What Actually Gets Executives Using AI

Christopher Parsons, founder and CEO of Knowledge Architecture, wrote about something that happened to him at a conference called AEC Innovate, a gathering for architecture, engineering, and construction firms. He'd just finished walking the room through a handful of AI agents his beta clients were building: one that reviews contracts against a firm's own accumulated judgment, one that drafts fee proposals the way a firm actually prices work, one that helps write project descriptions the way the firm's best writers already write them.

A prospective client's CEO came up afterward, energized. He could picture a whole shelf of these things his firm might build.

Then the CEO's shoulders slumped as he thought through the question he was about to ask Christopher. Was he supposed to sit down alone and write all of that from scratch? Were his leaders supposed to somehow find the hours to document everything they know before any of this becomes real?

Parsons could have answered with a framework: change management principles, adoption curves, a slide on organizational readiness. That's the reflex, and it's usually where these conversations go and then just fizzle out, because a framework doesn't fix a slumped shoulder. It just gives the person a more sophisticated reason to stay overwhelmed.

Instead, Christopher told him about a CFO who was retiring.

Before the CFO left, the firm sat him down for a few short conversations about how he reviewed NDAs: which clauses were dealbreakers, which he'd negotiate, and why. Decades of judgment, pulled out through nothing more complicated than good questions.

The transcripts of those conversations became the training data for an agent that now reviews NDAs the way that CFO did, for a firm that no longer has him.

The CEO's reaction was immediate. Parsons quoted it in his writeup afterward: "It's so obvious now that you are saying it." (His full piece on what he calls Modern Knowledge Capture is worth reading in full.)

Explaining a capability doesn't make anyone want it. Recognizing yourself inside someone else's example does.

That's the part worth noticing because it contradicts how most of us try to bring leadership along on AI. The instinct is to explain: here's what the tools can do now, here's where the industry is heading, here's the framework for thinking about it. All true, all reasonable, and none of it would have moved that CEO an inch, because an explanation asks something of him that a story doesn't. An explanation asks him to trust a claim about a future he can't picture yet. A story hands him an instance he can already picture, sitting inside his own building. Chip and Dan Heath, who spent a career studying why some ideas take hold and others don't, put it plainly: skip the concrete case and you're "building a roof in the air." There's no foundation under the claim for anyone to stand on, no matter how well-reasoned it is.

What moved the CEO wasn't a better argument. It was a story with a name and a job title and a specific kind of document in it, close enough to his own firm that he could see himself standing inside it. He didn't need to understand AI. He needed to recognize his own firm in someone else's story, and see his own NDA agent, even though it hadn't been built yet.

I've talked before about the two threats that make AI adoption an identity problem rather than a training problem: the fear of returning to beginner status, and the fear of finding out you're no longer needed. The beginner threat concentrates hardest at the top of an org chart, among people whose authority rests on already knowing things, and it's usually what an explanation triggers without meaning to. Explain AI's capability to an executive in the abstract and you're implicitly asking them to imagine fumbling with something new in front of the people who report to them. A concrete example does something different. It hands them a template close enough to their own situation that a first real attempt feels like imitation rather than exposure.

That gap, between knowing AI exists and actually practicing with it, is where most of the value gets lost. Misha Sulpovar's read on enterprise AI adoption is blunt: "AI's ROI is not evenly distributed. It is lumpy, unpredictable, and unequally captured." A small number of people become real practitioners and get outsized results. Most people dabble once, see nothing, and quietly stop. Showing rather than explaining doesn't let an executive skip becoming a practitioner. It's what actually gets them to become one, by giving them a low-stakes way to start.

I've watched this play out directly. At an architecture firm I worked with, close to a third of the staff were waiting for direction from senior leadership before touching AI themselves. The problem wasn't that leadership opposed it. It was that their own usage amounted to fixing a typo here and there, nowhere near enough to set a real tone for the firm. The fix wasn't a better explanation of AI's potential delivered to that leadership team. It was getting them personally using the tools enough to be, in the phrase I used with them at the time, knowledgeable enough to be dangerous, so they had something real to point to when they told their own people to start.

Which brings me to a question I get quite often, in some version of: how do I get my leadership to actually engage with this stuff instead of just mandating it from a distance? The honest answer isn't a briefing. It's finding the equivalent of that CFO conversation inside your own company; the person whose judgment is genuinely valuable and genuinely at risk of walking out the door with them one day, and turning thirty minutes of their thinking into something a skeptical executive can see themselves using. You don't need executive buy-in to start that conversation. You need one expert, one recorder, and enough curiosity to ask the same question five different ways until the instinct becomes explicit.

This is a piece of a bigger argument I'll be making this Thursday for Sidebar's Speaker Series: Your executives don't understand AI. That's your problem to solve. It's not a fair ask, putting the job of educating leadership on the person who reports to them. But waiting for fluency to arrive from above is a bet against your own timeline, and the fix usually isn't a better explanation. It's a better story, built once, aimed at the exact problem the person above you is already wrestling with.

The talk is open to the public, noon PT this Thursday, September 10th, part of Sidebar's Speaker Series. You can sign up here.

I wanted two examples of this in hand before sending the newsletter, so I went looking this week for a second story to complement the CFO one, a story about someone with no formal authority who found their own leader's actual pain point and built one small thing that changed the relationship. Unfortunately, I didn't find a clean one. I share that because it demonstrates that doing this is a lot easier to describe from the outside than to actually do from the middle of an org chart. If you try it this week (and I think you should) and you find that it's harder than it sounded here, that's not a sign you're doing it wrong. Keep at it anyway.

So I ask you to think about it: what's the CFO conversation sitting inside your organization right now, waiting for someone to ask the right questions before it walks out the door?

Break a Pencil,

Michael

www.breakapencil.com