Practice Beats Paperwork

I had to tell my friend and client Krysta this week that the policy she wanted to publish doesn't exist...and then tell her she was already ahead of companies ten times her size.

This exchange took place after the second session of my AI Fluency for a 100% Human Workforce program for her BuildingPPL team. Specifically, I had just sent over the Gut-Check Card from Tuesday's session and 90 minutes later, Krysta was so jazzed that she wanted to circulate her "policy" to her clients and social media.

She should have been jazzed: her team had high-level alignment on why AI governance mattered, a strong statement on how their organizational values guide their AI use, and a pragmatic checklist of how to check that AI-generated work adhered to those organizational values.

Not really a policy, per se, but way better.

Compare that with how most organizations approach AI governance: paper tigers. A document that satisfies Legal, satisfies a board, and has a Notion page with a nice header. A bloodless set of abstract risk mitigation "Thou Shalt Nots".

It just can't bite.

It can't change what someone actually does on a Tuesday. It can't tell a manager how to identify whether their employee did an AI no-no. It can't help an AI luddite identify if AI output is 100% true, or interesting, or useful, or relevant.

A policy written before you know what your people already do is governing a workforce that doesn't exist.

Organizations who roll out policies (and kudos to them for trying, it's better than "Here's Claude, let it rip") tend to fail to reckon with three AI realities:

1. Your people are already using AI under the table. See also: Shadow AI.
2. Your people are being told loudly by their leaders to "use more AI" without a "and here's how..."
3. Your people are being force-fed AI in the tools they already use (I'm looking at you, Gemini and Copilot).

And they already have opinions — the scared one, the evangelist, the quiet one who thinks it's a fad. A top-down rulebook doesn't land on a blank slate. It lands on habits and beliefs that are already there. Best case it's ignored; worst case it drives the behavior further underground.

The paper tiger concept really works here. Back in 1946, Mao Zedong popularized the phrase- already 600 years old- to wave off the atom bomb as a paper tiger: fearsome on the surface, but no match for the will of the people underneath. Three cheers for Communism.

The threat is the key: a paper tiger isn't a small tiger or a weak one. It's tiger-shaped paper. The silhouette of authority, none of the teeth.

This pretty much sums up many AI Acceptable Use Policies: looks real, looks scary, and doesn't say anything about using ChatGPT to generate a generic-sounding blog post. Too many AUPs say "Don't put Personally Identifiable Information (PII) into AI" but give no guidance on whether a networking event sign-in sheet is considered PII...and there's no way to tell that your marketing associate has been doing exactly that for the past three events.

But this is why I told Krysta that her team was ahead of the curve on this. Instead of drafting an unenforceable policy document, we took an alternative route:

1. Look first. I asked what her people are already doing and what already feels icky. Session 1 was mostly "what are you doing, what feels gross, and what feels right." You can't govern what you refuse to look at.
2. Draft a vision from what you saw. Coming out of Session 1, I took their answers and assembled them as a manifesto — values, not rules. What BuildingPPL stands for: to our clients, to our people, to each other.
3. Build the practice. The Gut-Check Card were simple quality check questions, generated by Krysta's team, that asks before they ship: Is this good enough to stand behind? Is this leaning on first-party vs third-party data? Does the rug fit the room?
4. Then, maybe, formalize. If Krysta wants to turn their Manifesto and Quality Assurance check into a formal AUP, she can because now the policy has teeth, built out of a practice that already lives in the room instead of declaring one that doesn't.

I don't mean to say that policies don't matter. A policy serves as critical training guardrails for new hires. It draws a bright line between Ok/Not Ok behaviors. It serves as a benchmark to be questioned in the future (which is actually Session 5 of the AI Fluency curriculum).

The nuance I want to really underline is the order of operations when creating AI governance. Observe, then articulate, then regulate.

Think you might have an AI governance paper tiger in your Employee Handbook? Here's a dead simple test: spend one meeting asking your team what they're already using AI for and what about it makes them uneasy. If their answers need to be turned sideways and overexplained to connect to your existing policy: meow.

PS. The course Krysta's team is taking, AI Fluency for a 100% Human Workforce, is 100% funded by the Massachusetts Workforce Development Express Grant. Reply back to this email if you want to see if you're eligible!

A TL;DR from the CRO

The 'no dogs on the bed' policy has been on the fridge for four years; I have never read it.

-Roman Noodles, Chief Ruff Officer

Show Me Something Cool With AI - Today!

Today, June 26th at 1pm ET, is our monthly show-and-tell meetup, Show Me Something Cool With AI. Each month has an optional challenge so everyone can come with something cool, and this month's challenge is to Give Something a Voice. Use AI to generate some kind of audio for a typically quiet thing: an AI-generated song about your credit card statement, an AI-generated voice for your mismatched sock, or an AI-generated podcast about your Tinder profile.

​Even CRO Roman Noodles is participating in this challenge.​

This week I travelled to New Jersey to watch my niece graduate high school. This is the same niece whose first day in 6th grade was profiled in one of my earliest blog posts for Learn to Scale, so yeah, not feeling old at all.

At her graduation dinner at the Cheesecake Factory, I asked her, "So what do you think about AI? Good? Bad? Lame? Ratchet?"

(I'm so hip with Gen Alpha lingo)

The most surprising thing she said was that she hates AI in Google search. "If I'm going to Google something, I want to Google something, not read what some AI told me. And the stuff they present is usually worse than what I can find through search."

I followed up with, "So what about college? Think you'll use it there?"

She gave me the narrowed eyes that only an 18-year old girl who worked at a lash salon could give. Of course she was going to use AI for her college homework.

How naive is her Uncle Danny?

Not very. I would assume that every college freshman is using AI to do some of their college work. I would!

But how naive are college administrators who think their AI policies are going to stop my niece?

meow.

This unc is low-key cringe but he's got aura for trying,

Dan from Learn to Scale​

​Opt-out from the newsletter | Unsubscribe from all emails | Update your Preferences | www.learntoscale.us, Boston, MA 02119 |

​PS. This paper tiger totally shreds.​