Every AI business case rests on the same promise: the technology absorbs the routine work so people can move up to higher-value thinking. The promise is usually kept, and that is exactly the problem. Strip the simple tasks out of someone’s day and what remains is a dense sequence of hard calls, unfamiliar problems, and accountability for outcomes, with nothing low-stakes in between. Julie A. Stone, who advises leaders on human performance in AI-enabled organizations, argues that this is the blind spot sitting underneath most enterprise AI strategies. The metrics being tracked (adoption, cost, throughput) all point up, while the human resource the entire system depends on quietly runs down. “AI can make individual tasks easier and still make the workday harder,” she says. That sentence should unsettle anyone who has signed off on a deployment this year.
Capacity Is The Constraint Nobody Measured
The resource Stone is pointing toward has a name but little presence in corporate reporting: cognitive capacity, the mental fuel people spend to think, evaluate, and decide. It is finite, it depletes over the course of a day, and AI deployment pulls on it harder than the work it replaced. Routine tasks were never just filler. They were the low-demand stretches that let people recover between hard decisions. Remove them and the workday becomes an uninterrupted run of judgment calls. “You can go from one demanding decision to the next, even as individual tasks take less time,” Stone says. The speed gain is real. So is the cost, and only one of them shows up in the quarterly numbers.
What Stone reports hearing from leaders across global markets fits the pattern: broader responsibilities, more complex decisions, less room to think or recover. That combination is not a wellbeing concern dressed up in operational language. It is a direct threat to the value of the AI investment itself, because the technology does not steer. People do. They choose the problem, supply the context, and decide whether the output holds. “When we’re mentally depleted, it’s harder to challenge a plausible answer or see a better way to use the technology,” Stone says. A plausible wrong answer is the most expensive thing an AI system can produce, and catching it is precisely the task that requires a mind with something left in reserve.
The Performance Review Is Asking The Wrong Question
Here is where the gap turns into a management failure. When a capable person starts missing things, the organizational reflex is immediate and familiar: question their skills, their effort, and their commitment. Stone wants leaders to interrupt that reflex and look at the workday first. The question is not whether the person had the capability. It is whether they had the capacity to use the capability they already had. Those are two different variables, and most performance conversations collapse them into one.
The diagnostic she offers is simple and leaders can run it without a new system. When several people are struggling under the same conditions, the problem is not distributed across several individuals. It is in the conditions, and conditions are something leadership controls. “Capabilities and cognitive capacity together determine the quality of human contribution,” Stone says. Treat that as an equation rather than a slogan and the implications for skilling programs are sharp. Training that adds capability to a workday with no capacity left is money spent on an input the system cannot absorb. Organizations rolling out AI literacy curricula, while compressing the time available to think, are solving one half of the problem and worsening the other.
Positive Friction And The Investment Review Question
The practical starting point Stone proposes costs nothing and belongs in a meeting most companies already hold. At the next AI investment review, ask what the technology will require of people. Not what it will save, but what it will demand. If a team will own more complex decisions as a result of the deployment, what comes off their plate to make room? That question is rarely asked because the business case is built on removal, not redistribution, and nobody wants to log a subtraction against a project pitched on addition. Asking it turns a vague productivity claim into a workload design decision with an owner.
The second move is counterintuitive enough that it needs a name, and Stone gives it one. If AI drafts a recommendation in minutes, some of the recovered time should go back into testing an assumption before anyone commits. She calls this positive friction, a deliberate pause that improves the decision. The instinct in most organizations is to bank every minute AI returns and push the cycle faster, which is how unexamined assumptions get locked in at speed. Reinvesting a fraction of that time in scrutiny is not a tax on the gain. It is the mechanism that protects it, and it is the kind of habit a skilling program can teach. “We can do more with AI than work faster,” Stone says. “We can solve harder problems and pursue ideas we didn’t have time to explore before.” Whether that upside arrives depends on something no productivity report will ever surface. A dashboard can show how much faster a team delivered. It cannot show which assumptions went unchallenged, or which ideas nobody had the capacity to develop.
Follow Julie A. Stone on LinkedIn for more insights on workforce skilling, cognitive capacity, and human performance in AI-enabled organizations.
