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

The workplace transition to AI can bring guilt, confusion, and anxiety, and Julie Bedard says these mixed feelings are normal. She predicts AI will dramatically reshape more than half of jobs in the coming years. The gains in output do not necessarily make work easier for everyone: one manager described a team producing 10 times more output 10 times faster, while her own job felt 20 times harder. Bedard says managers in her research often saw productivity rise alongside declining connection and cohesion. The central leadership challenge is therefore not simply adopting tools, but changing work so higher output does not come at the cost of people’s capacity and collaboration.

Bedard distinguishes “brain fry” from burnout. Brain fry is acute mental strain associated with overseeing AI tools, and can show up as mental fog, slower decisions, and more errors. Workers experiencing it reported 39% more major errors and were 39% more likely to intend to leave. These findings make the way AI is introduced important: adding tools while leaving existing work untouched, or failing to redesign workflows as a team, can contribute to the problem. Higher productivity alone does not establish that a workflow is sustainable or that employees have the attention needed to check AI output.

Manager involvement is one practical lever. Employees whose managers answered their AI questions had mental fatigue scores 15% lower, according to Bedard. She recommends redesigning workflows around intended outcomes, involving managers in that work, and keeping attention, accountability, and apprenticeship people-led responsibilities. That means treating AI integration as a team and management problem, not only as a matter of individual tool use. The implication is that leaders need to make space for employees to learn and ask questions while clarifying how work should change, rather than expecting people to absorb new oversight demands on top of unchanged responsibilities.

AI agents also complicate accountability. In an experiment, managers reviewed the same document containing the same errors, but those familiar with AI agents on org charts found 18% fewer errors. When the work was attributed to an AI employee, they were 9 percentage points less likely to hold a drafter accountable, while escalations rose 44%. Bedard’s point is not that responsibility disappears when AI is involved. Agentic systems need a human overseer with clear decision rights, an escalation path, and career consequences. Employees also need training to review AI-generated content, so delegation does not obscure who is expected to catch mistakes or act when something goes wrong.

The effects on early-career development require particular attention. Junior employees may use AI to produce strong work on stretch assignments without developing the judgment behind that work, especially if they lack expert support. Bedard suggests rotational or project-based jobs, pairing junior employees with senior experts, and using junior employees as forward-deployed AI enablers. These approaches connect AI use with guidance and learning rather than treating finished output as evidence that an employee has acquired the underlying expertise. Leaders need to preserve opportunities for people to build judgment as tasks and workflows change.

Companies getting the most from AI were 12 times more likely to engage executives and leadership teams, four times more likely to dedicate time and structure to AI learning, and 2.5 times more likely to reshape workflows, Bedard says. Those figures point to organizational choices, not just individual enthusiasm: leadership engagement, protected learning, and redesigned work are associated with stronger AI results. Bedard also urges companies to update manager competencies to include coaching, AI upskilling, and team-based work redesign. The broader message is that leaders should pair adoption with clear accountability, human expertise, and deliberate attention to how AI changes the work people do.