There is a paradox at the heart of the AI age, and it is urgent. Organisations have invested in AI tools with the expectation that they will reduce workload, increase efficiency, and give their people more capacity for high-value work. The reality, according to a growing body of 2026 research, is more complicated — and more concerning.
An eight-month study from UC Berkeley found that while AI tools do increase short-term productivity, they simultaneously intensify workload, accelerate task-switching, and blur the boundaries between work and recovery. The result is cognitive fatigue at a scale that organisations are only beginning to understand. Researchers coined a term for the worst cases: "AI brain fry" — a state of cognitive depletion that occurs when people are managing four or more AI tools simultaneously, constantly processing AI outputs while maintaining the relational and judgment demands of their roles.
Burnout is now at a seven-year high. A 2026 Aflac study found that 52% of workers say burnout is dragging down their engagement — up from 34% in 2025. And McKinsey's Health Institute research, drawing on 30,000 workers across 30 countries, found that one in five professionals is currently experiencing burnout symptoms.
The Mechanism: How AI Creates the Overload It Was Meant to Solve
Understanding how AI drives burnout requires understanding a shift in the nature of work. Before AI tools, cognitive work had natural stopping points — the analysis was done, the report was written, the decision was made. AI eliminates many of these stopping points. There is always another query to run, another output to review, another document to refine. The cognitive load is not reduced; it is transformed into a relentless stream of augmentation tasks that never feel finished.
Deloitte's 2025 research identified mental fatigue as now surpassing workload volume as the leading burnout predictor. This is significant. It means that even when people work fewer hours, they can experience greater burnout if the work they do is cognitively demanding, emotionally draining, and never fully resolved.
Amy Edmondson's Framework for Navigating This
Amy Edmondson — ranked second on the Thinkers50 list in 2025 and a keynote speaker at UNLEASH America 2026 — offers a framework that cuts through the complexity. Her recent work argues that in conditions of significant uncertainty and rapid change, the formula for organisational success is: "Aim high, team up, fail well, learn fast, repeat."
Each element of this formula addresses an aspect of the burnout challenge. Aiming high gives work meaning — and meaning is one of the strongest buffers against burnout. Teaming up reduces the isolation that amplifies stress. Failing well — what Edmondson calls "intelligent failure" — requires psychological safety, which also reduces burnout risk. Learning fast keeps people oriented toward growth rather than overwhelm.
Edmondson's research on the perils of replacing entry-level roles with AI adds an important dimension. When organisations remove the roles that provide learning through doing — the junior analyst, the executive assistant, the associate — they also remove the developmental pathways that build human capability over time. The short-term cost savings are real. The long-term capability costs may be greater.
What Leaders Must Do Differently
The 2026 research points to three specific shifts in how leaders need to approach AI adoption.
First, treat cognitive load as a real resource. Just as financial capital needs to be invested deliberately, cognitive capital needs to be managed. Leaders need to ask: what is the total cognitive demand we are placing on this team right now? What can we remove, simplify, or defer to make space for the AI-related tasks we are adding? This is not a soft question. It is a performance management question.
Second, build recovery into the workflow. The UC Berkeley research found that the most effective protection against AI-driven burnout is structured recovery — periods of time that are genuinely free from AI-tool engagement. This is not the same as having a lunch break with your phone. It means leaders actively protecting thinking time, deep work periods, and social connection that is not mediated by screens or AI outputs.
Third, make psychological safety non-negotiable. Edmondson's research is consistent: psychological safety is the single most important environmental factor in determining whether teams experience AI adoption as energising or depleting. When people feel safe to voice concerns, admit uncertainty, and make mistakes without blame, they can engage with AI tools as learning partners rather than threats to their competence.
A Note on the Younger Workforce
The burnout data contains a specific warning about younger workers. McKinsey's research found that people in their 20s and early 30s report higher stress and emotional exhaustion than any other demographic — a reversal of historical patterns. They are navigating AI-augmented workplaces without the experience base that older workers use to contextualise and regulate the demands placed on them.
This has direct implications for how organisations think about early career development. The investment in human skills — emotional intelligence, self-regulation, resilience, and learning agility — cannot be deferred to mid-career. It needs to start at the beginning.
The AI burnout paradox does not mean organisations should slow AI adoption. It means they should pace it thoughtfully, invest in human capability alongside technological capability, and build the leadership conditions that allow their people to thrive rather than just survive.
References
Aflac (2026) Aflac WorkForces Report 2026. Columbus, GA: Aflac.
Edmondson, A.C. (2025) 'The perils of using AI to replace entry-level jobs', Harvard Business Review, September.
McKinsey Health Institute (2026) Addressing employee burnout. New York: McKinsey & Company.
UC Berkeley (2025) AI tools and workplace cognitive load: An 8-month longitudinal study. Berkeley, CA: Haas School of Business.