Clover ERA.
The arc that lifted output per hour for 75 years has bent, broken, and inverted. What used to be a story about capital is now a story about people, and the people are burning out. The arc shows productivity adjusted for human capacity, not raw output per hour.
Workers got more productive each hour because the tools around them got better, not because they worked harder. Output per hour rose with every wave of equipment, software, and process. The factory floor automated. The office digitised. The supply chain globalised. At each turn the worker rode a curve someone else had built.
The arithmetic was simple. Capital deepening did roughly six-tenths of the work. Technology and process improvements did roughly three-tenths. The human contribution was the residual, and the residual was small.
Each of the engines that produced 75 years of gains has stalled or reversed. The breaks were not simultaneous; they unfolded across two decades and have compounded sharply since 2020. Capital deepening slowed first in 2005. Manager engagement weakened through the late 2010s and has dropped 9 points since 2022. AI began increasing cognitive demand from 2024 onward. The lines that used to climb together now diverge.
Capital contribution to productivity growth has dropped from roughly 60% (1970–2005) to under 20% (2010–2025). The big infrastructure waves are behind us. Each marginal dollar of capital buys less marginal output than it did a generation ago.
The post-2008 recovery was capital-light by design. Fewer factories, more software. Software depreciates faster than steel.
AI is now increasing cognitive demand rather than reducing it. Whether AI infrastructure costs eventually fall (as historical precedent suggests) or continue rising as demand outpaces efficiency gains (as Jevons Paradox in current AI economics suggests), the human-side cost is real now. Companies are paying for AI in two ways: in capex that is consuming operating cash flow and in cognitive load on workers that is depleting their capacity to deliver. Neither cost is being tracked or managed in any meaningful way. The capex appears on the cash flow statement; the cognitive load does not appear on any dashboard. Both costs are real, both are growing, neither is being seen for what it is.
AI was supposed to reduce cognitive demand. The 2025 data shows the opposite. Heavy AI users report higher cognitive effort, not lower. Three structural mechanisms compound:
Microsoft Research (Jan 2025) found heavy AI users report higher cognitive effort. UC Berkeley’s longitudinal study found 67% of workers who adopted AI tools in 2025 worked more hours, not fewer, by year-end. Deloitte’s 2025 Workforce Intelligence Report identified mental fatigue and cognitive strain as the leading predictors of burnout, surpassing workload volume for the first time.
Gallup’s 2026 State of the Global Workplace report shows global engagement at 20%, the lowest level measured since 2020. Manager engagement has dropped 9 points since 2022. The historical “manager engagement premium,” the gap between manager and individual-contributor engagement that used to subsidise team morale, has effectively closed.
The retention layer is camouflaging the problem. MetLife 2026: 56% of employees stay out of necessity, only 18% out of genuine commitment.
The composition of growth has flipped. The factor that used to be a small residual is now the residual that matters most, and the only one still moving. Capital is exhausted. Technology is producing cognitive load instead of relief. What is left is the person standing in the workplace.
The median company in our cohort loses twenty million dollars a year to the depletion of the people growth now depends on. The bill is no longer abstract. It already lands every quarter, it just isn’t named.
As the cohort grows toward the Q2 2026 publication, the median has held steady. New companies confirm the pattern rather than dilute it. We call the whole thing Silent Degradation.
$20M
Range $8M to $65M across our growing Q1–Q2 2026 cohort. As of Apr 28, 2026 · n=11
Less than half of that cost is voluntary departures. The rest is what depletion looks like before someone leaves: disengagement, manager drag, promotion risk, innovation suppression, customer impact.
The cost layers leaders haven’t been tracking are larger than the one they have. The dashboards are looking in the right place but at the wrong layer.
Below the four-move thesis sits a simpler claim. Productivity matters. Managers are the lever. The manager group gets the least attention. If those three land, the next step becomes obvious.