The Comfortable Crutch: Dependency and the Slow Erosion of Skill at Work

What do we quietly surrender when a tool begins to think for us? Can competence survive the comfort of constant assistance? And if skill is a muscle, what happens to the body that stops exercising it?

There is a particular kind of ease that arrives with powerful tools. At first it feels like liberation. The heavy cognitive lifting, the painstaking drafting, the careful calculation that once demanded full concentration, now happens almost instantly. The machine obliges, the task is completed, and somewhere in the background, barely noticed, a small light dims in the mind of the worker. This is not a dramatic collapse. It is quieter than that, and therefore more dangerous.

The phenomenon has a name in cognitive psychology: skill atrophy, sometimes framed within the broader concept of cognitive offloading. When we delegate mental effort consistently to an external system, the neural pathways associated with that effort are used less frequently. Like a path through a field that no one walks any more, they gradually disappear into the grass. The research of psychologists who study human-automation interaction has long suggested that pilots who rely heavily on autopilot systems can lose the fine-grained manual skills needed in emergencies. The parallel with knowledge workers and artificial intelligence is not a distant metaphor; it is an immediate and practical concern.

Consider what happens in a typical professional context. A writer who once wrestled with a blank page, feeling the productive discomfort of searching for exactly the right structure, now prompts a language model and receives a polished draft within seconds. A junior analyst who might have spent an afternoon learning to interpret ambiguous data now accepts a summary generated by an AI system without interrogating its assumptions. A manager who once composed careful, considered communications now edits machine-generated text rather than constructing original thought. Each of these is, in isolation, a reasonable efficiency. Taken together, over months and years, they represent a slow and cumulative withdrawal from the very practices that build expertise.

What makes this erosion so insidious is that it rarely feels like loss. It feels like progress. Productivity metrics may even improve in the short term. The danger is concentrated not in what is measured but in what is not: the depth of understanding, the capacity for independent judgement, the resilience of the practitioner when the tool is unavailable or, more critically, when the tool is wrong. Because tools are wrong, often in ways that are difficult to detect without the very expertise that has been quietly hollowed out.

This is the paradox at the centre of the comfortable crutch. The more we rely on a system to perform a cognitive task, the less capable we become of verifying whether it has performed that task well. An experienced writer can read a generated draft and sense where the logic is thin or the tone is misaligned. A novice writer, or an experienced writer who has spent years offloading the craft, may no longer possess that felt sense of quality. The crutch undermines the very capacity needed to evaluate the crutch.

It would be easy, at this point, to slide into the familiar rhetoric of technophobia, to argue that we should resist these tools or refuse them altogether. That argument is neither realistic nor, in the end, necessary. The question is not whether to use AI at work, but how to use it in a way that preserves the cognitive engagement which keeps expertise alive. There is a meaningful difference between using a calculator after one has understood the mathematics, and using it instead of understanding the mathematics. The former is a legitimate extension of capability; the latter is a substitution that leaves the mind impoverished.

The philosopher Michael Polanyi wrote extensively, though we shall paraphrase rather than cite directly, on the idea that skill contains a dimension of tacit knowledge: knowing that cannot be fully articulated, that lives in practice, in repetition, in the accumulated experience of doing. This kind of knowing cannot be outsourced, because it exists precisely in the act of doing. When we stop doing, we stop knowing, even if we remain unaware of the loss. The professional who has not written a difficult letter without assistance for two years has not simply saved time; that professional has quietly surrendered something that belongs to the deeper architecture of expertise.

Keeping the mind in practice requires deliberate effort, and this is perhaps the most important insight. It means choosing, sometimes, the slower path. It means treating certain tasks as practice rather than merely as output, recognising that the value of a task is not only in its completion but in what the process of completion builds in the person completing it. Organisations that care about the long-term capability of their people will need to think carefully about when AI assistance serves development and when it short-circuits it. They will need to ask not only whether something was done efficiently, but whether the person doing it emerged from the process more capable than before.

Individuals, too, bear responsibility here. There is something to be said for the intentional practice of unassisted work: the deliberate exercise of a skill without the crutch, not out of stubbornness but out of respect for the mind and what it can do when it is given the chance. A musician who performs only to a backing track loses something essential. So does a thinker who thinks only with a machine.

So we return, as we must, to questions. How often do we pause to notice what we are no longer doing for ourselves? How would we fare if the tools we depend on were taken away for a month, a week, even a day? And perhaps most importantly: what kind of worker, thinker, and professional do we wish to become, and are the habits we are forming now actually moving us in that direction?