Moving From Adoption to Adaptation for Explainable AI

Simon Sinek, motivational speaker on business leadership and book author, tells a story about a Millennial training program at a company with 100,000 employees (simonsinek.com/stories/why-your-best-ideas-might-needa-waiting-list). The company hired him to build the program. He said fine, but we’re doing it my way.

He taught one class. A hundred seats. People had to apply, write essays, fly to New York on their own time with no extra pay and no promise it would help them get promoted. He barred anyone born before 1984 from the room—every senior leader was locked out. At the end, he asked for volunteers to build the rest of the program. No budget, no title bump, just this question: Do you think this matters? Fifty people raised their hands.

Two weeks later, the corporate sponsor called him, furious. Managers across the country were calling in, demanding to know why their people weren’t invited. Sinek leaned in and told him that’s what demand is.

No PowerPoints. No marketing campaign. No executive mandate. The early adopters went back to their teams and said, “This is something,” and the system tipped.

DATA GOVERNANCE PROGRAMS

It was 2018. IBM had decided I was going to be the chief data officer for the $34 billion managed services division and that my job was to create the standard for data governance. The way it was presented to me—the way these things always get presented—was as a compliance exercise. Build the framework. Get the sign-offs. Roll it out. Make people adopt it. Nope.

I’d spent enough years watching governance programs die on the vine to know what would happen. You can build the most elegant data governance framework in existence, laminate it, distribute it to every desk in the building, and absolutely nothing changes. People don’t resist governance because they’re lazy or ignorant. They resist it because nobody ever tells them why it matters to them. The framework shows up as another mandate from another executive with another acronym, and it goes in the same mental drawer as last quarter’s compliance training.

So instead of building a standard, I started a movement. The question I asked wasn’t, “How do we get people to comply with data governance?” It was, “Why would any human being in this building care about whether AI is trustworthy?” Different question. Different answers. Different everything.

I found the early adopters—the people who already felt something was off, who were uneasy about the black boxes shipping under their names, who lay awake wondering what happened when the model was wrong and nobody could explain why. They weren’t hard to find. They were relieved someone was asking. I didn’t have to convince them. I just had to name what they already felt.

We started by inviting folks to a group, had speakers come from outside of IBM, had my friends come give their perspectives, and invited everyone to tell their story. We established an active Slack channel with massive amounts of moderation to teach manners and that reciprocity is the only metric that matters. Conversations, not committees.

We used every resource IBM had in its disposal—the IBV (Institute for Business Value) to create studies and do qualitative work—and get folks together from engineering and design. Collaborate with those who were already doing the work and just giving them a platform. There were working sessions and hundreds of conversations where people could articulate what worried them—not in the sanitized language of risk frameworks but in the plain language of, “I don’t trust this thing, and I can’t tell you why.” We gave the discomfort a vocabulary. And once people had language for the problem, they couldn’t unsee it.

That’s how IBM ended up with the largest trustworthy AI Center of Excellence (COE) anywhere. Not because someone mandated it from the top. Because we activated the humans around why it matters—and they carried it into every room we couldn’t reach. My famous quote that I made up at the time was, “You have a team of 350K humans in 175 countries, what is it that you think you can’t do again?”

ADOPTION VS. ADAPTATION

There’s a distinction that Sinek circles without quite naming, and it’s the one I’ve been turning over for months: the difference between adoption and adaptation.

Adoption is a transaction. Someone builds a thing, packages it, and works to get other people to use it. The product is fixed; the humans are the variable you’re trying to move.

This is the logic underneath every enterprise rollout, every change management playbook, every adoption curve. Get the humans to accept the artifact.

Adaptation is biological. It’s what organisms do when their environment shifts—not because someone made a compelling slide deck, but because survival demands it.

Adaptation happens when the organism encounters a genuine pressure and reorganizes around it. The change isn’t imposed. It emerges from the interaction between the organism and its context.

Sinek’s Millennial program worked because he didn’t try to get 100,000 people to adopt a training initiative. He created conditions where a hundred people adapted—recognized something real, reorganized their priorities around it, and then became the pressure that made adaptation spread. The sponsor’s angry phone call wasn’t a problem. It was proof the environment had shifted.

My IBM story follows the same pattern. I wasn’t trying to get thousands of employees to adopt a data governance framework. I was trying to find the people who had already adapted internally—who already sensed that untraceable AI was a problem—and give them the vocabulary and permission to act on it. The COE wasn’t adopted. It grew the way things do when you plant in fertile ground instead of paving over it.

This is what Geoffrey Moore’s chasm framework misses when you apply it to movements rather than products. Moore describes a gap between early adopters and the mainstream majority, and the conventional wisdom is that you cross it by packaging your innovation into a “whole product” that reduces risk for pragmatists. That works when you’re selling a widget.

But when you’re trying to change how people think about trust, accountability, and the role of machines in human decisions, the packaging metaphor collapses. You don’t shrink-wrap a shift in consciousness.

What actually crosses that gap is what Sinek describes and what I lived at IBM: You don’t cross it at all. You create enough pressure on one side, so the environment itself shifts. The early adopters don’t convince the majority. They change the conditions the majority operates in. The majority doesn’t adopt; they adapt—because the world around them has changed enough that the old way stops working.