All resources

Building AI governance your team can actually follow

AI governance often fails for a simple reason: it is written for auditors, not for the people using the tools.

A policy that says “use AI responsibly” does not help a marketing assistant decide whether to paste client data into a chatbot. A 30-page document nobody opens does not prevent an employee from using an unapproved tool under deadline pressure.

Governance becomes real when it helps people make better decisions in the flow of work.

Why AI governance matters

Teams are already using AI: to draft content, summarise documents, analyse data, answer customer questions and speed up research. The question is no longer whether AI is present in your organisation.

The question is whether it is being used with appropriate judgement.

Without governance, common risks include:

  • Sharing confidential or personal information with tools that are not approved
  • Publishing inaccurate or biased AI-generated content
  • Losing track of which decisions were made with AI support
  • Creating dependency on tools without understanding their limits
  • Using AI in ways that conflict with your values, contracts or legal obligations

Governance is not about stopping AI use. It is about making responsible use practical.

What useful AI governance includes

1. Clear principles

Start with a small number of principles your team can remember.

For example:

  • Protect people’s data and privacy.
  • Check AI outputs before using them.
  • Be transparent when AI has a material role in work.
  • Keep humans accountable for decisions.
  • Use AI to support your mission, not replace judgement.

Principles give people a shared reference point when a policy does not cover every situation.

2. Approved tools and use cases

People need to know what they can use, for what, and under which conditions.

A practical approach is to define:

  • Approved tools
  • Tools that require approval
  • Tools that are not permitted
  • Acceptable use cases
  • Prohibited use cases
  • Data that can and cannot be entered

This turns an abstract policy into a clear decision guide.

3. Data rules people understand

Data rules are often where governance breaks down.

Instead of legal language alone, give examples:

  • Do not enter client names, contracts, employee records or unpublished financial data into unapproved tools.
  • Use approved enterprise tools when working with confidential information.
  • Remove identifying details before using a tool for drafting or analysis.
  • Ask before using customer or beneficiary data in a new way.

People follow rules more easily when they can recognise the situation in front of them.

4. Human review and accountability

AI can support work, but it should not remove responsibility.

Define who reviews outputs before they are used, especially for:

  • Client-facing communication
  • Public statements
  • Reports and data analysis
  • Recruitment or people-related decisions
  • Legal, financial or compliance-related work

A useful rule is simple: the person using AI remains responsible for the final output.

5. Training and shared language

Governance fails when people do not understand the tools they are using.

Training should cover:

  • What AI can do well
  • Where it makes mistakes
  • How to check outputs
  • How to protect sensitive information
  • When to ask for help
  • How to document AI use where needed

This does not require everyone to become technical. It requires everyone to develop good judgement.

6. A way to raise concerns

People need a safe, simple way to say: “I am not sure whether this is appropriate.”

Create a named contact, channel or process for questions. If raising a concern is difficult, people will make decisions alone, and that is when risk increases.

Make it usable

A governance approach sticks when it is easy to find, understand and apply.

Consider creating:

  • A one-page AI use guide
  • A short list of approved tools
  • Examples of safe and unsafe use
  • A checklist before publishing or sending AI-supported work
  • A short induction module for new team members
  • A regular review of tools, risks and lessons learned

The goal is not perfection. The goal is a system your team can follow on a busy Tuesday afternoon.

Start small

If you do not have governance yet, do not begin with a long policy.

Start with these questions:

  1. Which AI tools are we already using?
  2. What information are people putting into them?
  3. Which uses create the greatest risk?
  4. Which tools should be approved, limited or stopped?
  5. Who is accountable for decisions and outputs?
  6. What training does the team need?

Answering those questions gives you the basis for governance that reflects reality.

At Proof of Change, we help teams build AI governance that is practical, responsible and aligned with how they actually work.

Need AI rules your team will use, not just file away? Book a call.

Ready to talk about where you're headed?