Every week brings another announcement. Another enterprise signs an agreement with Anthropic or OpenAI or Google. Another internal memo goes out: "We're excited to share that [AI tool] is now available to all employees." The rollout meetings get scheduled. IT walks through the connectors. Someone from the vendor shows a few demos. Questions get answered about data security and acceptable use policies.
And then, for most organizations, not much changes.
This pattern is playing out across companies of every size right now. The tools are being acquired. The licenses are being distributed. The access is being granted. But the transformation that leadership expected? It's not materializing at the pace anyone hoped.
The Problem Isn't the Technology
The problem is that we're treating AI like software.
For the past two decades, enterprises have been trained to solve problems by buying tools. Need better project management? Buy Asana or Monday. Need better customer data? Buy Salesforce. Need better communication? Buy Slack or Teams. The pattern is deeply ingrained: identify a problem, evaluate vendors, select a tool, roll it out, train people on the features.
This works because traditional software does specific things in specific ways. You learn the interface. You learn the workflow. You use the tool as designed.
AI doesn't work like that.
When you give someone access to Claude or ChatGPT or Gemini, you're not giving them a tool with a defined workflow. You're giving them a capability that can do almost anything, but only if they know how to direct it. The interface is deceptively simple: a text box. But using that text box effectively requires a completely different way of thinking about work.
A Different Kind of Skill
The shift isn't about learning features. It's about learning how to decompose problems, organize information, and provide context in ways that most knowledge workers have never had to do before.
Consider what happens when someone encounters a complex task. In the old model, they would ask: "What tool do I need for this?" Maybe they'd search for a template, or look for software that handles that type of work, or ask a colleague who's done it before.
With AI, the question changes: "How do I explain this problem clearly enough that an intelligent system can help me solve it?" That's a fundamentally different skill. It requires thinking about what information matters, what context is missing, what assumptions need to be made explicit, and what a good outcome actually looks like.
Most people have never been asked to think this way. And no rollout meeting is going to teach them.
The Data Tells the Story
This is why so many AI deployments plateau quickly. The data bears this out: 74% of companies have yet to see tangible value from their AI initiatives, and nearly two-thirds of organizations remain stuck in the pilot stage1. The early adopters figure it out through experimentation. They become the "AI people" in their departments, the ones everyone asks for help. But the broader workforce, the people who were promised productivity gains, end up using these powerful systems for the same basic tasks: drafting emails, summarizing documents, answering simple questions. Helpful, sure. Transformative? Not even close.
The gap between adoption and value is widening. In 2025, 42% of companies abandoned the majority of their AI initiatives before reaching production, up from just 17% the previous year2. These aren't failed experiments with immature technology. These are implementations of proven tools that simply didn't deliver on their promise. The technology worked. The operationalization didn't.
What Actually Works
The organizations that will actually capture value from AI are the ones that recognize the rollout is just the beginning. The real work is helping people develop new mental models for problem-solving. It's teaching them to see their work differently. To notice when a task could be decomposed into smaller pieces. To recognize when the real problem is context organization, not execution. To understand when they need a quick answer versus when they need to build something more structured.
This isn't a training problem in the traditional sense. You can't solve it with a two-hour workshop on prompt engineering. It requires ongoing practice, feedback, and a willingness to rethink workflows that have been calcified for years.
The companies announcing their AI rollouts this month aren't wrong to be excited. The technology genuinely is powerful. But somewhere between the press release and the productivity gains, there's a gap that most of them haven't fully reckoned with yet.
Access is the easy part. Adoption is harder. But the real challenge is transformation: helping people think differently about how they solve problems in the first place.
That's the work that matters. And it's just getting started.