AI brain fry is real. Here’s what leaders can do | Julie Bedard

Host Elise Hu introduces Julie Bedard’s argument that the uneasy feelings surrounding workplace AI—guilt about cutting corners, anxiety about falling behind, confusion and even embarrassment—are not merely personal reactions. Bedard says she has advised executives and board leaders for several years and has seen leadership teams become either “frantic or frozen”: some launch too many changes at once, while others leave employees to redesign their jobs individually through do-it-yourself AI adoption. Bedard says this person-by-person redesign can improve individual productivity but does not scale across an organization. She cites a seasoned manager whose team was producing “10x more output, 10x faster,” while the manager’s own job felt “20x harder.” The manager felt lonelier rather than more collaborative, and the quality of the output varied. Bedard says this experience was common among the managers she interviewed. Her central claim is that AI will dramatically reshape more than half of jobs in the coming years, so the important question is not whether work will be redesigned but how leaders will guide that redesign. She organizes her recommendations around three areas: attention, accountability and apprenticeship.

Bedard begins with attention, describing research into what it feels like to work alongside AI. A senior engineering manager told her that it was like having “a dozen browser tabs open in my head, all fighting for attention,” and that managing the tools had become harder than solving the underlying problem. Bedard calls this condition “brain fry,” distinguishing it from burnout. It is the mental fog and reduced decision quality caused by acute oversight of AI tools, and it can produce slower decisions and more errors. According to the figures Bedard cites, workers who experience brain fry report 39% more major errors and are 39% more likely to intend to leave. She says the problem arose because organizations layered AI tools onto existing workflows without removing obsolete work or considering the changes as a team; the added work was often more cognitively demanding. Her remedy is to redesign workflows around outcomes, with managers actively involved. When managers answer employees’ AI questions, mental-fatigue scores are 15% lower. She describes one company that rewrote job descriptions to emphasize strategic work, removed unnecessary handoffs and created a broad upskilling plan. The changes were coordinated rather than left to employees one by one.

Bedard’s second leadership concern is accountability. She says companies are placing AI agents on organizational charts and giving them names and job titles, then describes an experiment using the same document and the same errors in three versions: one said to be drafted by an AI tool, one by an “AI employee,” and one by an actual employee. Among managers already familiar with the idea of AI agents on org charts, personal accountability dropped by 9 percentage points when they reviewed work attributed to an AI employee. Those managers found 18% fewer errors, while escalations increased by 44%. Bedard’s interpretation is that reviewers scrutinized the work less carefully, found fewer problems and more often passed responsibility to someone else. She recalls one company that named its AI “Bob,” so that when something went wrong, people could say Bob had made the mistake. She warns that leaders did not intend to rewire accountability by making agents seem familiar, but the naming decision had that effect. Bedard recommends that every agentic system have a human overseer with clear decision rights, a defined escalation path and career consequences if the system causes harm. Employees also need training in reviewing AI-generated material, and she advises organizations to provide that training before naming their AI systems.

The third area is apprenticeship. Bedard says entry-level tasks such as drafting, analysis, first cuts and research can increasingly be performed, at least partially, by AI. Those tasks were not just low-level production work: they gave early-career employees the repeated practice, or “at-bats,” needed to develop judgment. Her research finds that junior employees can produce good output with AI on stretch assignments, but without expert support they do not necessarily develop the reasoning underneath that output. Bedard describes the result as “output without learning.” She says chief executives ask her repeatedly how to build talent in the age of AI, and points to several responses: rewrite jobs to be more rotational and project-based, pair junior employees with senior experts, and use younger workers’ natural AI fluency by positioning them as forward-deployed AI enablers. She also argues that companies should rewrite manager and leadership competencies to include coaching, AI upskilling and team-level work redesign. Rather than allowing AI to eliminate the development work companies have long claimed to value, she sees an opportunity to invest in managers who actively help people learn.

Bedard closes by returning to attention, accountability and apprenticeship, saying these are places where work redesign is happening to employees instead of being led by them. She defines the leader’s central job as optimizing work for good outcomes while combining human and machine intelligence. The companies getting the most from AI, she says, are 12 times more likely to have executives and leadership teams engaged with AI, four times more likely to dedicate structured time to AI learning, and two and a half times more likely to reshape workflows. She argues that the shame, fear of missing out and exposure people feel are understandable symptoms of being inside a major work transition, not evidence that an individual is handling AI incorrectly. Leaders should name what is changing, actively solve the organizational problems created by adoption and coordinate changes across teams. Bedard’s broader conclusion is that the AI conversation has focused too heavily on technology. It should now become a conversation about people, organizational change and leadership. The episode ends with Hu identifying the talk as Julie Bedard’s presentation at TED at BCG in 2026.