Use more AI, they said.

Here’s how it usually starts.

The CEO comes back from Davos, or a board meeting, or a vendor dinner. A new line appears in the quarterly priorities: “Accelerate AI adoption.” Within a month, every function has an OKR. Usage dashboards appear. Someone in the CTO’s office starts tracking tokens.

Meta built an internal leaderboard called “Claudeonomics” that tracked AI token consumption across more than 85,000 employees, with gamified status markers like “Token Legend” for top users (‘Meta killed employee AI token dashboard’, Fortune, April 9, 2026). Shopify’s CEO declared that “reflexive AI usage is now a baseline expectation” (‘Unserious Exploration’, Shopify, October 28, 2025). These are not edge cases. They are the template.

The directive is clear. Use more AI. The question nobody can answer is: for what?

The mandate without a thesis

Microsoft’s 2026 Work Trend Index found that only one in four AI users say their leadership is clearly and consistently aligned on AI (‘Agents, Human Agency, and the Opportunity for Every Organization’, Microsoft Work Trend Index, 2026). Three out of four employees experiencing the mandate without the map.

The pressure is real. Boards are asking about AI. Competitors are announcing AI. LinkedIn is a wall of AI transformation posts. The fear of being the company that missed the wave is driving more decisions than any analysis of where AI would actually help.

So the directive goes out. And because it arrives without a thesis – without a clear argument for what AI should change about the business – it lands as a metric. Adoption percentage. Active users. Token consumption. Sessions per week.

The thing about metrics without a thesis is that they get gamed. They always get gamed.

Where teams land

When teams receive a mandate to “use more AI” without clarity on what that means, they do the rational thing. They automate what they already do.

Same reports. Same dashboards. Same weekly emails. Same workflows. Just faster. Maybe with a chatbot in the middle. Maybe with a summarizer on top. The work does not change. The tools change.

A vague AI mandate does not transform work. It automates whatever was already on the calendar.

Asana’s research found that companies are asking employees to figure out AI adoption “on their own, without a clear blueprint for success,” while more than half of knowledge workers’ time is already wasted on low-value tasks (‘Scaling AI in 2025’, Asana, 2025). The AI gets bolted onto the low-value work rather than replacing it.

Harvard Business Review published research in early 2026 with a title that should have been a wake-up call: “AI Doesn’t Reduce Work – It Intensifies It” (‘AI Doesn’t Reduce Work – It Intensifies It’, Harvard Business Review, February 2026).

Microsoft’s same 2026 index found that 45% of AI users say it feels safer to focus on current goals than redesign work with AI. Only 13% say they are rewarded for reinvention even if results are not met.

The incentive structure punishes the exact behavior the mandate pretends to want.

The scream test

There is an engineering practice called a scream test. Turn something off – a report, a dashboard, an automated email, a data pipeline – and do not tell anyone. If nobody screams within two weeks, the thing was never needed.

Data teams do this routinely. It is one of the most effective ways to kill work that exists only because someone created it once and nobody ever asked whether it should keep running.

Most companies have never run a scream test on anything. And they are about to spend millions using AI to make those same never-needed artifacts faster.

That is the gap. The directive is “use more AI.” The question that should come first is: “what would happen if we stopped doing this?” If the answer is nothing, AI did not create value. It accelerated waste.

The real opportunity

Every company has a list of things they always wanted to do but could not. Analysis that was too complex. Products that needed engineering bandwidth that did not exist. Customer segments too expensive to serve. Internal processes everyone knew were broken but nobody had the resources to rebuild.

That list is the AI opportunity. Not faster versions of what already exists. New work that was previously too expensive, too slow, or too messy to attempt.

McKinsey’s 2025 State of AI found the same divide. The companies pulling real value from AI are distinguished by workflow redesign and embedding AI into business processes – not by adoption metrics (‘The State of AI’, McKinsey / QuantumBlack, November 2025).

The companies that understand this are not measuring token consumption. They are asking: what could we build that we could not build before?

The resistance is a signal

A stat has been circulating on LinkedIn: 41% of Gen Z employees are sabotaging their companies’ AI rollouts (‘New Data: 41% Of Gen Z Workers Are Sabotaging Their Employer’s AI Strategy’, Mark C. Perna, Forbes, March 24, 2025). The framing is dramatic. The signal underneath it is not.

Pew Research found that 52% of U.S. workers are worried about the future impact of AI in the workplace, while only 36% say they feel hopeful. Sixty-three percent say they do not use AI much or at all in their job (‘U.S. Workers Are More Worried Than Hopeful About Future AI Use in the Workplace’, Pew Research Center, February 25, 2025).

That is not sabotage. That is what happens when the directive arrives without the reasoning. If the mandate is “use AI” and nobody explains what problem it solves, the people closest to the work conclude that the point is performance, not improvement. And they respond accordingly.

The question that should have come first

The directive was never supposed to be “use more AI.” It was supposed to be: what would you build if complexity and resource constraints were not the limiting factor?

That is the generational opportunity. Not faster reports. Not automated status updates. Not token consumption leaderboards. The work that was always too hard, too slow, or too expensive to attempt.

But that requires a thesis. And a thesis requires thinking about the work before thinking about the tool.

Most companies skipped that step. Did yours?