OP-ED. If Expertise Is One Prompt Away, Are We All Experts Now?

Someone approaching an unfamiliar topic can, within a remarkably short time, move from basic principles to highly specialized knowledge. Along the way, it becomes possible to ask for explanations, challenge an answer, compare alternatives, request examples and follow a line of reasoning deeper into a subject. When something is too complex, we can continue asking questions until the concept is explained in terms we understand. What once required knowing where to look, which sources to consult and often whom to ask can now begin with a single prompt.

This does not make us experts. It does, however, change something fundamental: access to specialized knowledge is no longer the privilege it once was. That raises an uncomfortable question: What makes an expert when knowledge is available to everyone?

Broader Access to Specialized Knowledge

Does the democratization of knowledge mean the end of expertise as we know it? Generative AI has significantly lowered the threshold for accessing specialized knowledge. Does broader access actually reduce the value of expertise, or does it merely change where that value lies?

Research suggests that generative AI can indeed narrow some of the gaps traditionally associated with experience and access to knowledge. Brynjolfsson et al.¹ found particularly strong benefits among less experienced and lower-skilled workers, helping them move faster along the experience curve. Dell’Acqua et al.² found a similar effect, while also showing that performance could deteriorate when tasks fell outside AI’s capabilities. Zhu and Walker³ identified another limitation: AI can lower barriers to knowledge production while its benefits remain unevenly distributed.

The evidence therefore suggests a more complex picture: generative AI may democratize access to knowledge and some of its practical application without necessarily democratizing expertise. If access to knowledge is no longer what separates the expert from the non-expert, what does?

In my own professional environment, I have noticed a paradox. Some of the strongest resistance to generative AI comes from highly experienced specialists. These are people with deep technical knowledge and years, sometimes decades, of practical experience.

It would be easy to interpret this simply as resistance to change. Such an explanation may be too convenient. What if some of these experts are sceptical precisely because they can see what AI does not know? What if experience allows them to recognize missing context, questionable assumptions or technically plausible answers that would not survive contact with industrial reality? This is only an observation, not evidence of a broader pattern. It nevertheless points to a question worth investigating: Could the people most skeptical of AI possess precisely the kind of expertise needed to use it responsibly?

Agentic AI Gets to Work: Are Manufacturers Ready to Let AI Act?

From Knowing More to Seeing Differently

Research on expertise suggests that the difference between experts and novices is not simply how much they know. Chi et al.⁴ found that experts and novices approached the same physics problems differently: while novices tended to focus on surface features, experts organized problems around the underlying principles needed to solve them. In other words, the difference lay not only in what they knew, but in how they saw and structured the problem.

I recently encountered a small example of this in my own work. Unusual black smut was appearing around MIG welds on an aluminium component. One obvious question was: What causes black smut in aluminium MIG welding? Today, AI can answer that question within seconds and provide a long list of technically plausible causes.

There was another observation: comparable TIG welds on the same component remained clean. The base material and filler material had the same chemical composition. The shielding gas was the same. The more interesting question was therefore no longer What causes black smut? It was: Why does it appear in one welding process but not in the other when several of the surrounding conditions are the same?

Anyone familiar with welding knows that MIG and TIG are fundamentally different processes. Knowing those differences, however, is not the same as knowing which of them could explain this particular observation. The next question therefore becomes more specific: Which process-specific differences could cause black smut to form in MIG welding while remaining absent, or significantly reduced, in TIG welding?

At this point, AI can suggest relevant parameters and mechanisms. An expert can continue the investigation: What physical phenomena do those parameters change? Each answer creates a more precise question, points towards possible mechanisms and helps eliminate explanations that do not fit the observation. The difference is important. The first question asks for knowledge. The questions that follow begin to structure the problem. Before an answer becomes useful, someone has to recognize which observation changes the question.

The Limits of Experience

However, experience does not automatically produce reliable judgment. Kahneman and Klein⁵ argued that expert intuition becomes reliable when the environment contains sufficiently predictable patterns and when experience provides opportunities to learn those patterns through feedback. Bilalić et al.⁶ demonstrated another risk: prior expertise can create cognitive fixation, directing an expert’s attention toward familiar solutions even when better alternatives exist.

Expertise, then, may lie less in possessing knowledge than in recognizing what matters, what does not fit and what deserves to be questioned. In the age of AI, that distinction becomes particularly important. The value of an expert may therefore lie less in having access to an answer and more in being able to judge whether that answer makes sense in the first place.

Deep knowledge does not become irrelevant simply because access to it becomes less exclusive. As access becomes easier, other dimensions of expertise become more visible: recognizing the right problem, asking questions that are not obvious, understanding context and knowing when an apparently convincing answer should not be trusted.

The easier it becomes to obtain an answer, the more important it becomes to know what to do with it. Perhaps that is where the distinction between having knowledge and having expertise becomes clearest.