The Subscription Graveyard
Open your company's expense report. Count the AI subscriptions. ChatGPT Team. Claude Pro. Copilot seats. Jasper. Otter. Notion AI. Midjourney. Perplexity. That one tool someone on the marketing team swore would change everything — what was it called again?
Now ask yourself: how many of these are actively used by more than two people? How many have changed the way your team actually works — not just sped up a task here and there, but fundamentally altered a workflow?
If you are being honest, the answer is uncomfortable. And you are not alone.
The Numbers Behind the Hangover
McKinsey's data tells the story in two statistics: 88% of organizations have adopted AI tools, but only one-third have scaled AI beyond a single function. That gap — between adoption and impact — is the defining challenge of 2026.
The 2023-2024 hype cycle was intoxicating. Every conference keynote promised transformation. Every vendor demo looked magical. Procurement teams approved budgets they did not fully understand because the fear of falling behind was louder than the demand for evidence.
The result is what we now have: tool sprawl. Organizations sitting on a pile of AI subscriptions that nobody fully uses, with no clear picture of what is working, what is redundant, and what is actively creating confusion.
What We Got Wrong
The mistake was not buying the tools. Many of them are genuinely capable. The mistake was treating AI adoption as a procurement problem rather than a behavior change problem.
The logic seemed sound at the time: give people access to powerful tools, and they will figure out how to use them. It is the same logic that gave every employee a gym membership and then wondered why nobody went. Access is not adoption. Adoption is not transformation.
Here is what actually happened in most organizations:
Phase 1: Excitement. A few enthusiasts — usually the most technically curious — started using AI tools and got impressive results. They became internal evangelists.
Phase 2: Rollout. Based on those early wins, leadership approved broader access. Everyone got seats. Training sessions were scheduled. Lunch-and-learns happened.
Phase 3: The Plateau. Usage data showed a familiar curve: initial spike, steady decline, flat line. Most people tried the tools a few times, got mixed results, and quietly went back to their old workflows. The enthusiasts kept using AI. Everyone else did not.
Phase 4: The Reckoning. Renewal conversations forced the question: are we getting value from this? The answer was usually "some people are, most are not." But canceling felt like going backward, so the subscriptions continued.
The Deeper Problem
The plateau is not a training problem. You can run more workshops. You can create prompt libraries. You can designate AI champions on every team. These help at the margins, but they do not solve the fundamental issue.
The fundamental issue is that integrating AI into how you work requires changing habits, not just learning features. And habit change is hard. It is hard for individuals, and it is exponentially harder for organizations.
Research from Anthropic's AI Fluency Index — a study of 9,830 conversations conducted in January 2026 — reveals the depth of the problem. The data shows that 85.7% of users iterate on AI outputs, which sounds encouraging until you read the next finding: far fewer critically evaluate those outputs. Polished AI responses actually reduce people's ability to identify gaps by 5.2 percentage points. The better the output looks, the less carefully people examine it.
This is not a tool problem. It is a cognitive problem. People are interacting with AI frequently but not effectively. They are generating outputs without building the judgment to evaluate them.
The Real 2026 Question
The question has shifted. In 2023, it was "which AI tools should we buy?" In 2024, it was "how do we get people to use them?" In 2026, it is something harder and more important:
How do we actually change how we work?
This question cannot be answered with a procurement decision or a training program. It requires rethinking workflows, decision-making processes, and the cognitive habits that underpin daily work.
Consider what it means to move from "I use AI sometimes" to "my operating model is built around AI." That transition requires:
- A persistent context system so AI understands your business, your standards, and your constraints without starting from zero every interaction
- Clear verification and control standards so people know when AI output is good enough and when it needs human judgment
- Codification and reuse practices so what works once becomes a foundation for what comes next, not a one-off
- Defined orchestration patterns so people know what to delegate and what to keep
These are not features of any tool. They are organizational capabilities. And building them is a fundamentally different activity than deploying software.
From Tool Adoption to Capability Building
The shift that matters in 2026 is the shift from tool adoption to capability building. Here is what that means in practice.
Tool adoption asks: "How do we get people to use this software?"
Capability building asks: "How do we redesign this workflow so that AI is integral to how it runs?"
Tool adoption is measured in seats and logins. Capability building is measured in process changes and outcome improvements.
Tool adoption can be delegated to IT. Capability building requires leadership to go first — because if the leader has not changed how they work, the team will not either. This is not theory. It is a pattern we see repeatedly. The organizations where AI adoption sticks are the ones where the founder or executive transformed their own workflow before asking the team to change.
This is the critical insight: AI transformation is a leadership behavior problem disguised as a technology problem.
The Leader-First Principle
There is a reason most AI rollouts stall at the team level. The leader approved the tools but did not change their own operating model. They still run meetings the same way. They still make decisions the same way. They still manage their time the same way.
The team reads this accurately. If the leader is not working differently, the implicit message is clear: this is optional. And optional things get dropped the moment workload increases.
The path forward starts with the leader. Not with a memo about AI strategy. Not with a training budget. With the leader sitting down and honestly assessing: how do I actually work with AI today? Where am I on the 7-Level AI Proficiency Framework? What is my weakest pillar?
For most leaders, the honest answer is L2 to L3. They use AI for specific tasks, but their operating model has not changed. And that is the ceiling for their entire organization.
What Comes After the Reckoning
The post-hype reality is not a failure story. It is a maturation story. The tools are good. The technology is ready. What was missing was the understanding that human behavior change — not software deployment — is the core challenge.
The organizations that will lead in the next phase are not the ones that bought the most tools. They are the ones that invested in building actual capabilities: context systems, verification standards, codified workflows, and the cognitive shifts required to use them.
The question is not whether your organization will make this shift. The question is whether you will make it deliberately — with a clear assessment of where you are and a structured path forward — or whether you will keep renewing subscriptions and hoping that adoption happens on its own.
If you are ready to have an honest conversation about where your organization actually stands — not where your tool stack suggests you should be — that is where we start.