Thou shalt not… likely read or remember this AI usage policy

There’s only one company policy that’s ever stuck in my head. It was from a tech company and it was to do with expenses. The policy was…

“Use your common sense.” 

Brilliant. Brave. And turns out highly effective. After the introduction of the policy, the average expense claim came down in dollars spent. 

That kind of policy is obviously not for every, or even most organisations. Too much grey and not enough down in black and white. 

But there are a few things we can take from that policy and the way it's written. 

1. It’s short.

2. It uses everyday language.

3. It’s easy to remember.

4. It puts the emphasis on the reader/employee to engage their brains and think before they act.

Expenses are one thing, AI usage policies are the real risky business

A recent article from the Forbes Communications Council predicts that clear AI usage policies will become increasingly important in 2026.

“Adopting comprehensive AI policies is a step in the right direction, but you can’t assume employees will read those policies, much less remember them.” 

Absolutely agree.

A policy that exists but isn’t read, understood, or recalled at the moment of decision isn’t a safeguard. It’s paperwork.

So, let’s take the spirit of that expenses policy and look at some practical ways to design AI usage policies that people will read and remember.

Getting people to read an AI usage policy

1. Structure and formatting are your friends here

Dense text tells people “this is going to take some effort”. 

People tell themselves “I’ll come back to this later”.

Later never comes. 

So...

• Embrace white space: give the text room to breathe. It’s not for decoration — it’s cognitive relief.

• Use summary subheadings: people will skim read the policy, so make the main points stand out, by sticking them in a subheading.

• Use bullet points as nature intended: a series of individual points. Not a big paragraphs with little dots by them. Those are sections that need subheadings. 

If you want people to take something away, make it impossible for them not to.

2. Use everyday language (and mean it)

If people have to decode the language, they’ll stop reading, or worse, they’ll misunderstand what you actually want them to do.

So: 

• Avoid business-speak and technical language: it only serves to slow people down and obscures meaning.

• Use metaphors and similes to make ideas land: abstract rules are forgettable. Concrete images stick. 

  
  

  For example: 

  Putting proprietary information into an unsanctioned AI tool is like giving your grandma’s secret sauce recipe to Heinz — with an invitation to use it. 

  
  

  You don’t need to love the metaphor. You just need to remember it.

  
  

• Use the active voice: the passive voice only gives people wiggle room. 

  
  

  You can’t upload client documentation to our AI tool. 

  That’s active. That’s clear. 

  
  

  Client documentation must not be uploaded to our AI tool. 

  That’s passive, and vague about who’s responsible.

  
  

Policies are behavioural tools. Write them in a way that plays into how humans behave. We all like clear, clean, easy to understand writing, because it’s low effort. 

3. Include a cheat sheet

Create a one-page summary or visual cheat sheet that captures the non-negotiables.

To make it work:

• Lean heavily on verbs so people are clear on what they should do (“Avoid…” “Check before…”, “Ask if you’re unsure…”)

• Focus on the highest-risk behaviours only, leave the rest for the body of the policy.

• Give people real scenarios so they recognise what to do in a certain situation. 

  You can try an ‘If this, then that’ structure. E.g. If X happens, do Y. 

Giving people a quick reference tool means they’re more likely to quickly reference it, funnily enough. 

Getting people to remember an AI usage policy

1. Tell them why it matters

Abstract principles fade quickly.

What sticks is:

• what’s at stake

• what could go wrong

• why the rule exists in the first place.

A short story will do more heavy lifting than a perfectly worded clause.

For example:

Deloitte had to refund part of their $440,000 consultancy fee to the Australian Government after submitting a report with AI-generated inaccuracies in it. We don’t want to do a Deloitte.  

People don’t need to memorise the policy. They need to remember why it matters.

2. Tie rules to real moments of decision

People will only be thinking about their company’s AI policy when they’re: 

• about to paste something into an AI tool

• under time pressure

• trying to be helpful or efficient.

So write for those moments.

Instead of:

Employees must not upload confidential information…

Try:

If you’re about to paste client data into an AI tool — stop.

That’s the kind of writing that'll be remembered in a risky moment. 

3. Make recall easy, not perfect

Employees only really need to remember:

• 3–5 default behaviours

• what to do when they’re unsure

• where to check before acting.

If someone remembers nothing else, they should still know how not to cause harm.

Apply this stuff to any and all policies

Policies don’t fail because people don’t care. They fail because they’re written in a way that’s easily forgettable, infinitely putdownable, exceptionally ignorable. 

Thankfully, that’s all very fixable.