Skip to main content

AI Policy vs AI Culture: Building AI Governance for UK Teams - Digital Compliance Academy

A policy document protects you from lawsuits. A culture protects you from irrelevance. Why most AI policies are useless, and how to build a culture of safe experimentation.

Jon McGreevy October 25, 2025 5 min read
Strategy Leadership Culture Management

Do you remember the corporate “Social Media Policies” of 2010?

They were usually written by lawyers who had never used Facebook. They said things like: “Employees must not use social networking sites during working hours.”

Fast forward to today. LinkedIn is the primary sales channel for B2B companies. Twitter (X) is where customer service happens. If you strictly followed that 2010 policy today, your business would be invisible.

We are seeing the exact same mistake happen with AI.

Companies are panic-writing 40-page PDFs called “Acceptable Use Policies.” They ban everything. They block URLs. They require three layers of sign-off to use a summarisation tool.

They are building a Policy. They are neglecting the Culture.

A Policy protects you from lawsuits. A Culture protects you from irrelevance. In 2026, irrelevance is the bigger risk.

The Flaw of “Compliance Theatre”

I call most corporate AI policies “Compliance Theatre.” They exist to make the Board feel safe, not to make the company safe.

  • The “Ban Everything” Fallacy: If you ban ChatGPT, your staff will just use it on their phones (Shadow AI).
  • The “Static Document” Problem: Policies are static. AI moves weekly. Your policy bans “uploading files to AI,” but doesn’t mention that Microsoft Copilot (which you just enabled) does exactly that automatically.
  • The “Department of No”: If the first response to every innovation is “No,” your best people—the ones who want to work smarter—will leave. They will go to a competitor who says “Yes.”

Defining “AI Culture”

Culture is not a poster on the wall. Culture is what happens when the manager leaves the room.

Does your team feel safe to say, “I used AI to do this”? Or do they hide it because they fear being lazily labelled?

  • Bad Culture: “I spent 4 hours writing this report.” (Lying, because they used Claude to do it in 20 minutes, but fear being punished for ‘cheating’).
  • Good Culture: “I used Claude to draft this report in 20 minutes, then spent 3 hours fact-checking and adding strategic insight.” (Honesty + Value Add).

Here is how you build that culture.

1. The “Demo Day” Ritual

You cannot mandate culture via email. You must build it through rituals.

Once a month, hold a strict 30-minute “AI Show & Tell”.

  • Rule 1: No Slide Decks. Live demos only.
  • Rule 2: Rank agnostic. The intern can present; the CEO can present. Use the hierarchy-flattening nature of AI.
  • Rule 3: Success AND Failure. Celebrate the prompts that failed. “I tried to get Gemini to write our newsletter and it sounded like a robot pirate.” This builds psychological safety.

Real World Example: A Junior HR Admin showed the Director how she used a “Chain of Thought” prompt in Claude to anonymise 50 CVs in 3 minutes—a task that used to take all Friday afternoon. The Director approved a Claude Team seat for her on the spot.

2. Reverse Mentoring

In the traditional world, experience equals wisdom. The Senior teaches the Junior. In the AI world, adaptability equals wisdom. Often, the 23-year-old grad is deeper into the prompt engineering rabbit hole than the 50-year-old partner.

Formalise this. Pair your ExCo members with “AI Buddies” from the junior ranks.

  • The Deal: Included in the “Buddy” role is a mandate to show the Senior one new tool or workflow every week.
  • The Benefit: The Senior gets upskilled without admitting ignorance in front of the board. The Junior gets face time with leadership.

3. The “Kill a Process” Bounty

Compliance culture adds process. AI culture should subtract it.

Offer a tangible reward (a £50 voucher, a half-day off, a bottle of wine) for anyone who can uses AI to kill a process.

  • “I used to manually copy data from email to Excel. I built a Zapier + OpenAI workflow to do it automatically. I have killed the ‘Data Entry’ process.”
  • “I used Gemini to summarise the weekly 2-hour status meeting into a succinct email. We have shortened the meeting to 30 minutes.”

Celebrate the destruction of drudgery.

4. Shared “Artifacts” as Knowledge

Don’t let knowledge live in private chats.

  • Claude Projects: Create a sophisticated “Project” in Claude with your brand tone of voice, your product manuals, and your style guide. Share it with the Marketing team. Now everyone writes with the same voice.
  • Gemini Gems: Create a specific “Gem” for “Objection Handling.” Feed it your best sales scripts. Share it with the Sales team.

Treat prompts as Company Assets, just like code or legal contracts.

5. The Golden Rule: “Human in the Loop”

The only non-negotiable “Policy” element in your Culture should be accountability.

Adopt the “Pilot in Command” mindset. When a pilot uses autopilot, they don’t take a nap. They monitor the instruments.

  • The Rule: “I don’t care if you used AI. I care if it’s right.”
  • The Consequence: If an employee sends out an AI hallucination to a client, do not blame the AI. Blame the employee for not checking. “The AI made me do it” is not a valid excuse in a grown-up business.

Summary: A “License to Experiment”

Give your people a license to experiment. Budget for failures. Buy the tools (see our Shadow AI post).

If you build a fence, your people will jump over it. If you build a playground, they will build castles.

Stop writing policies. Start building culture.