What is the AI Adoption Chasm?

When I walk into an organization, I start by getting a read of the room. How many people have been using AI? How many feel confident with it?

Then I ask individuals about their work. What's your role? What are you doing in your role right now?

The answers are usually pretty generic. They use it to rewrite emails. They use it to summarize documents. That's where the majority of people are.

Then we dig into their role, and I show them specific ways they could be using AI in the work they already do. That's when their minds are blown.

"I didn't even know we could do that." Some get excited right there. Others tell me, "I'm still skeptical, but I'm ready to take my first step."

What changed was what they saw: AI used on work like theirs.

That gap, between the few who already work differently with AI and everyone else, is what I call the AI Adoption Chasm.

What does the AI Adoption Chasm look like inside an organization?

Inside an organization, the AI Adoption Chasm looks like a small group changing how they work while everyone else stays where they were. Everyone has access. A few people rebuilt their work. The rest tried it on an email or two.

The numbers behind it come from Everett Rogers, whose book on how new ideas spread, Diffusion of Innovations, ran to five editions. He sorted people by how early they adopt: innovators (2.5%), early adopters (13.5%), early majority (34%), late majority (34%) and laggards (16%).

Innovators and early adopters add up to 16 percent. Round up, and you get the first 16 to 20 percent. The majority is everyone else, roughly the other 80 percent.

Rogers called his groups "ideal types," a way to make comparisons. Your organization won't split exactly this way. Treat the numbers as a rough map of where the stall tends to happen.

The national data shows the same shape. Gallup, the polling and workplace research firm, surveyed 22,573 US workers in May 2026.

Gallup counts anyone who uses AI at least a few times a year. Raise the bar to a few times a week, and the share drops to 30%. Raise it to daily, and it's 15%.

Even 30% is past the 16 to 20 percent line. So has the US already crossed?

By the measure that matters, no. The chasm is about changed work, and frequency can't show you that. Someone can open a chatbot every day to rewrite emails and still do the job exactly the way they did it last year.

That matches what I hear in the rooms. Plenty of use. Very little changed work.

Is the AI Adoption Chasm the same as pilot purgatory?

They're related, and they're different.

The term "pilot purgatory" comes from manufacturing. A January 2018 World Economic Forum paper written with McKinsey used it for factories stuck piloting Industry 4.0 technology (the push to connect and digitize the factory floor). McKinsey popularized it through 2018. It wasn't coined for AI.

A pilot can succeed and still leave the chasm untouched. The tool works, it rolls out to everyone, and the majority's work stays the same.

I'd argue pilot purgatory is worse than failure. A failed experiment teaches you something. A pilot that never scales and never dies just drains energy. The chasm is quieter than that. Nothing looks stuck, because everyone has a login.

Why doesn't the majority follow the AI enthusiasts?

Because most people take their cues from people like them.

Rogers found this decades ago. "Most individuals evaluate an innovation, not on the basis of scientific research by experts, but through the subjective evaluations of near-peers who have adopted the innovation." A near-peer is someone close to your own role and situation: same kind of job, same kind of pressure.

He also found that adoption usually "takes off" at about 10 to 20 percent, once people start hearing about an innovation from peers. That's the same stretch where the chasm sits. In his model, it's where things speed up.

So why does it stall inside so many organizations?

Here's my read. Inside an organization, the first AI users are often the tech-curious, and they aren't always the people their colleagues already look to. When the peer conversations never start, take-off never comes.

That's why a few star users can't carry the organization. A hero isn't a system. Your best AI user can show what's possible, and the people in other roles still have to see it working on their own kind of work.

Rogers drew one more line that matters here. Inside an organization, the decision to buy a tool and each person's decision to use it are separate. He wrote that "the individual office worker in the organization may have little or no say in the innovation-decision."

The license is the leader's decision. Changing how someone works is theirs. That's why buying licenses doesn't get you across.

Underneath the stall sit what I call The 5 Invisible Forces, the hidden barriers that hold an organization at the chasm. You have to name them to tame them.

