The Cognitive Toll of Automation
Mitchell Coburn, a 24-year-old business analyst from Manchester, resigned from his £50,000-a-year position at a mid-sized financial services firm in April 2026. His resignation letter, which later went viral on LinkedIn, described a slow erosion of professional competence that he termed "AI brain rot." Coburn's experience represents what sociologists and cognitive scientists call "occupational deskilling"—a phenomenon where technology, rather than enhancing human capabilities, begins to atrophy the very skills it's meant to augment. This case study examines the psychological and intellectual impact of constant AI mediation in professional tasks, exploring how cognitive offloading to large language models has led to a measurable decline in critical thinking ability and professional agency among knowledge workers.
For additional context on why AI systems appear intelligent without conscious understanding, see our related analysis of the illusion of AI consciousness.
Occupational Deskilling and Loss of Agency
Coburn's role involved data analysis and report generation using traditional tools like Excel and SQL. After his company implemented generative AI assistants for all routine analytical tasks, Coburn reports that his own proficiency declined within six months. "I found myself asking the AI to do what used to be my core competence," he stated in an interview. "I wasn't learning; I was just training the technology to replace me."
This experience aligns with decades of research on technological deskilling. Sociologist Harry Braverman first articulated the theory in his 1974 work "Labor and Monopoly Capital," describing how automation fragments skilled labor into repetitive, unskilled tasks. The digital era has extended this phenomenon beyond manufacturing to knowledge work, where AI mediates cognitive labor itself.
Key observations from Coburn's experience:
- Critical thinking atrophied through lack of use rather than active replacement
- Confidence in independent analysis decreased as AI outputs became the default reference point
- The ability to detect and correct AI errors diminished alongside foundational knowledge
- Professional identity shifted from "analyst" to "AI prompt engineer"
Cognitive offloading—the practice of delegating mental tasks to external systems—is not new. Calculators replaced slide rules; spellcheck replaced handwriting proficiency; GPS replaced map-reading skills. What distinguishes AI-mediated offloading is its comprehensiveness and the speed at which it erodes both skill and awareness of skill loss.
The psychological mechanism behind this phenomenon is closely related to what researchers call the attention wall, where sustained focus and executive control collapse under task complexity. When AI provides ready-made answers, human cognitive muscles lose their tone.
Cognitive Science Perspectives on AI-Mediated Work
From a cognitive psychology standpoint, Coburn's experience illustrates several well-documented phenomena:
The Automation Bias: Humans disproportionately trust automated outputs over their own judgments, even when the automation is incorrect. Studies show this bias intensifies with repeated exposure to AI systems, creating a dependency loop where users increasingly defer to automated outputs without verification.
The Generation Effect: Research demonstrates that generating information oneself (even imperfectly) leads to better retention than passively receiving information. AI-generated content, by providing ready-made answers, bypasses this active learning process entirely.
Implicit Task Decomposition Loss: Skilled professionals develop the ability to break down complex problems intuitively. When AI handles decomposition by providing pre-structured outputs, this meta-skill atrophies, leaving workers unable to approach problems without technological scaffolding.
Coburn described his breakdown in problem-solving ability: "When the AI gave me a response, I stopped thinking about the problem. I'd just edit the output instead of formulating my own analysis first. Before long, I couldn't even identify what questions to ask anymore."
The Transformation of Professional Identity
The shift from skilled practitioner to AI interpreter has profound implications for professional identity and workplace value:
Erosion of Tacit Knowledge: Traditional skill acquisition involves implicit, embodied knowledge gained through practice—knowing how to do something before knowing how to explain it. AI-mediated work bypasses this tacit layer entirely, producing workers who can operate tools but cannot troubleshoot them.
The Prompt Engineering Paradox: Proponents of AI integration argued that prompt engineering would become a valuable skill. In practice, Coburn found the opposite: the more he relied on AI for analysis, the less capable he became of crafting effective prompts or evaluating outputs critically.
Loss of Intellectual Ownership: When AI generates the core deliverable, what remains for the human to claim? Coburn reported feeling like a "content curator" rather than an analyst, with his primary contribution becoming the selection and editing of machine-generated outputs.
