The AI Productivity Paradox: Your Best Adopters Are Burning Out
By Chintan Dhanji, Managing Director, SC Strategy Consulting
Your best AI adopters are burning out. And most leadership teams don't see it coming.
The narrative is simple: give employees AI tools, they get more done, everyone wins. New research says that's dangerously incomplete.
Three signals point to an AI productivity paradox that deserves boardroom attention.
AI Doesn't Reduce Workload. It Intensifies It.
UC Berkeley researchers embedded themselves in a 200-person tech company for eight months. Employees using AI worked at a faster pace, took on broader tasks, and extended their hours - often without being asked.
Product managers started writing code. Researchers took on engineering work. Role boundaries blurred as AI made cross-functional tasks feel feasible. One engineer put it plainly: "You don't work less. You just work the same amount or even more."
The Berkeley team calls this "workload creep." It looks like productivity. But it quietly snowballs into cognitive fatigue, weakened decision-making, and turnover.
This pattern is consistent with what I've seen in enterprise AI engagements. When we evaluated 50+ AI use cases for a Fortune 500 healthcare company, one of the critical factors in our feasibility assessment was organizational readiness - not just whether teams could use AI, but whether the surrounding work design could absorb the change without overloading people.
(Source: Ranganathan & Ye, Harvard Business Review, Feb 2026)
The Data Confirms the Pattern at Scale
An NBER working paper analyzing nearly two decades of time-use data found that workers in high AI-exposure jobs are working roughly 3 additional hours per week since ChatGPT launched. That extra time comes directly from leisure, socialization, and exercise.
Faros AI's research across 10,000+ developers tells the same story: individual output increases, but organizational delivery velocity doesn't. AI-augmented code is getting bigger and buggier, shifting the bottleneck from creation to review. New supervisory labor that nobody budgeted for.
The pattern is consistent: individual productivity up, organizational productivity flat, human cost rising.
This mirrors a dynamic I've observed repeatedly in strategy work. When you optimize one part of a system without redesigning the whole, the bottleneck doesn't disappear - it moves. In AI adoption, the bottleneck is shifting from production to quality assurance, from creation to coordination, from individual output to organizational throughput.
(Sources: Jiang et al., NBER Working Paper 33536; Faros AI Productivity Paradox Report 2025)
McKinsey's Own Playbook Acknowledges the Gap
McKinsey's "Seizing the Agentic AI Advantage" report finds that nearly 8 in 10 companies have deployed gen AI, yet roughly the same proportion report no material earnings impact. Fewer than 10% of vertical use cases make it past pilot.
Their prescription: stop layering AI onto existing processes. Redesign workflows with AI at the core. Upskill teams not just on tools but on the cognitive demands of working alongside them.
The organizations seeing real returns treat AI adoption as a change management challenge, not an IT deployment.
This is the same principle that drives successful M&A integration. In a ~$300M healthcare acquisition I led, the ~$100M in captured revenue synergies didn't come from layering the acquirer's processes onto the target. They came from redesigning how the combined entity worked. AI adoption requires the same discipline.
What the Research Says to Do About It
The Berkeley researchers recommend building an "AI practice" - intentional norms around how AI gets used. This includes:
The Strategic Implication
For C-suite leaders, the question isn't whether your teams are using AI. It's whether you've redesigned the work around it - or just added AI on top of the same expectations and headcount plans.
The companies that will win the AI adoption curve aren't the ones that deploy the most tools. They're the ones that treat AI as a workforce design challenge: redesigning roles, workflows, and expectations to capture AI's value without burning out the people who deliver it.
AI governance isn't just about model lifecycle management and steering committees - though those matter. It's also about the human systems that determine whether AI adoption creates lasting value or an expensive burnout cycle.
The productivity paradox is real. The question is whether your organization will recognize it before your best people start leaving.