The Psychology of AI Adoption

One assumption skeptics or unskilled AI users often make is that AI replaces individual thought. Some view it as cheating, taking shortcuts, or a less valuable way to think than doing all the work from idea to execution.

Skilled users of AI understand that the expertise and intelligence of the user govern the outcomes. From this standpoint, AI is not a replacement for human genius; it amplifies it. However, if the people using AI fear judgment or misunderstanding from those who don’t, they become less likely to talk about it — which shuts down the social contagion that open conversation and modeling would otherwise create.

Application: If you are trying to grow the use of AI in your company, talk about the process of using AI, not just the end results. For example, one of my CEOs used AI to create traction on critical items that move the needle for the company. I saw the end result. When I asked about the process underlying it, it turned out to be an iterative, collaborative exchange between the key stakeholders and AI. Human judgment and creativity drove the process; AI helped execute. Even as a CEO, he worried whether people would view the outcomes as “his work.” So I suggested he tell his senior team about his process, so everyone felt more comfortable talking about AI as part of their results.

“How interested are you in learning AI to help with any of the tasks above?”

What actually played out was bigger than interest in artificial intelligence. It was the degree to which someone was tied to old and familiar systems that offered the perception of efficiency and excellence. It tied to whether candidates preferred predictable sequences over creative exploration. To be clear, I love things that are predictable — but not at the expense of growth and excellence. I saw differences in critical thought: the willingness to personally research and explore rather than relying on friends, family, and headlines to make decisions.

When I asked the candidates that question — “How interested are you in learning AI to help with any of the tasks above?” — I was assessing a construct called Openness, one of the most well-researched traits in personality science. Openness is one of the “Big Five” traits in the Five-Factor Model of Personality. The Big Five model meets the standards of rigorous science that psychologists use to separate legitimate constructs from the pseudoscience of most commercial personality tests. Its five core traits are openness, conscientiousness, extraversion, agreeableness, and neuroticism (OCEAN).

Openness refers to each individual’s interest in engaging in new things and their ability to adapt. It applies not only to tech but to the willingness to pursue or adapt to any change. Openness is substantially heritable and relatively stable, which is part of why it predicts where someone sits on the adoption curve — but it is not destiny. Like all traits, it operates on a continuum rather than in categories, and situational variables can raise or lower its expression. People high on Openness tend to try new things. They are not necessarily threatened by change, and if they are, they tend to be more curious than scared, more optimistic about possibilities than fearful of risks. Other variables in the GIST framework will increase or decrease the expression of Openness. For example, social psychology predicts that our social environment influences adoption, so even people who are low in Openness will be more likely to adopt AI if their friends are using it.

The other Big Five trait that can influence adoption is neuroticism. Neuroticism is the dimension of emotional stability: high neuroticism means a greater tendency toward negative emotion and a heightened vulnerability to anxiety and stress. People who have read negative predictions about AI, or who feel high anxiety about unknown or unpredictable outcomes, are less likely to experiment with it. At the same time, people high in Openness may feel stressed about AI but adopt anyway, because they are curious about its impact and the learning it affords. Thus, neuroticism has the most adverse impact when paired with low openness. One role where these traits most directly affect organizational adoption is that of the organization’s tech experts: whether they are open to new learning while also addressing their concerns about security and the new threats AI may bring.

In real time, I have watched people who are retired — or have the option of retiring — say, “I don’t have to think about that.” Someone who finds change exhausting, dislikes technology, or fears the changes AI will bring needs external motivation to adopt it. Without a strong reason, they are unlikely to endure the discomfort. People high in openness who love learning are more likely to adopt, regardless of their work status. Those who like things the way they’ve always been, and who aren’t directly affected, will likely chalk AI up to a new thing from a different generation and file it under “not me.”

IMPORTANT: One blind spot I see here involves senior leaders who may personally retire before AI fully changes the landscape. Because of where they sit, they risk underestimating both the impact on the company and the secondary and tertiary effects on talent attraction and retention.

Bluntly, people don’t like to be told what to do. We all want a sense of control in our decision-making, but some people are more sensitive than others to actions that can feel controlling or manipulative. I am one of those people — which means I have high psychological reactance.

Application Scenario for High Achievers: For one year, I talked non-stop about AI, on purpose, with all of my clients. I felt I was doing a disservice to them and their organizations if I didn’t lead and assist by example. After that, I shut up. A curvilinear relationship exists between exposure to an option and follow-through — and it is easy to miss the turn. Exposure helps, up to a point: initial offerings are often received with openness, and repeated exposure builds familiarity, which tends to breed positivity (the mere-exposure effect, Zajonc). But the curve reverses. Once someone who isn’t psychologically or behaviorally oriented toward adoption has had enough exposure, repetition stops building familiarity and starts generating aversion. Think about anyone you know who repeatedly gets on a soapbox about an issue that is irrelevant or annoying to you.