Imposter Syndrome Amplification: Paradoxically, heavy reliance on AI can intensify imposter syndrome. Workers know they're producing lower-quality work than the AI could, yet they lack the skills to match or exceed it independently. Coburn described checking his own work against AI outputs "like a security guard checking his own ID."
Organizational Implications and the Deskilling Trap
Organizations implementing AI tools often cite efficiency gains, but Coburn's experience reveals darker consequences:
Productivity Paradox: Initial adoption produces dramatic productivity boosts, but over time, workers become dependent on AI to perform basic analytical tasks. When the system fails—due to downtime, rate limits, or model errors—the entire workflow collapses.
Knowledge Transfer Erosion: As senior employees rely on AI for their work, they stop developing the deep expertise that would enable them to mentor others. The replacement cycle becomes broken: new hires cannot learn from seniors because both are operating at surface level.
Innovation Stagnation: True innovation requires the ability to connect disparate concepts and ask novel questions. When AI mediates all information processing, workers operate within predefined parameters, making truly original contributions increasingly rare.
Coburn's employer was unable to replace him with someone equally skilled in both analysis and AI oversight. They hired a junior analyst who was proficient with the tools but lacked the foundational understanding to develop better tools—or even recognize when the tools were failing. The result was a deskilled workforce, more efficient but less resilient.
The transition is not merely technical; it's epistemological. Work moves from a state of knowing (understanding why an analysis works) to having (producing outputs through prompt engineering). For another perspective on how AI systems mimic intelligence without understanding, see our exploration of the illusion of machine mind.
Paths Toward Human-Centered AI Adoption
Coburn's resignation wasn't surrender—it was a refusal to participate in his own deskilling. His story resonates because it represents a growing resistance movement among knowledge workers:
Deliberate Disruption: Some professionals have begun intentionally avoiding AI tools to preserve their skills. Coburn reports colleagues who turn off AI assistance during core working hours, using it only for review or administrative tasks.
Manual Reclamation: A fringe movement encourages workers to perform critical tasks entirely by hand before turning to AI for refinement. This preserves cognitive muscles while still leveraging AI's efficiency gains.
Skill Arbitrage: The most resilient workers are those who develop skills AI cannot easily replicate—domain expertise, stakeholder management, ethical judgment. Coburn is now pursuing certification in compliance and regulatory work, fields where human judgment remains paramount.
Organizational Countermeasures: Forward-thinking companies are implementing policies to prevent deskilling: mandatory skill-building time, limited AI access for junior staff, and periodic "AI detox" periods where teams work without assistance to maintain baseline capability.
The most promising approach combines AI's breadth with human depth: using AI for initial exploration and broad pattern recognition, while reserving deep analysis, synthesis, and judgment to human experts. This requires intentional design—not just of the technology but of work processes themselves.
As Coburn puts it: "AI shouldn't replace your thinking; it should do the boring parts so you have time to think harder about what matters." The challenge is designing work structures that make this distinction intentional rather than accidental.
Seeking the Human Element
In his search for something "more human," Coburn's story serves as a cautionary tale for organizations rapidly integrating generative AI. While efficiency gains are clear, the long-term impact on employee mental health and cognitive development remains a critical area for study in AI psychology.
The question isn't whether to use AI—it's how to structure work so that AI enhances rather than replaces human cognition. This requires recognizing deskilling as a real and measurable phenomenon, not just a side effect of technological progress.
For organizational leaders: Monitor not just productivity metrics but skill retention. Watch for signs of dependency before AI becomes the sole arbiter of professional competence.
For individual workers: Preserve your skills deliberately. The tools change, but the capacity for independent thought is your most valuable and irreplaceable asset.
For researchers: The deskilling phenomenon demands sustained study. We need longitudinal data on AI adoption's cognitive effects, not just the immediate efficiency metrics that dominate current evaluations.
Coburn's resignation letter concluded with a simple question: "If I'm just editing what the AI tells me to say, what am I really doing?" His answer was to walk away. The professional community's answer should be to build work structures that make him—and others like him—redundant, not obsolete.