Your Governance Gap Is a Culture Metric

The most useful indicator of whether AI is changing your culture is who's hiding their usage. That's one of the more concrete intangible AI benefits to track: whether people feel safe surfacing what they're actually doing, and it's more concrete than output quality or hours saved.

We spent week six on governance, but the real thread running through the class was behavioral. Someone in the cohort shared that staff had been quietly using AI for external communications without flagging it to anyone. When they dug in, the people who'd gone underground with the tools were the ones who felt deficient in certain skills. They weren't being reckless. They were compensating. That's a culture signal. That's data.

Shadow AI Tells You Something Before Governance Does

Shadow AI (tools people are using that leadership doesn't know about) typically gets treated as a compliance problem. It is. But before you send a policy memo, sit with what it's telling you. If people are using AI without flagging it, a few things are probably true at once: the tools feel useful enough to bother with, the formal channels feel slow or punishing, and there's some unmet need underneath the behavior.

The simplest version of measuring this: ask people what they're actually using. Not as an audit. Canva, Zoom's auto-summarization, Microsoft Copilot, the email connectors inside ChatGPT that hook into Gmail and Google Calendar. None of these register in most people's heads as "AI tools," but they're all running on models.

Ask five people in different functions what they'd do if they wanted to try AI for something new. If you get five different answers, or if several of them say they'd just quietly try it, you've found your baseline. That inconsistency is the culture metric.

Governance Response Time Is an Innovation Metric

The tension most organizations hit: governance reads as restriction. The analogy that landed in class was brakes on a car. Brakes aren't there to slow you down. They're what let you go fast and stay in control. Without them you're not free, you're dangerous.

When you have clear tiers (approved tools that need no conversation, uses that warrant a manager check-in, high-risk decisions that go to leadership) people move faster. The uncertainty disappears. Track how long it takes someone to get a yes or a no on a new tool request. If no one knows whose call it is, that latency is measurable. It shows up as inertia and missed experiments. Shorten it, and you'll see more initiative.

Rebecca mentioned before class that engineers at a company she knows aren't engaging with AI at all. Not from fear. Just because no one has made it top of mind, and there's no clear AI-first stance from leadership. That's a different problem from paralysis, but it shows up the same way: nothing being tried, nothing to measure.

Coaching Conversations Are Your Soft ROI Receipts

Smaller organizations have an advantage here that larger ones genuinely don't. When someone uses AI in a way that's off, you can have a direct one-on-one conversation. Ask what brought them to the tool. Work through whether the output actually served the organization. That exchange is where culture change happens, more than any policy document ever will.

Track those conversations, even informally. A running note in a shared doc is enough. Zero conversations about AI use probably means one of two things: no one's using it, or no one feels safe surfacing it. Either reading matters. Both are actionable. And the pattern those conversations reveal over time is one of the clearer intangible AI benefits you can actually point to: a shift in how people relate to the tools, not just whether they use them.

Ross described a manufacturing environment where the default stance is: if you think a new tool might help, try it and report back. That low-friction permission structure doesn't require a formal policy. It requires a leadership position that's been communicated consistently.

The One Question That Surfaces Your Actual Baseline

Ask five or six people this: "If you wanted to use an AI tool for something you've never used it for before, what would you do?"

Varied answers, or answers that involve going quiet and figuring it out alone, tell you exactly where the governance gap is. That gap is your culture baseline. What you do with it determines whether AI governance enables your people or just documents your anxiety about them.

Documenting it doesn't have to be a long project. One to three pages. What's approved, what needs a conversation, who owns the high-risk calls. Over a thousand AI bills are pending across US states and only 68 have been enacted so far. The EU AI Act hits full compliance requirements in 2027. Organizations with something written down now, even something that's only directionally correct, will adjust faster when requirements land. Iterating on a draft is a lot easier than starting from scratch under pressure.

You don't have to be right about every governance decision. You just can't be wrong in a way that compounds.