Application in Work Settings: You may need to tell employees that AI learning is mandatory. That’s fine. What will help adoption, though, is to give choice wherever you can. We all understand that we don’t have complete freedom of choice over everything, and employees as a group tend to respect clear rules. There is a very cool option that works for both 5-year-olds and 50-year-olds: provide choices about the process instead of the outcome.

For example, if I say, “Would you like to join an AI learning group, or would you prefer a self-paced training program?” I haven’t given you the option of whether to become more familiar with AI. As soon as I ask that question, it engages your brain to think about the options — the how of learning — rather than whether you want to learn at all.

Even those of us who learn through listening, reading, or vicarious witnessing encode information differently when we manipulate it (Kolb, 1984). At a neurological level, we naturally elaborate the information, which is part of deep learning. When we work with information, it encodes into memory — which is why many people say they need hands-on learning to remember new processes (Craik & Lockhart, levels of processing). Further, when we experience something, we are more likely to make it personally relevant, which attaches it to motivation, emotion, and reward. Together, the learning moves from something we understand in the abstract to a concrete thing that becomes real.

Application of Experiential Learning to Clients and Organizations

1. Business Client. One of my clients was open but skeptical. I heard a clear way of using AI to elevate an existing process that required integrating the multiple factors and contingencies common to manufacturing. Instead of talking about it, I asked him to describe what he was trying to do. I opened my computer and started typing, which I never, ever do. I often research things in real time, so he probably thought I was trying to figure something out. Instead of typing into a document, I typed directly into the AI platform that best fit the query. This was an in-person meeting, so I added some questions, submitted the prompt, moved beside him, and showed him the answer. All of my clients are fast learners, and he was off and running… it was the initial exposure to tangible ROI that made the difference.

2. Business Client. One of my clients wanted to learn about AI by reading about it. That’s impossible. For people who love to learn, reading is safe and fun but mostly useless here, because part of learning AI comes directly from using AI. So over a few meetings, I had him sign up, sign in, and start doing research with me in real time—the kind I often do behind the scenes. He is now not only using it but integrating it into his business strategy as an essential component of development with his team, rather than as an add-on.

3. Organizational Client. Wherever possible, I operate on a “show vs. tell” model of facilitating change. One of the organizations I assist struggles with alignment on general tech practices and AI. In this case, I’m not trying to change their DNA, but I am concerned about their missed opportunities for efficiency. Instead of educating, debating, or converting, I simply bring my own tools to the game. The introduction to new tech or AI usage happens through interacting with me, rather than as a separate body of work.

Regardless of openness and interest, energy is finite. Many people know physical energy is finite, but what I call “emotional energy” is, too. Emotional energy fuels drive, ambition, frustration tolerance, and perseverance. It intersects with the other individual variables. Learning takes energy, even for those of us who are refueled by it. For a High Achiever (or anyone else) already maxed out navigating personal and professional crises, asking for a new area of learning — regardless of its value — is simply too much.

Think about how everything you know about sleep, nutrition, and exercise dissolves into the habits of a 5-year-old in times of stress. Without the buffer to implement, witnessing and experiencing have limited utility. When individuals and teams have competing demands and high stress, problem-solving the bandwidth constraint is as important as every other variable in the GIST framework.

I’ll start with a non-textbook example of social psychology, drawn from years of trying to “help” in different ways. While it may sound egotistical, it is hard for me to watch people I care about carry unnecessary risk, stress, or burdens. When I first started investing time into health as a lifestyle, I talked about it. At some point, I’m sure I got tired of hearing myself, so I stopped and simply continued doing my thing. I remember two friends starting to change their behaviors. When I asked one what she’d read or seen that clicked, her response was “watching you.”

Watching other people and adopting their behavior results from social learning (observational learning) and social contagion. The research on these concepts traces back to Albert Bandura, whose work taught us that we do not have to directly experience reward and punishment for learning to occur. We can learn vicariously.  Bandura’s work also taught us that when we witness people behaving a certain way (for better or worse), those behaviors provide an option for how we might behave ourselves, if we are so inclined.

Success Stories on How My Clients and I Have Influenced Each Other

1. J Genius — J Genius was ahead of me on the curve of AI adoption (still is). We aligned on openness, but he leaned in harder on time investment and application. He shared information about platforms, and I followed his lead. Our businesses differ and have different use cases, which allows for multiple learning angles that cross-pollinate.

2. B Genius — B Genius’s amplification of technology and his application of artificial intelligence are breathtaking. I moved through the existential angst before he did, so I helped him through it. He has built more complex AI tools than I have, which inspires me to up my game. Our relationship and connection on AI also helps me discuss AI adoption with his senior leaders.

“I feel more peer pressure to use AI than I did to use drugs in high school.”

