No AI Policy at Work? How to Write Your Team's AI Rules Before Someone Writes Them Without You

22 minutes ago

7 min read

Diverse women in AI at a dinner table frowning at their phones, why every team needs an AI policy at work

Erica Anderson Rooney is an ex-CPO turned executive coach and AI educator on a mission to get more women into positions of power and keep them there. She is the author of The AI Gap (2026), host of the Glass Ceilings & Sticky Floors and AI Voice or Victim podcasts, and NC Chapter Lead for Women Applying AI. Her SNAP and AI Continuum frameworks are taught to thousands of women through CHIEF, WICT, and Fortune 500 leadership programs. She is the sought-after speaker at the intersection of women, AI, and leadership.

A year ago I was at the AI-Powered Women's Conference with about 100 of my closest AI female friends, and many of them went out to dinner, a table full of the smartest women I know in AI, founders and CEOs of AI companies, CTOs, Chief AI Officers and women who teach this stuff for a living. I stayed back at the hotel. One of them had an idea and said, okay, everybody type this into ChatGPT: "Based on everything you know about me, generate an image of what you think I look like."

Then the messages came rolling in on WhatsApp, and I read every single one. One after another, they got the same guy back, an old, pale, stale white male, with glasses on one screen and a goatee on the next. (The women building AI asked it to picture them, and it handed them a dude. You cannot make this up.)

One woman at that table got a picture that looked like her. But hanging on the wall behind her, in the background of the image, was a piece of artwork that read, "We men belong in the rooms where decisions are made."

My girl RBG would be rolling over in her grave. (Ruth Bader Ginsburg said women belong in all places where decisions are being made. The machine kept her sentence and swapped out the women. Freaking ridiculous.)

That night is the reason I wrote The AI Gap.

Nobody typed that rule in

What I found so interesting is that nobody sat down and wrote a rule that said the default expert is a man in glasses. Nobody typed it, approved it or signed off on it in a meeting. So why is this happening? Well… AI is a pattern recognition machine, and it learned from every pattern we have ever fed it, including the one about who gets to hold the power. History is written in the training data.

And that is what makes an unwritten rule so dangerous at work. You can't point to it in your performance review, you can't forward it to legal, and you can't argue with it, because officially it doesn't exist.

The first step is the least exciting word in business

I had Anne Cantera on Glass Ceilings & Sticky Floors this week. She is the founder and CEO of Elemental Intelligence, a responsible AI consulting firm for small and mid-sized businesses, and she has been building on the internet since 2001. So I asked her what the very first step is for a business that wants to do AI the right way. She told me the size of the business doesn't even matter. "It's governance and education."

Governance. I know, I know, it sounds like a binder nobody reads. But here is how Anne described the AI constitution she just built with her own team: "It's really just a set of rules. What will you, what won't you, what's acceptable, what isn't?" That's the whole thing, and most companies still don't have it.

Almost everybody is using AI, and barely a third wrote it down

In May 2026, ISACA surveyed more than 3,400 digital trust professionals and found that 90 percent believe employees in their organization are already using AI. Only 38 percent have a formal, comprehensive AI policy.

Hold up. Nine out of ten companies know their people are using it, and most of them never wrote down how? That means the people doing the work are making up the rules as they go, and they are the ones who get blamed when something breaks. (As a Chief People Officer, I'll tell you this for free: a rule nobody wrote down is a rule nobody can be held to, and that cuts both ways.)

The same survey found just 33 percent of organizations train ALL of their employees on AI. Anne told me about a designer whose team got in trouble for using up their whole week of paid AI usage by Wednesday. Nobody had ever shown them a better way to work with the tool. Her response was perfect: "That is a failure of leadership. He gave you a Ferrari, and you don't have a license." (And guess who gets written up when the Ferrari ends up in a ditch?)

Red sports car key and a blank learner permit beside a laptop on an office desk

Who gets a seat when the rules do get written

The World Economic Forum's Global Gender Gap Report 2026 used LinkedIn data to look at the people building these systems. Women make up 44.8 percent of data annotators, the people labeling the data that teaches the models. Women hold just 8.6 percent of chief technology officer roles.

So we are almost half of the people teaching the machine and fewer than one in ten of the CTOs deciding what it gets used for. Oh, hells no. Whoever sits in that CTO seat decides which tools get bought, what they are allowed to touch, and whose job gets "streamlined" first.

Woman in a navy blazer walking toward an empty chair at the head of a boardroom table

When AI becomes the excuse

Anne said something on the show that I haven't been able to shake. One of the reasons she left corporate was that she realized the multi-agent systems she was building "are going to be an excuse to fire people."

The data says she's onto something. In March 2026, Orgvue published a survey of 300 US HR managers and found that 32 percent of the organizations that made layoffs on the cost-saving promise of AI have already had to rehire staff. Almost a third of those decisions are getting walked back, and the people who got cut don't get those months back.

And the people most exposed are us. In April 2026, the National Partnership for Women & Families found that women make up 83 percent of the people working in AI-vulnerable jobs, about 6 million women. When the next one of those decisions comes around, the person holding the written rules is the person with the authority in that room.

What do I do if my company has no AI policy?

You write the first draft, the working one for your own team and your own scope. Open ChatGPT, Claude, Copilot or Gemini, whichever one your company lets you use, and paste this in:

You are an experienced AI governance advisor for small teams. Help me write a one-page draft of AI rules for my team of [number] people who [what your team does]. Use four headings: what we use AI for, what we never use AI for, what a person has to review before anything leaves our team, and what we do when the AI gets something wrong. Keep it in plain language and short enough that my team will follow it. Before you start, ask me up to five questions, one at a time, to help you get to certainty.

Put your name and the date on it, and send it to your boss as a draft for input. A draft that asks for input is really hard to say no to, and the woman who wrote the first draft is the one with her name on the rules when the official version gets written.

Woman at a desk writing a one-page draft with four sections, laptop open and a cup of tea nearby

Anne's advice for every woman who has real concerns about AI but knows she can't sit this one out was simple: "Don't be quiet." Your needs won't get fulfilled if you don't state them.

Run the prompt yourself

I still tell every audience to try that dinner prompt. At a keynote not too long ago, a Black woman, an Asian woman and white women in the room ran it, and they got that white man back too.

So run it tonight. Based on everything you know about me, generate an image of what you think I look like. What did yours give you? And who is writing the rules on your team?

Go write that first draft. I'm cheering you on.

Erica

Frequently Asked Questions

What should an AI policy at work include?

Start with four things: what your team uses AI for, what it never uses AI for, what a person has to review before anything leaves the team, and what happens when the AI gets something wrong. Add a date and an owner so it stays a living document. A one-page working draft beats a perfect policy that never gets written.

How many companies have an AI policy?

ISACA's 2026 AI Pulse Poll of more than 3,400 digital trust professionals found that only 38 percent of organizations have a formal, comprehensive AI policy, while 90 percent believe employees are already using AI at work.

What should I do if my company has no AI policy?

Draft one for your own team. Use an AI tool to build a one-page draft with four headings (what you use AI for, what you never use it for, what needs human review, and what you do when it gets something wrong), put your name and date on it, and send it to your manager as a draft for input.