munkapalyakezdoknek.hu / Articles / Analysis
Why Friction Matters: AI, Deskilling and Learning at Work
Relying too heavily on AI removes the struggle required to build deep professional judgment and tacit knowledge.
The Triple Deskilling Pattern
Automation does not simply remove tasks. It removes the practice required to maintain expertise. This creates a pattern where the human operator loses the ability to perform the core function when the machine fails. The risk is not that the AI takes the job, but that it takes the training ground.
Consider the baker who uses a machine to knead dough. The machine produces consistent bread. Yet the baker forgets how the dough feels when it is ready. The tactile feedback loop breaks. The skill resides in the hands and the senses, not just the final output. Without the physical resistance of the dough, the baker cannot judge quality.
This pattern appears across industries. In call centers, scripts handle routine inquiries. Agents lose the ability to improvise when a customer presents a unique problem. In data analysis, automated tools generate reports instantly. Analysts stop questioning the underlying assumptions because the answer appears without effort. The ease of the process masks the loss of critical thinking.
The result is a workforce that looks efficient but lacks resilience. When the system encounters an edge case or a novel situation, the human operator has no framework to respond. They have outsourced the very cognitive work that defined their profession. The machine becomes a crutch rather than a tool.
- Automation removes the practice needed to sustain expertise.
- Ease of output often masks the loss of critical judgment.
- Human operators lose the ability to handle novel situations.
Friction as the Engine of Learning
Learning requires resistance. Psychologist Robert Bjork called this 'desirable difficulties.' When learning is easy, retention is poor. When the brain struggles to retrieve information or solve a problem, neural pathways strengthen. The difficulty itself is the signal that learning is occurring. Remove the struggle, and you remove the growth.
AI excels at removing friction. It provides answers before questions are fully formed. It summarizes texts without the need for note-taking. It corrects code before the developer understands the error. The speed is impressive. The cost is the elimination of the mental processing time required to build memory and understanding.
In personal knowledge management communities, members observe that organizing notes manually is more valuable than the notes themselves. The act of deciding where a thought belongs forces the brain to connect ideas. Automated sorting saves time but bypasses the cognitive work of synthesis. The user ends up with a library they do not understand.
This phenomenon creates an illusion of competence. You have the information saved. You feel you know the material. But you cannot apply it without the tool. The skill was never internalized. The friction was the teacher, and the algorithm skipped the lesson. You must reintroduce difficulty to reclaim the learning process.
- Desirable difficulties force deeper processing and stronger memory.
- Automated answers bypass the cognitive work of synthesis.
- Illusion of competence arises when information is stored but not understood.
Pilots, Surgeons and Programmers
Aviation history offers a stark warning. In 2009, Air France Flight 447 crashed after pilots failed to recover from a stall. The aircraft sensors froze, disabling the autopilot. The crew had relied on automation for years. They lacked the manual flying skills needed to react. Their experience was valid on paper, but their muscle memory had atrophied.
Surgeons face a similar challenge with robotic systems. The da Vinci robot allows for precise incisions. However, it filters out the tactile feedback of tissue resistance. Young surgeons trained exclusively on robots cannot feel the difference between healthy and damaged tissue. Experienced surgeons who transitioned later retain the sense of touch. The difference lies in the years of manual practice.
Software development shows the same trend. GitHub Copilot generates nearly half of all new code. Junior hiring rates have dropped significantly because companies assume AI replaces entry-level work. Yet debugging and code review were the primary ways juniors learned system architecture. Now they review code they did not write, missing the foundational logic.
Stack Overflow traffic has declined sharply. Developers no longer ask questions because AI provides immediate answers. The community dialogue that once clarified complex concepts has vanished. The collective intelligence of the field relies on this friction. When the struggle disappears, the shared understanding erodes. The profession loses its ability to self-correct.
- Pilot training shows automation leads to loss of manual recovery skills.
- Robotic surgery removes tactile feedback essential for tissue assessment.
- Junior developers miss foundational learning when AI writes the code.
When to Let AI Help
You should not reject AI entirely. The goal is strategic friction. Use AI for tasks that do not require deep understanding. Drafting routine emails, formatting data, or summarizing meeting notes are suitable candidates. These activities benefit from speed without risking the loss of core competencies. The output matters more than the process here.
Avoid AI when the task is the learning opportunity. If you are trying to understand a new framework, write the code yourself first. If you are analyzing a dataset, explore the raw numbers before asking for a summary. If you are preparing a presentation, structure the arguments manually. The struggle ensures you own the knowledge.
Set boundaries for your workflow. Define specific hours or projects where AI assistance is prohibited. Treat these periods as deliberate practice sessions. Just as an athlete trains without equipment to build strength, you train without algorithms to build judgment. This discipline prevents the slow erosion of your professional capabilities.
Monitor your confidence levels. If you feel you understand a topic only because the AI explained it, you are likely in danger. True understanding comes from the ability to explain the concept without aid. Test yourself regularly. Can you solve the problem if the tool goes offline? If the answer is no, you have outsourced your skill.
- Use AI for routine tasks where the process is not the learning goal.
- Perform core learning tasks manually to ensure internalization.
- Test your understanding by attempting problems without AI assistance.
Designing Your Own Practice
Create a personal protocol for skill acquisition. Identify the fundamental skills of your profession that rely on manual execution. For a writer, this might be outlining without suggestions. For an analyst, it might be building a model from scratch. Isolate these skills and practice them daily. Make the friction a habit.
Seek out mentors who emphasize the process over the result. Ask them to critique your manual work before they offer solutions. Encourage them to ask questions rather than give answers. This dynamic replicates the traditional apprenticeship model. It forces you to articulate your reasoning and defend your choices.
Document your failures. When you struggle with a problem, record what went wrong and how you resolved it. This log becomes a repository of tacit knowledge. AI cannot generate these insights because it has not experienced the failure. Your personal history of errors is your most valuable asset.
Reframe efficiency. Do not measure success solely by speed. Measure it by the depth of your understanding and your ability to adapt. A slower process that builds robust skills is more efficient in the long run. The goal is not to finish the task quickly. The goal is to become the person who can handle any variation of the task.
- Isolate and practice fundamental manual skills daily.
- Build a personal log of errors to capture tacit knowledge.
- Measure success by adaptability and depth of understanding.
Key takeaways
- Productive friction is the mechanism that converts effort into lasting professional skill.
- Automation removes the practice required to maintain expertise, leading to deskilling.
- Juniors must intentionally struggle with core tasks to build judgment and tacit knowledge.
- Strategic use of AI requires distinguishing between tasks that need speed and tasks that need learning.
- True competence is measured by the ability to perform without algorithmic assistance.
Related questions
Does using AI for coding prevent me from becoming a good developer?
Using AI for every line of code prevents you from learning the underlying logic of programming. If you rely on AI to generate code without understanding it, you cannot debug effectively or design robust systems. You must write code manually to build the mental models required for expertise.
How can I tell if I am losing my skills due to AI reliance?
You are losing skills if you cannot explain a concept or solve a problem without the AI tool. If you feel confident only when the algorithm is present, you have outsourced your judgment. Test your abilities by attempting tasks in an environment where AI is unavailable.
Should I avoid AI completely to learn my job better?
No, avoiding AI entirely is unnecessary and inefficient. Instead, use it selectively for routine tasks while reserving core learning activities for manual practice. The goal is to balance speed with the deliberate struggle required to build deep professional judgment.
AI-assisted: drafted with a language model from the author's research notes, then edited and checked before publication. Author: Zoltán Varga, AI visibility researcher and AI ecosystem architect, Budapest.