–LinkedIn post

As social creatures, one of the most common heuristics we use to gauge our own value and status is comparing ourselves to other people. When we compare ourselves to people we perceive as better than us (upward social comparison), we feel worse about ourselves. When we compare ourselves to people we perceive as worse off (downward social comparison), we feel better. While this quirk of the psyche is a horrible strategy for life adjustment, it exists and it shapes our behavior. We have Leon Festinger (1954) to thank for the research on social comparison: in the absence of objective standards, we look around us for ways to evaluate ourselves. Listen to conversations about AI (or weight, or money) and you will instantly hear people gathering and assessing information to understand how they stack up relative to others.

IMPORTANT: Social comparison has the most power when we compare ourselves to those we perceive as similar to us.

Application: I’ll use myself as the example. When I compare my own usage and knowledge of AI to people I respect, I’m motivated to be competent and to keep up or surpass them. I stretch. But when I’m talking with someone whose use of AI is at a lower level, it’s easier to feel like the expert — and additional adoption on my part doesn’t improve my status.

Workplace Application: Publicly praise the colleagues or employees who are advancing in their AI dexterity. It sends a clear message about what we value and what we’ll praise, and it gives others a tangible example to compare themselves against.

I need to speak to this personally first, because it has impacted, and will continue to impact, the breadth, speed, and depth of my own AI adoption.
On a personal level, I see confidentiality as a measure of who to trust and whether I can be trusted. That assessment began in 7th grade. Years of extensive training in ethics, HIPAA, board complaints, and litigation have sharpened, honed, and pounded it home. At an extremely personal and professional level, people trust me with information that is sacred. Yes, sometimes it is intellectual property or sensitive data, but it can also be deeply personal.

So the amount of time I’ve spent assessing the security of AI usage has, on some days, rivaled my AI usage itself. I then bounced my knowledge off security experts. Those conversations made me more confident in my security knowledge, but they did little to inform or assuage my concerns about AI usage specifically.

Application: Knowledge is always helpful, but in this case it is imperative. Whether you are making the judgment or trying to reassure colleagues or employees with security concerns, knowledge lets us make informed trade-offs instead of avoiding adoption or blindly hoping for the best.

At a tech conference, I heard a recurring theme: AI both poses a security risk and serves as the best defense against risk. I found this apropos when, while building an MCP to mitigate risk, the AI thought I had inadvertently shared sensitive data with it. (I hadn’t. I’d tested an outdated key.)

Knowledge helps with the decisions each company and each individual faces. Build a private LLM? Use only on-premise AI? Separate the hardware to restrict access for desktop machines? I am at the edge of my expertise here, because the field is enormous. But for people who have good judgment and are security-conscious, a lack of information will restrict adoption. 

IMPORTANT: Your AI Policy – A Tool to Increase or Decrease Risk

For those of you who are tech experts: if you haven’t figured it out already, I am not a tech expert. So I’ll use plain language instead of jargon for the following.

All companies rely on different types of software and hardware to do their work. Some have systems where everything plays nicely together; others have tech stacks that look like a pile of mismatched toys.

A conversation about AI adoption built on a duct-taped tech stack makes no sense, for several reasons:

1. If the company hasn’t invested in tech, it already has security gaps it doesn’t know about. These gaps amplify risk as AI enters the equation.

2. If the company doesn’t know what software people are already using and the associated workflows, it has no way to assess where AI integration adds value.

3. In my opinion, it is unfathomable to expect employees with little exposure to or training in non-AI tech resources (document sharing, video conferencing) to confidently move into AI learning.

Face Validity

If your company has a decent tech infrastructure and adoption work is helping people change their usage, the ask needs face validity. No one wants to spend an extra hour on something they used to do in 20 minutes unless you show them how it reduces the recurring workload to 5 minutes. One way to increase face validity (“this makes sense”) is to help people know what to anticipate.

In therapy, there is a term called “therapeutic induction” (also called role induction). It describes the process of a therapist telling a new client what to expect. That information reduces anxiety and fear, and it paves the way for the speed bumps to come. The same goes for any encounter with newness. If we are told that something will be frustrating for eight hours, we perceive the frustration as normal and expected rather than as evidence that we should give up. Asking people to adopt a new technology without helping them build the cognitive expectation and the physical time to adjust for the frustration punishes their initiation.

Application: Ask people to identify the parts of what they do that are most frustrating, or the items they’d most like to research or build. Everyone has a wish list that gets pushed aside for the fires of the day. When people can connect the pain of the learning curve to a personally meaningful outcome, it helps them get through what Gartner’s Hype Cycle calls the trough of disillusionment. Ensure space for people to share their frustrations and cross-pollinate learning. The shared applications, both the frustrations and the triumphs, adjust everyone’s expectations and help them manage the adoption process.