AI can be present across the enterprise and still nowhere in actual work.

That is the challenge many organizations are running into right now. As AI moves into enterprises, IT leaders are increasingly being asked to do more than deploy the technology, according to CIO’s 2025 State of the CIO Research. They are being asked to drive adoption across their workforce, a responsibility that has traditionally sat with Talent.

That shift in ownership matters because it is exposing a gap. Yes, AI requires governance, security, and technical oversight (all buckets squarely in IT’s wheelhouse). But AI rollouts also live and die on something messier: culture (i.e., whether people build the habits, confidence, and new ways of working needed to actually use the technology well).

Unfortunately, that human side of AI adoption is easy to miss when the rollout is led too heavily through a technical lens. And as great IT leaders understand, that is exactly where a false start can happen, too.

For organizations that want to draw real value from AI, the challenge is bigger than deployment. It is about creating the conditions for adoption to take hold by treating AI as a change problem just as much as a technology one (if not more).

What IT leaders may get wrong about AI adoption

Most organizations do not struggle to make AI available. They struggle to make it usable.

Many IT leaders start by doing the sensible thing from a systems perspective. They secure the platform, communicate availability, and may even offer introductory training through vendor resources or self-serve videos.

All of that is useful, but none of it is enough.

Earlier this year, we saw this play out with an insurance company that had cycled through three Chief AI Officers (CAIOs) in just two years. The technology was there. The training rollout was happening, but no one was using it. Adoption was still lagging because the organization had approached AI too technically.

Employees may understand the range of possibilities AI opens up. But when they feel stuck, their questions often point to a larger people challenge that still has not been addressed:

• Where does this fit into my role?
• What does good use look like here?
• What happens if I get it wrong?

These may sound like soft questions, but they have hard consequences. They shape whether people build real habits around AI or decide it is easier and safer to stick with old ways of working.

That is why AI adoption is not just about putting a tool in people’s hands. It is about building the confidence, judgment, and shared norms that help people use it well.

For IT leaders, that is the real mindset shift. And that is why it helps to think about AI adoption as a three-phase human-centered journey, not a one-time rollout.

The three phases of effective, human-centered AI adoption

In our experience, organizations that succeed with AI adoption tend to move through three phases:

Phase 1: Early friction

Diagnostic: The tool is available, but people are not using it consistently. A few early adopters jump in. Most employees hang back. Teams have uneven norms. Managers are not aligned on expectations. People are unsure when to use AI, how much to trust it, or whether experimenting with it is even safe.

This is where we see most AI rollouts stall. At this stage, leaders often respond by doubling down on access or feature awareness.

But that rarely solves the real issue. The problem is not that people fail to see the tool’s value. The problem is that they have not yet built the agency, confidence or shared understanding to use it in the context of their real work.

What to do: Turn curiosity into daily habits. Help people use AI in their existing work so it feels practical, safe, and valuable.

Phase 2: Optimization value

Diagnostic: You are seeing your first real gains from AI. Your people are taking work that already happens and doing it faster or better with the new technology. Use becomes less theoretical and more embedded in how work actually gets done.

This is what we call “replacement thinking”: same work, better tools. AI is helping teams accelerate, offload, or streamline repetitive tasks, but the fundamental shape of work has not changed yet.

What to do: Move beyond encouraging more use for its own sake. Researchers at Stanford University found that people are significantly better at using AI when a colleague or manager shows them how it solves a concrete, high-value problem.

Help teams uncover AI use cases from first principles so they can spot where AI fits valuably into their work and where it can create more value.

Phase 3: Explosive innovation value

Diagnostic: At this stage, the question is no longer, “How can AI improve the work we already do?” It becomes, “How should we redesign our work, roles, processes, and even products now that AI is a foundational capability?”

This is where organizations unlock what ExperiencePoint’s VP of Organizational Innovation, Andrew Webster, calls “reimagination value.” In his work with organizations becoming AI-native, he has seen that this kind of value usually appears only after teams have spent enough time with the technology to understand not just what it can do, but what new possibilities it creates.

This is the difference between digitizing Blockbuster’s existing rental model online and reinventing it entirely, as Netflix did.

What to do: Stop treating AI as an add-on to the current model and start using it to invent a new one. This is the moment to rethink how work gets done and how value gets created.

Organizations do not get here by skipping ahead. They get there by helping people move through the earlier phases first.

The IT leader’s guide to AI adoption success

Most organizations stall in phase one of our three-phase AI adoption journey, where early friction keeps people from turning initial curiosity into real, repeatable use. If your rollout has been live for weeks, months, or even years, and adoption is still hovering around 15 to 20 percent, that is usually why. The good news is that organizations do overcome this false-start stage. And the ones that do tend to get four things consistently right:

1. They treat adoption as a change challenge, not just a technical one. They know awareness is only the first step, and that building agency and confidence, what we like to call “What’s in it for me?”, goes a long way.

2. They give teams clearer guidance on where AI should be used, where human judgment still matters most, and what responsible use looks like in practice.

3. They create opportunities for real practice. Not just demos or self-serve training videos, but applied learning with peers around the kinds of scenarios people actually face in their work. At ExperiencePoint, we have seen this kind of AI training increase adoption by at least 40 points.

4. They foster the psychological safety people need to move past fear and start experimenting, learning, and improving. Bringing people together helps reinforce that they are not figuring it out alone. They are in it together.

These are the moves that help organizations break out of early friction and turn AI from an interesting tool into a real way of working.

Why the human side of AI adoption matters more than most leaders think

Yes, AI rollout requires the technical foundation to be right. But those things alone do not create adoption. Adoption happens when people have the confidence to use AI, the clarity to know where it fits, and the support to build new habits around it. In other words, adoption happens when the rollout accounts for the human side of change.

That is why the best way to avoid a false start is not simply to launch the tool and hope people figure it out. It is to give them the chance to practice with peers, in the context of their real work, with enough guidance and safety to build confidence as they go. Because the real measure of an AI rollout is not whether the platform went live. It is whether people have changed the way they work for the better.