Picture a 30-person company that decided this was the year to get serious about AI. Leadership bought enterprise licenses for ChatGPT, Claude, or Copilot and spent a few months rolling them out to every account. HR was asked to run the training. The usage report shows plenty of logins.
Look a little closer, and the picture gets murky. A few people use the tool to clean up emails, and others chat with it in conversations that start from scratch every time. Meanwhile, some employees are using personal AI accounts on the side, and leadership has no idea. Nobody can say whether any of it has saved the company time or improved anyone’s work.
Lesley Bodine spends her days on that gap. Her LinkedIn summary puts the problem in one line: “AI can be implemented quickly, and adoption takes structure.” She has spent more than 20 years helping teams change how they work, and her focus now is what she calls AI change management, mostly with small and mid-sized, founder-led companies. When we talked on Transform Your Workplace, I asked her about a rollout just like this one. She called it a “loaded question,” then pulled it apart.
AI implementation and AI adoption are two different jobs.
Lesley starts with definitions, because most companies use the two words as if they mean the same thing. Implementation is the tool. It covers what you bought and who has access to it. It’s also a moving target. She told me the setup she uses today looks completely different from the one she had three months ago, because so many programs keep folding more features into a single tool.
Adoption is harder to see. “Adoption to me is when we can see the organization, the employees, are using it and they’re getting measurable benefits,” she said.
That definition rules out most of what companies report on. A login count tells you people opened the tool. It doesn’t tell you whether anything improved. That’s one of the problems with a tech-first approach to AI: implementation can look successful even when the organization isn’t seeing meaningful results.
Traditional change management was built around a finish line: the day every user is on the new system. AI doesn’t give you one. The tools keep changing, so the culture has to keep adapting with them, which makes intentional AI adoption just as important as implementation, and Lesley argues that calls for a different kind of change management altogether.
The research points the same way. A July 2025 report from MIT’s Project NANDA found that 95 percent of organizations were seeing no business return on their generative AI spending. Some researchers think that number is overcooked, and it’s worth holding loosely. Lesley wasn’t attached to it either. What she trusts is that leaders are seeing the gap inside their own companies.
She has a theory about where it comes from. She recently talked with someone who builds AI tools for companies. His tool worked. The ROI was getting stuck on the people who resisted using it, and that was costing him sales. Most implementers will tell you they already handle training and adoption. In her view, that disconnect is where a number like 95 percent comes from.
Why employees resist AI tools at work
When I asked whether the people who won’t touch the tools are really resisting, Lesley didn’t hesitate. It’s identity, she said. Leaders tend to describe AI change as workflow or role redesign, “but really what it is is identity redesign”. People protect their identity and their role, especially when AI is changing how people think about their roles and what makes their work valuable.
She learned this in a previous role, on the executive team of a small company. Their approach to change was top-down: this is what we’re doing; everyone get on board. It missed the human side. Looking back, she says they needed to go a layer deeper than behavior and skills, into the environment and beliefs people were working inside. Shift those, and the behavior change follows. That’s where training and tools finally do their part. They implemented the change. Adoption, as she put it, wasn’t so great.
She lived it from the other side too. As that change rolled through, she looked at her own position and thought, “I think I just worked myself out of a job.” She could handle change. What she wanted was communication and clarity, including clarity about how her role would be redesigned. That missing clarity, she says, creates much of the discomfort at the bottom layers of a company.
She also knows what it feels like when the communication never comes. She once found out she was losing her job in a 9:30 a.m. phone call. She had just come back from vacation, where she’d worked through the weekend because it was month-end. Looking back, she realized everyone else knew the call was coming. She understands the numbers were down and business is business. But the company had always felt like family; the call was quick and factual, and “it just felt disposable.”
I think a lot of employees are bracing for that same feeling with AI. Lesley’s fix starts with telling people the truth about where you are. Most small companies she talks with are nowhere near AI-native. Many are still using it to organize email, which is miles from what people fear. A leader who explains where the company is with AI today, and where it’s likely headed over the next couple of years, removes a lot of that fear with the conversation alone. Nobody has a full map anyway. “Everyone’s behind,” she told me. “As of 9 a.m., you and I are behind.” She believes small, founder-led companies that handle this well will probably hire people before they lay people off.