Here's how three of them show up at the chasm. Status Quo Inertia is the pull of business as usual, and Gallup found preferring the current way of working is one of the most frequently cited reasons people give for not using AI where it's available. The Perfection Fallacy is waiting for the tool to be fully ready. Misaligned Incentives show up when the reward system pays for efficiency and gives nothing for experimenting.

Isn't AI just spreading faster than past technology?

Yes. Use is spreading faster than PCs or the internet did. Three economists writing for the National Bureau of Economic Research, Alexander Bick, Adam Blandin and David Deming, found generative AI adoption "faster than either PCs or the internet." I walked through that comparison in why new technology takes decades to pay off.

Faster use still doesn't get the majority across.

Boston Consulting Group shows where it stalls. In BCG's June 2025 survey of more than 10,600 leaders, managers and frontline employees in 11 countries and regions, more than three-quarters of leaders and managers said they use generative AI several times a week.

Regular use among frontline employees stalled at 51%. BCG calls it the "silicon ceiling." Frontline here means white-collar individual contributors, people with no one reporting to them.

So the people using AI most are the leaders and managers. The individual contributors, usually most of the headcount, use it less. That's the group that decides whether you cross.

Fast use and slow change can happen at the same time. In the post on why new technology takes decades to pay off, the profit numbers told the same story: people feel faster, and the business mostly looks the same.

What helps the majority cross?

The majority crosses when they see someone in a role like theirs get a real result on real work. I've watched it happen.

In one organization, an AI power user spent about two hours with two of their peers, walking them through how to build a financial model from scratch with an AI tool. By hand, that kind of model usually takes them a week to a week and a half. With AI, it came together in about 45 minutes.

It blew their peers' minds.

Look at what was in the room. A peer, two people who do the same kind of work, a piece of their own work, and a result they could measure: more than a week of work down to under an hour.

That's Rogers' near-peer finding, live. It's also the pattern I see most. The person who gets a team across is usually a power user spending time with their peers, showing them how they use AI in their own way.

Manager support carries the biggest number. In organizations that make AI available, Gallup found 78% of employees who strongly agree their manager actively supports AI use it frequently, vs. 44% of those who don't. Supportive managers may simply sit where AI already fits the work, so read it as a strong link. It doesn't prove cause.

This is how I build those conditions. The ADAPT Framework™ is the structure, and its Train pillar is hands-on, on real work, with managers trained too. The Culture Mortar holds it together with four parts: Safe to Try, Curiosity Is Rewarded, Time Is Protected and Leaders Go First.

Two of those line up with the list above. Leaders Go First is the manager who backs it. Time Is Protected is the time to learn.

Here's what that can look like in practice:

1. 01Pick one team and ask three questions. How many people used AI on real work this week? How many changed a step in how they do the job? Can anyone show a result?
2. 02Find the person on that team who already got a result, and give them time with peers who do the same kind of work. Two hours with two people was enough in the example above.
3. 03Ask managers for one standing slot where people show what they tried, including what didn't work, and recognize both.
4. 04Put time to learn on the calendar, so nobody has to steal it from their real job.

What this means for you

If a few of your people are doing amazing things with AI, that's a good start. Celebrate it.

Then remember what got their peers moving. Abstract strategy doesn't move behavior. Visible examples do.

The majority is waiting to see someone with their job get a result on their work. Your job as a leader is to make that happen on purpose, and to make room for people to try.

So who on your team already has the result, and who's seen it?

The majority crosses one peer at a time.

Key takeaways

1. 01The AI Adoption Chasm is the stall between an organization's first AI users, roughly the first 16 to 20 percent, and everyone else.
2. 02Access spreads fast. Changed work doesn't. Most people still use AI for generic tasks like rewriting emails and summarizing documents.
3. 03People judge a new tool by what peers in jobs like theirs got out of it. The majority crosses when someone in their role gets a real result on real work.
4. 04Pilots are about projects. The chasm is about people. They're related, and they're different.
5. 05Make the crossing happen on purpose. Put your power users in front of their peers, have managers back it, and protect time to learn.

Questions

Sources

Frameworks in this piece: The 5 Invisible Forces, The ADAPT Framework™, The Culture Mortar