🎧 Listen to this article

Narrated by Talon · The Noble House

On February 9, 2026, TechCrunch published a piece that should have gotten more attention: "The first signs of burnout are coming from the people who embraced AI the most." The article described a pattern showing up across companies that adopted AI tools aggressively. Employees who used AI to accelerate their work didn't end up working less. They ended up working more — faster, broader, longer hours than before.

Two days later, Fortune's UC Berkeley coverage confirmed the mechanism: "Nonstop productivity may come at the cost of rest and work quality." HBR published its own analysis the same week: "AI Doesn't Reduce Work — It Intensifies It." The Harvard review found employees worked at a faster pace, took on a broader scope of tasks, and extended work into more hours of the day, often without being asked to do so.

The Pattern Is 160 Years Old

In 1865, William Stanley Jevons published The Coal Question and made an observation nobody wanted to hear: as steam engines became more fuel-efficient, total coal consumption went up, not down. More efficient engines made coal-powered production cheaper, so more industries adopted steam power. The efficiency gains created more demand, not less.

This became known as the Jevons Paradox. Cars got more fuel-efficient; people drove more miles. LED lights use a fraction of incandescent energy; total lighting consumption held relatively flat as we illuminated more things. Nuclear power was supposed to make electricity "too cheap to meter"; total global electricity consumption has increased every decade since.

AI is the Jevons Paradox applied to human attention. When you give a knowledge worker a tool that cuts report-writing from four hours to forty minutes, you'd expect them to reclaim three hours. That's the vendor pitch. What ManpowerGroup's 2026 Global Talent Barometer found — surveying nearly 14,000 workers in 19 countries — was that workers' regular AI use increased 13% in 2025, but confidence in the technology's utility fell 18%: people are using the tools more and trusting them less, experiencing the productivity gains as pressure rather than relief.

Jevons Paradox applied to AI productivity and attention visualization
The Jevons Paradox: efficiency improvements historically increase total consumption rather than reducing it. Steam engines made coal cheaper to use; people built more steam-powered machines. AI makes knowledge work faster; organizations assign more knowledge work.

The Bottleneck Moved

The original AI productivity studies (Stanford HAI's 2023 GitHub Copilot study, Nielsen Norman Group's 2023 analysis, others) measured task-level performance. AI users completed specific defined tasks 25–55% faster. These studies were real. The productivity gains were real. What the studies didn't measure: what happened after the task was complete.

When an AI tool drafts five versions of a marketing email, someone has to choose between them. When it generates a risk analysis for a project, someone has to evaluate it against institutional knowledge the AI doesn't have. When it accelerates report production from four hours to forty minutes, the number of reports requested tends to increase proportionally to the available capacity. The bottleneck moved from production to evaluation, and evaluation is cognitively expensive.

Fortune's AI Productivity Paradox reporting (February 17, 2026) noted that thousands of CEOs surveyed by the Federal Reserve Bank of Atlanta admitted AI had no measurable impact on employment or productivity at the firm level — echoing a point about Robert Solow's original "productivity paradox" from the 1980s, when computers were visibly everywhere but didn't show up in productivity statistics for a decade.

Three Responses That Work

Define the ceiling, not just the floor. If AI cuts report creation time from four hours to forty minutes, the answer to "what should I do with the remaining 3.5 hours?" cannot be "more reports." The organization needs to define maximum productive output, not just track velocity. Without a ceiling on output, AI tools become demand generation for cognitive load.

Protect evaluation time explicitly. The new bottleneck is reviewing, verifying, and deciding based on AI-generated material. This work looks passive but isn't — evaluating five AI-generated options requires active judgment. Organizations that treat evaluation as "not real work" because it's shorter than production will systematically exhaust their highest-judgment people.

Watch the quality signal, not just the speed signal. The UC Berkeley finding was specific: nonstop productivity may come at the cost of work quality. The pattern follows from the mechanics — when evaluation time is compressed, error rates increase in ways that are hard to measure in the short term and expensive in the long term.

AI attention tax worker cognitive load evaluation burden visualization
The shift in where cognitive work happens: AI tools reduce production time but increase evaluation load. The bottleneck moves from "can we create this?" to "can we judge which version is correct?" Judgment doesn't compress at the same rate as production.

The Structural Question

The TechCrunch finding — burnout appearing first in the power users — is a leading indicator, not a lagging one. It shows up in the most productive employees before it shows up in aggregate workforce metrics. Those are the employees worth understanding and protecting. They're running the Jevons experiment on themselves and showing you the results before your efficiency dashboard does.

The useful question for 2026 is not "how do we get more productivity from AI?" Most organizations have figured that part out. The useful question is: "Given that AI expands output capacity, what will we not do with that capacity?" Answering that question is harder than deploying the tool. It's also the question that determines whether AI deployment generates sustainable value or organizational stress.

Sources: TechCrunch, "The first signs of burnout are coming from the people who embraced AI the most," February 9, 2026; Fortune/UC Berkeley, "In the workforce, AI is having the opposite effect it was supposed to," February 10, 2026; Harvard Business Review, "AI Doesn't Reduce Work — It Intensifies It," February 2026; Fortune, "AI Productivity Paradox — CEOs survey," February 17, 2026; ManpowerGroup 2026 Global Talent Barometer (n=14,000, 19 countries); William Stanley Jevons, The Coal Question, 1865; Federal Reserve Bank of Atlanta CEO survey (via Fortune)

Sources