Shadow AI and the missing guardrails
The same MIT report found that only 40 percent of companies had bought an official LLM subscription, while workers at more than 90 percent of the companies surveyed were using personal AI tools for work. Much of that happens without IT knowing or approving it. Lesley files this under AI governance and ethics, which makes it a CEO conversation before it’s an IT one.
She sees a quieter version of the problem in companies that did buy the license. When a small business tells her it has AI, she asks what that means. Usually it means Enterprise ChatGPT, Claude, or Copilot, with people having back-and-forth conversations in it. Then she asks what guardrails they’ve given the model. Is it connected to a folder that identifies you and the company? Most of the time she gets a blank look.
In her view, bolting a language model onto people’s work with no guardrails won’t produce a return. The MIT researchers landed somewhere similar. They traced the gap to learning, finding that most generative AI systems don’t retain feedback or adapt to context. In a small company, Lesley says, one wiring fix like that can move mountains. In an enterprise, the same fix takes far longer.
Who should own AI adoption in a small business?
Lesley admits she hasn’t fully pinned this one down, and it depends on company size. What she sees most often is adoption landing on HR’s plate as a training assignment. HR teams are skilled, she said, but “adoption is different than training”. HR gets asked to handle AI training and produce results, then runs into resistance that training was never designed to fix. She sees that as setting them up to fail.
The job market is just as muddled. When she started reading job postings, she found the same work called AI enablement in some places and AI implementation and adoption in others. If you’re hiring for it, define the terms before you write the posting.
She also described my own situation a little too accurately. The person running implementation is usually capable of adoption work too, she said. Their day just gets eaten by implementation.
When you spend the budget on implementation and treat adoption as a side project, you get side results.
I’ve been there. Lesley’s argument is that implementation and adoption are two jobs that work in tandem, with HR and learning and development alongside them.
How to start AI adoption at a 30-person company
I asked Lesley what a leader at a 30-person company should do in the next week or two. She’d start by mapping out where the company wants to go. Then pick one department and systematically build a skill for it, a repeatable set of instructions that automates work the team does over and over. Then sit down with the people in that department and talk through how the change affects their work. She’d reframe it with them right away and keep the conversation optimistic, focusing on how their role improves.
That first department is your test case. When the person using it gets excited, they become your champion, and you move to the next department. She compares it to incubating something away from the core of the business so the company’s antibodies don’t attack it before it proves itself. Once you get to teams of teams, she’d bring in an adoption specialist. For a company of five or ten people, she’d skip the formal readiness assessment altogether.
Under all of it is a change in how you think about roles. “AI is about outcomes,” she said, while most company structures are built around tasks. She used marketing as the example. Picture walking in on Monday to find your market research done and your outreach drafts queued up, with you deciding what gets sent. Once you trust the output, you let it send, and your morning goes to the people who wrote back.
I’ve seen this in my own work. I built a skill for this podcast. After an interview, I drop in the transcript, and it drafts show notes and a companion article in the same voice and format every time. That used to take me four or five hours. Now it takes minutes. The work still matters, but it was never the best use of my time. The hours went into more interviews and more important conversations, and it worked because I started with the outcome I wanted.
Lesley’s first step for any leader is a gut check. On a scale of one to ten, where do you sit between excitement and resistance? Then look at where the company really is, decide the outcome you want to reach, and start with clarity and alignment. The approach looks different for a solopreneur, a company of 5 to 50, and anything bigger. The principle holds at every size, she said, because everyone has an identity tied to a role. For the 30-person company with every license in place and no clear return, that means going back to the people, one department at a time, and having the conversation the rollout skipped.
—
Brandon Laws is the host of Transform Your Workplace and the VP of Marketing and Product at Xenium HR, an outsourced HR and payroll partner for small and mid-sized businesses in Portland, Oregon.