All Posts

Your team is already using AI. Did anyone decide the rules?

Someone on your team pasted a customer email into a chatbot this week. Maybe a quote, maybe an invoice. Nobody decided whether that was OK. A one-page AI acceptable use policy fixes that. Here's a template you can copy, and what to check before anyone connects AI to your email, files or CRM.

November 16, 20266 min read

The short version.

Your team is already using AI. Many small businesses never decided the rules for it.

An AI acceptable use policy is one page that answers four questions:

  • Which tools can we use?
  • What data never goes into them?
  • Who checks the output?
  • Who approves connecting AI to our systems?

The template below covers all four. Copy it, change the names, and hand it to your team this week.

If you'd rather have help, writing an AI use policy with you is part of our AI consulting work. But you can do this one yourself.

The problem nobody decided.

Here's how it usually goes.

Someone on the team finds a chatbot that saves them an hour. They start using it for emails. Then for quotes. Then they paste in a customer's file to get a summary.

Nobody did anything wrong. Nobody said it was OK, either.

The questions come later:

  • Was that the free version?
  • Does it keep what you paste in?
  • Was there a Social Security number in that file?

In our experience, owners find out what their team is doing with AI by accident. A policy means you decide on purpose.

What an AI acceptable use policy is.

It's not a legal document. It's a set of plain rules your team can read in two minutes.

A good one is short. If it's longer than a page, nobody reads it, and a policy nobody reads protects nobody.

It should answer:

  • Which tools are approved. By name. Not "AI tools," but which ones.
  • Which plan. Many AI tools offer a business plan that doesn't use your data to train their models. The free version may not. Check the settings and the terms for each tool.
  • What data never goes in. Customer financial details, passwords, health information, and anything a client asked you to keep private.
  • Who checks the output. AI writes the first draft. A person signs off.
  • Who approves connections. Linking AI to your email, files or CRM is a bigger decision than using a chatbot. Someone specific should own it.

If you're in a regulated field like health care, finance or law, have your attorney or compliance lead review it. This post isn't legal advice.

The template.

Copy this, replace the brackets, and cut anything that doesn't fit your business.

[Company name] AI Use Policy

Updated: [date]. Owner: [name], who answers questions and approves changes.

1. Why we have this. AI helps us work faster. These rules keep our customers' information safe while we use it.

2. Approved tools. You may use these tools for work, on our business accounts only:

  • [Tool name], [business plan name]
  • [Tool name], [business plan name]

Any other AI tool needs [owner]'s OK first. That includes browser extensions and phone apps.

3. Never put these into any AI tool:

  • Social Security numbers, bank or card numbers
  • Passwords or login codes
  • Health or medical information
  • Anything a customer or client asked us to keep private
  • [Anything specific to your business]

4. A person checks everything. AI can write first drafts, summaries and suggestions. Before anything goes to a customer, gets filed, or changes a record, a person reads it and owns it.

5. Connecting AI to our systems. Don't connect any AI tool to company email, files, calendars, our CRM or accounting without [owner]'s written OK. Answer the connection checklist questions first.

6. When something goes wrong. If you put something into an AI tool that you shouldn't have, tell [owner] the same day. Nobody gets in trouble for reporting it. We'd rather fix it early.

7. Review. [Owner] reviews this policy and our list of connected tools every [three] months.

Before anyone clicks Allow.

Connecting AI to your business systems is where the real risk lives.

A chatbot only sees what you paste into it. A connected tool can see everything it's given access to. Some can also act on it: send emails, change records, delete files.

That screen asking you to click "Allow" is the whole decision. Most people click through it in two seconds.

Before anyone approves a connection, check these:

  • What does it want to see? All of your email, or one folder? Every file, or one shared drive?
  • Can it only read, or can it act? If it only needs to read your inbox, it shouldn't be able to send from it. Look for words like "send," "modify," "delete" and "manage" in the permission list.
  • Does it need all of that? If a tool asks for more than the job requires, pick a narrower option or a different tool.
  • What needs a person's OK? Anything that sends to a customer, moves money, or deletes something should wait for a person to approve it.
  • Who approved it, and when? Write it down. A simple list works.
  • How do we turn it off? Know where to remove the connection before you need to.

This is the same rule we follow when we build automations for clients. Each one gets only the access it needs, and anything that sends, changes or deletes waits for a person's OK when the stakes call for it.

A policy written once isn't enough.

In our experience, businesses that write an AI policy write it once and move on.

Meanwhile the tools changed. Someone added a new extension. A connection made for one project is still sitting there with access to everything.

The fix is boring. Put it on the calendar.

Every few months, the policy owner:

  • Lists every AI tool the team uses. Ask. People will tell you.
  • Checks each connected tool's access, and removes what it doesn't need.
  • Removes connections for tools nobody uses anymore.
  • Updates the approved list and the date on the policy.

It takes an hour. It's the difference between having a policy and following one.

Where this rule came from for us.

We ran a four-hour AI training for CPAs at Wilkes University. Accountants hold some of the most sensitive data any small business touches.

One rule came up all session. Client data never goes into a free tool. AI does the first draft and the prep. A person does the planning and signs off.

That rule fits almost every small business, not just accounting firms. It's the core of the template above.

You can read how that session went in the CPA training case study.

Want help putting it in place?

You can write this policy yourself. The template above is enough to start.

If you want a second set of eyes, we review which AI tools your team already uses, what each one can see and do, and what it doesn't need. Then we write the policy with you, in your words.

Book a 20-minute call and we'll start with what your team is using today.

Common Questions

Frequently asked questions

Does a small business need an AI policy?+

If anyone on your team uses AI for work, yes. It doesn't need to be long. One page that names the approved tools, the data that never goes in, who checks the output, and who approves connections is enough for most small businesses.

Can employees use ChatGPT for work?+

They can if your policy allows it. Most businesses approve a business plan of one or two tools, check that the plan doesn't use their data for training, and set clear rules for what data never goes in.

What should an AI acceptable use policy include?+

The approved tools and plans, the data that never goes into any AI tool, a rule that a person checks AI output before it's used, who approves connecting AI to company systems, what to do when something goes wrong, and when the policy gets reviewed.

Is it safe to connect AI to company email?+

It can be, if you check the access first. Give the tool only what the job needs. If it only needs to read, it shouldn't be able to send, change or delete. Anything that acts on your behalf should wait for a person's OK.

How often should we update our AI policy?+

Every few months, and whenever the team starts using a new tool. Review the approved list, check what each connected tool can access, and remove anything nobody uses.

Let's talk.

20 minutes. No pitch. No pressure. We'll ask questions, listen to how things work, and tell you whether there's something worth fixing.