As AI becomes part of everyday work, organizations face a new challenge: distinguishing genuine expertise from the increasingly convincing appearance of it.
A recent conversation with one of my colleagues left me thinking about a challenge that I suspect many organizations will begin facing more frequently over the next few years.
We were discussing an engagement where an individual had spent months advising a client on an initiative that appeared, at least from the outside, to be progressing well. Meetings were taking place regularly, recommendations were being presented with confidence and documents were being produced. The conversations sounded thoughtful and sophisticated.
Yet months later, very little meaningful progress had been made, and as we reflected on what had happened, one observation stood out. The problem was that the person leading the initiative had begun relying on AI as the primary source of expertise while lacking the experience required to challenge its recommendations, adapt them to the client's context, or recognize when they simply did not apply.
This raised a question that has stayed with me ever since. As AI becomes increasingly capable of producing articulate, convincing, and technically sound responses, how do organizations distinguish genuine expertise from someone who has simply become very good at asking AI the right questions?
I don't believe we have a good answer yet, and that uncertainty is important because I think many of us are focused on just one side of AI.
Much of the conversation surrounding AI has focused on how quickly it can generate content, write code, summarize research, or automate repetitive work. Those are remarkable capabilities, and I believe AI will become one of the most transformative productivity tools of our generation.
The more interesting question, however, is not whether AI can generate answers, but whether the person using those answers understands them well enough to evaluate their quality.
Over the years, one of the ways we naturally identified expertise was through knowledge. Experts knew things that others did not. They had accumulated years of experience, learned from failures, developed intuition, and understood the tradeoffs that rarely appear in textbooks or best-practice guides.
AI fundamentally changes that dynamic because almost anyone can now generate a polished explanation of an unfamiliar topic within seconds. Technical language, structured frameworks, and convincing recommendations are no longer reliable indicators of expertise because they have become increasingly accessible to everyone.
I think this requires us to distinguish between two very different ways of working with AI.
The first is what I would describe as expert-led AI, where the human remains the expert and AI functions as an accelerator. Experience provides the context, judgment determines which recommendations make sense, which assumptions should be challenged, and which ideas should be discarded, while AI helps organize information, explore alternatives, identify blind spots, and increase productivity without ever replacing the thinking itself.
The second model, which I would describe as AI-led human, reverses that relationship. Instead of using AI to strengthen existing expertise, the individual begins relying on AI as the primary source of knowledge. Recommendations are accepted with little critical evaluation, technical language replaces understanding, and confidence begins to substitute for competence. The person is no longer directing technology; they are following it.
At first glance, the difference between these two approaches can be remarkably difficult to detect. Both individuals may speak confidently, produce polished presentations, reference industry frameworks, establish best practices and even appear equally convincing during meetings.
The difference often becomes visible only when the conversation moves beyond general recommendations and into real-world complexity. Experienced professionals rarely believe there is a single correct answer. They talk about tradeoffs. They explain why one approach may work well in one organization but fail in another. They acknowledge uncertainty, discuss competing priorities, and adapt recommendations as new information emerges.
Someone relying primarily on AI, by contrast, may deliver recommendations that sound entirely reasonable until they encounter a situation that requires judgment rather than information. That distinction has become increasingly important to me, and I don't think the solution is becoming more skeptical of AI. If anything, I believe organizations should and will continue embracing it. What I do believe will help is becoming better critical thinkers. AI is remarkably effective at generating possibilities. It is far less capable of understanding the unique organizational dynamics, political realities, historical context, and strategic tradeoffs that influence almost every important business decision. Those remain fundamentally human responsibilities.
Perhaps the greatest risk organizations face is not that AI will provide incorrect answers, but that we begin mistaking fluency for expertise. That possibility raises another uncomfortable question. How should leaders evaluate expertise in a world where everyone has access to the same extraordinary technology?
I honestly don't know and I don't yet have a reliable framework for distinguishing genuine experts from individuals who have simply become exceptionally good at presenting AI-generated thinking. I suspect many teams are wrestling with the same uncertainty as AI becomes more deeply integrated into everyday work.
What I do believe is that critical thinking has never been more valuable. Leaders should be asking more questions, challenging assumptions more frequently, and exploring recommendations from multiple perspectives before turning them into decisions. AI has dramatically reduced the cost of producing answers, but it has done nothing to reduce the importance of evaluating them.
Technology will continue evolving, and so will the ways we use it. The bigger question, at least in my mind, is whether we will become equally disciplined in evaluating the recommendations it produces. I don't think we have a clear answer yet, but I do think this is becoming one of the most important leadership conversations of the AI era.