Almost everything measured about AI at work looks at the present: how many hours it saves, how many tickets it resolves, how much faster a draft comes out. There is a cost none of those metrics registers, and it is worth naming before it becomes irreversible: what happens to your people’s skill when AI takes off their plate, precisely, the hard tasks through which they learned the craft.
Matt Beane studied it in operating rooms and warehouses (The Skill Code, 2024). The path by which a novice becomes an expert rests on three conditions —what he calls the three C’s: challenge (facing tasks at the limit of one’s ability), complexity (seeing the whole problem, not a slice) and connection (the bond with an expert who corrects in context)—. When an intelligent system absorbs the entry-level tasks —the “easy” ones that were really the practice ground—, all three weaken at once: the novice is no longer challenged, no longer sees the whole problem, and stops needing the expert. Productivity rises today; the training of tomorrow’s expert goes dark with no one measuring it.
Erosion of the scaffold
I call this the erosion of the scaffold: the silent decay of the path by which an organization forms its own experts. It is not automation unemployment —that is seen and debated—; it is a loss of future capacity that the productivity metric does not register, because that metric improves while the scaffold falls. It is the second, quieter floor of a more familiar fear: the expert who feels AI discredits what they know. Here the problem is even harder to see, because no one complains: the experts who would come next simply stop being born.
A field experiment makes it tangible from another angle. Dell’Acqua et al. (2025), with 776 professionals at Procter & Gamble, found that individuals with AI matched the performance of teams without AI, and that the tool broke functional silos —R&D and Commercial people producing balanced solutions, as if AI lent them the other side’s vocabulary—. It is powerful. But the same study saw something worth underlining: AI covered part of the social role of the teammate, motivating without correcting. A teammate who motivates but does not train is, once again, scaffold that erodes: Beane’s connection is precisely what a team provides and a lone individual with AI does not.
What to do differently (it is design, not destiny)
The good news is that the erosion of the scaffold is not an inevitable effect of AI: it is a design risk. You can intervene without giving up productivity:
- Reserve learning tasks. Deliberately decide which tasks are kept for the person even though AI could do them, because their value lies not in today’s output but in tomorrow’s skill.
- Keep the novice exposed to the whole problem. Do not give them only the slice AI does not cover: make sure they keep seeing the whole, which is where judgment is formed.
- Protect the expert-novice bond. AI can motivate; it cannot correct in context. That bond is an asset, not a cost to optimize away.
- Measure what productivity hides. If experts stop being formed, the scaffold is falling —and no usage metric will warn you—.
AI well placed can level up: journal evidence shows the least experienced workers gain most when well assisted (Brynjolfsson, Li and Raymond, 2025). But “assisting” and “replacing the practice ground” are not the same. The difference between the two —between an AI that trains and one that dims— is not decided by the model: it is decided by the design of how it is brought into the work. That is, exactly, the bridge’s job.
The framework is developed in The sociotechnical bridge and in My approach. Related essays: Your team already adopted AI without you · SME maturity is not a straight line.