Transformidable Advisory · AI Governance

Your staff are already using AI. Start with an acceptable use policy.

In most organizations, AI arrived before the policy did. Staff are drafting, summarizing, and analyzing with generative AI tools, vendors are switching on AI features in software you already own, and some departments are buying AI on their own.

Banning it rarely works. What works is deciding, deliberately, which uses matter, who gets to say yes, and who answers for the results.

Signs you need a policy now

  • Staff use ChatGPT, Copilot, Gemini, or similar tools, and no one has said what’s allowed.
  • Someone has pasted donor, student, member, patient, or employee information into an AI tool.
  • A vendor has switched on AI features in software you already pay for.
  • A department wants to buy an AI product, and there’s no process for evaluating it.
  • Your board has asked “what are we doing about AI?” and the answer was a shrug.

Having a policy isn’t the same as governing AI

In EY’s 2026 AI Risk and Governance Survey, 98% of organizations said they have formal AI governance policies. 47% said they had skipped their own governance process to meet an urgent deadline.

That isn’t a discipline problem. It’s governance built for a slower world: when the process is slower than the technology, people go around it. A policy proves you have a policy. It doesn’t show anyone followed it.

Govern the decisions, not the tools

AI changes by the week: new tools, new models, new features in software you already own. A policy built around a list of approved tools is out of date almost as soon as it’s written. What stays put are the decisions: when a use matters, who decides, who owns the result, and what you can show when someone asks how it’s working. Build the policy on those.

What an AI acceptable use policy should cover

A useful policy is short enough that people read it. It should answer:

  1. When a use matters. Clear thresholds for when an AI use needs more attention, such as when it touches sensitive information or influences decisions about people.
  2. Who decides. Who can approve which uses, at the right level, so low-stakes uses aren’t stuck waiting on the top of the organization.
  3. A fast lane. Light checks for everyday, low-stakes uses, so the legitimate path is faster than going around it.
  4. Information that must never go in. Confidential, personal, and regulated information, named specifically for your organization.
  5. Who owns the result. IT runs the system; the business owns what it produces. Each use that matters needs a named owner.
  6. Approval with conditions. Approving a use isn’t a one-time event. Decide up front what would send it back for review.
  7. Human review and disclosure. Who checks AI output before it’s relied on, and when to say AI was used.
  8. Vendors. What contracts must say about your data, and how you’ll notice AI features a vendor switches on.
  9. Training. How staff learn what the policy means in their own work.

Five questions to test your policy

The most important question isn’t “Do we have a policy?” It’s “Could we explain how it actually works, in our organization, right now?” Five questions, drawn from the Transformidable Governance Method, put it to the test:

  • See. Can we spot the AI uses that matter, including what vendors switch on?
  • Decide. Is it clear who can approve them?
  • Own. Is someone on the business side accountable for the outcome?
  • Escalate. Do we know what needs leadership's attention, and who can pause it?
  • Evidence. Could we show that governance is actually working?

None of them is really a technology question. They’re questions about how the organization runs, which is why AI governance can’t live only in IT.

The briefing: “Are we governing AI, or just letting it happen?” Dr. Jerri Bland walks through each of the five questions, where to draw the lines, what should send an approved use back for review, and a real scenario: the day your HR vendor switches on AI. See the dates and register.

AI policy for nonprofits

Nonprofits often face the same AI decisions as large organizations with a fraction of the staff. The biggest risks are usually donor and client information entered into free tools, grant and funder requirements about data, and AI-generated content that goes out under the organization’s name. Keep the policy short, name the information that must never be entered, and give staff a fast, approved path for everyday uses rather than none.

AI policy for colleges and universities

Universities need to separate two conversations that often get blended: AI in teaching and academic integrity, which belongs with faculty and shared governance, and AI in administration (admissions, financial aid, advising, HR, and operations), which involves student records and regulatory obligations such as FERPA. Many institutions have the first policy and not the second. Governing boards should ask which one they’re looking at.

How Transformidable helps

  • Advisory. We can draft your AI acceptable use policy with you, or review the one you have, and shape the governance around it: who decides, who owns each use, what escalates, and what you report to the board.
  • Academy. TA-101 Know What You’re Governing, a 90-minute live course, gives your board a shared vocabulary, the questions to ask, and the Before You Vote on AI card. See all Academy courses.
  • Briefings. Upcoming live briefings and recorded briefings on AI governance for boards and executives.

Questions

Is a generative AI policy different from an AI acceptable use policy?

They usually mean the same thing in practice: rules for how staff may use AI tools like ChatGPT, Copilot, and Gemini. "Acceptable use" emphasizes what's allowed; "generative AI policy" names the kind of tool. A broader AI governance approach also covers how the organization decides, owns, and oversees the AI uses that matter.

Should our policy list approved AI tools?

A short list can help with everyday, low-stakes uses. But a list of tools is out of date almost as soon as it's written, because tools, models, and features change by the week. Base the policy on what stays put: when a use matters, who decides, and who owns the result.

Should we just ban AI tools?

Bans are hard to enforce and tend to push use out of sight, where the risks are worse. A fast, legitimate path for low-stakes uses keeps people inside the process instead of around it.

Who should own the AI policy?

A named executive, not a committee. Committees can advise, but one person should be accountable for the policy, its exceptions, and reporting to the board. Each AI use that matters also needs a named business owner for its results.

How often should an AI policy be reviewed?

Review the policy itself at least annually. Individual uses need a second look sooner, whenever something about them changes or goes wrong, so decide when you approve a use what would send it back for review.

Do we need a lawyer to write it?

Involve legal counsel, especially where you handle regulated information. Transformidable advises on governance and oversight; we are not a law firm and don't give legal advice.

Tell us where you are with AI.

No policy yet, a policy nobody follows, or a board asking questions management can't yet answer. We'll tell you honestly whether and how we can help.