tool. If you ban public tools and give people nothing, they will keep using them quietly. Pay for a licensed option with business data protections.
- Explain the why in ten minutes. Show one example of what happens when a customer list is pasted into a public tool. Stories stick better than rules.
- Make asking easy. Name a person, not a department. A fast "yes, that's fine" builds the habit of asking.
- Treat the first mistake as information. If the first person to report a slip is punished, you will not hear about the second one.
Who should own the policy, and what should it not try to do?
Give it to one executive. IT should help set up the tools and the technical controls, but they should not be left to decide what the business accepts as risk. That is a management decision. If you are not sure who that is, that is a finding in itself, and it usually points to a larger gap in technology leadership. We write about that in Is your IT manager your technology strategy?
The policy should also not try to cover every future scenario. It will not tell people how to build an AI model, choose a vendor or run a pilot. Those belong in your broader AI plan. For a view of the whole picture, our AI Readiness Assessment covers the six pillars and ends with a short list of what to fix first.
Can employees paste company data into ChatGPT?
Not unless you have approved that tool for that kind of data. Public, consumer versions of AI tools may keep what people type in, and you have no say over where it goes next. A company-licensed plan with business data protections is a different thing, and your policy should name which one people may use for what.
The safe default is a short list of data that never goes into any AI tool unless it is specifically approved: customer names and contact details, employee records, pricing and margins, contracts, financial statements, passwords and anything regulated. Pasting a customer list or a draft contract into a free public tool is a common mistake. Check the terms of any tool you approve, because they differ by plan, and have your attorney look at the policy before you issue it.
Who approves new AI tools?
One named person, with a short path to a yes or a no. Usually that is the executive who owns the policy, with input from whoever runs IT and from legal or HR for anything that touches customer, employee or regulated data. Do not leave it to individual managers, because every team will approve a different tool.
Make the process small enough to use. Ask for the tool, the plan, what data it would see and what problem it solves. Check the vendor's terms on data use and retention, decide, and add the tool to the approved list with its plan and who has access. Aim for an answer in days. If asking is slow, people stop asking and use the tool anyway. Review the list every quarter and remove tools nobody uses.
Common questions
Do we need an AI policy if we have not rolled out any AI tools?
Yes. Your people are probably already using AI on their own. A policy tells them what is allowed, which is better than finding out after something sensitive has been pasted into a public tool.
Should we just ban AI tools?
A ban rarely works. People keep using the tools on personal devices and accounts, and now you cannot see it. A short policy with approved tools and clear red lines gets you more safety than a ban.
Who should own the AI policy?
One named executive, usually the CEO, COO or CFO, with input from whoever runs IT and from legal or HR. Policies owned by nobody go stale in a quarter.
How often should we update it?
Review it every quarter for the first year, and any time you approve a new tool or have an incident. The tools change fast, so the policy needs a date and an owner.
Not sure where to start?
Take a free three-minute score. If a conversation makes sense, it will offer one.