Fools’ Gold … the deceptive allure of easy answers.
Four findings from inside consultants' working lives — and what they reveal about the future of knowledge work
Executive Summary at the bottom of the article
AI can now summarize documents, analyze transcripts, build first-draft frameworks, compare options, draft recommendations, produce slides, generate ideas and create plausible arguments. It can generate the visible artefacts of knowledge work at extraordinary speed. So, the question is no longer whether AI can make us more productive. It can. The more important question is what happens to human value when machines become increasingly good at producing the outputs that used to signal expertise.
That question sits at the heart of Adam’s academic work, looking at how generative AI is reshaping cognition, judgement and expertise in knowledge work. Some of his recent research focused on management consultants, partly because consulting is such a useful early-warning system, and partly because that is a world we are part of. Consultants operate at the intersection of analysis, interpretation, persuasion and commercial judgement. Their value depends not only on what they know, but on how they frame problems, synthesize complexity and help clients make decisions. If AI changes consulting, it is likely to change many other knowledge-based roles too.
This recent research explored how consultants are actually experiencing AI in their day-to-day work. Not the corporate press release version. Not the shiny “AI transformation” keynote version. The lived experience of working with a technology that is useful, unsettling, empowering, tempting and professionally destabilizing all at once. Four findings stood out.
The first is that AI is not simply being used to save time. It is becoming a scaffold for thinking. Consultants are using it to structure unstructured problems, summarize large amounts of material, create first drafts, identify gaps, generate hypotheses and provide alternative perspectives. In that sense, AI is not just sitting at the end of the work process as a production tool. It is moving upstream into the formation of thought.
This is an important distinction. When AI is used well, it can help people think. It can provide a starting point, create useful friction, challenge a frame, offer a different angle, or help someone step back from the detail and see a broader pattern. But there is a thin line between scaffold and substitute. A scaffold supports the human thinker. A substitute replaces the human thinker. That distinction is becoming one of the most important questions in knowledge work. Are we using AI to extend our thinking, or to avoid it? Are we becoming more reflective, or simply more efficient? Are we using the machine to provoke judgement, or to bypass judgement? The answer depends less on the technology than on the human habits around it.
The second finding is that AI increases the need for judgement rather than reducing it. This runs against much of the productivity narrative. We often talk as though AI reduces cognitive effort. It does some of the work, so the human has less to do. But in practice, good AI use requires constant evaluation. Is the answer right? Is it relevant? Is it complete? Is it biased? Is it generic? Is it beautifully answering the wrong question? Is it smoothing over the very ambiguity that matters?
AI produces fluency very easily. That is both its power and its danger. A bland recommendation can feel authoritative because it is presented with confidence. This means the human role shifts. The knowledge worker becomes less of a producer and more of an evaluator, editor, challenger and interpreter. That is not a lesser role. It is arguably a more demanding one, but it requires a different kind of capability. Not just technical competence. Not just prompt craft. Not just “AI literacy”. It requires metacognition, which is the ability to think about your own thinking whilst working with a system that is also shaping the thinking process. That is a much higher bar than most AI training currently recognizes.
The third finding is perhaps the most interesting: AI use is not experienced as emotionally neutral. There is excitement, of course. People are impressed by what it can do. They value the speed. They can see how it helps them produce better outputs faster. But there is also unease. Some of that unease is practical … can I trust this? What has it missed? Where has the answer come from? And some of it is existential … what is my value if the machine can do this work? What happens to my role? What happens to junior people? What happens to the structure of the firm?
And then (perhaps most interestingly) … some of it is moral. There is a sense of guilt around using AI to do work that previously required effort. Not because people think using AI is wrong in itself, but because effort has always been part of the DNA of professional work. Professionals are paid not merely for outputs, but for earned expertise. They have struggled, learned, practised, failed, improved and accumulated judgement over time. So, when AI produces a strong answer in seconds, it can create a strange discomfort – the feeling that I am claiming credit (and billing) for thinking I did not really do.
That discomfort should not be dismissed as technophobia. It may be a signal of professional integrity. People are sensing that something important is being renegotiated … the relationship between effort, expertise, value and trust. The commercial risk is that organizations ignore this. They treat hesitation as resistance, when it may actually be a form of quality control. The people who feel discomfort may be the ones most alert to where AI use could damage judgement, learning or client trust.
The fourth finding is that AI threatens the traditional learning model of consulting. Consulting firms have long depended on a pyramid structure. Junior people do much of the research, analysis, modelling, note-taking, synthesis and slide production (most of the billable hours). Managers review and shape the work. Senior people, at the pinnacle, steer the client relationship, provide judgement and carry accountability. AI puts pressure on the base of that pyramid. If AI can do more of the junior work, firms may need fewer juniors. That may look commercially attractive in the short term. Lower cost. Faster work. Leaner teams.
But the base of the pyramid is not just a cost layer. It is also the apprenticeship system. Junior consultants learn by doing the work. They learn by wrestling with data, drafting badly, being corrected, sitting in meetings, building slides, making mistakes, seeing patterns and gradually understanding what good judgement looks like. They learn by osmosis. If AI removes too much of that struggle, where will future senior judgement come from?
This is not only a consulting issue. Every knowledge function has its own apprenticeship pathway. Junior researchers learn by reading and coding interviews. Junior analysts learn by cleaning and interrogating data. Junior marketers learn by drafting, testing and refining. Junior lawyers learn by reviewing documents. Junior strategists learn by wrestling with messy uncertainty. If AI takes over formative work, organizations may accidentally hollow out the development of expertise. The work that looks inefficient, to be replaced by AI systems, may be the work through which expertise is formed.
Taken together, the research points to a bigger shift. Human value is moving from analysis to orchestration. That does not mean analysis disappears. Nor does it mean humans no longer need analytical skills. Quite the opposite. You need enough expertise to know whether the AI-generated analysis is any good. But as AI becomes more capable of producing summaries, structures, comparisons and first-draft recommendations, the premium shifts elsewhere. It shifts to framing the right question, understanding context, interpreting ambiguity, knowing what matters, spotting what is missing, making judgement calls under uncertainty, communicating meaning in a way that creates action, and carrying accountability for the advice being given.
That is true in consulting, but it is also true across knowledge work. The lawyer, researcher, strategist, marketer, analyst, product manager and policy adviser all face a similar question. If AI can produce the first version of the work, where does my value now sit? The answer is not in being better than AI at producing content. That is a losing game. The answer lies in becoming better at the human side of the human-AI equation.
This is where we think many organizations are making a strategic mistake. They are investing heavily in artificial intelligence while underinvesting in the human intelligence required to use it well. They are upgrading the machine side of the equation while assuming the human side can remain largely as it is. That is not sustainable or optimal. If the future of work is moving toward something closer to collective intelligence — humans and machines thinking together — then the quality of that collective intelligence depends on both sides of the relationship.
Better AI will not automatically produce better organizational thinking. It may produce faster mediocre thinking, faster groupthink, faster shallow answers, faster overconfidence and faster strategic drift. The competitive advantage will come from the capabilities wrapped around the tools … curiosity, judgement, critical thinking, creativity, sensemaking, communication and leadership.
This is the central argument behind PolymathMind. The future does not belong to people who try to compete with machines on machine terms. Nor does it belong to people who refuse to change. It belongs to people who can combine human and machine intelligence in more deliberate, reflective and powerful ways. That requires a new emphasis on human development, not as a sentimental add-on but as strategy.
So what should leaders do? First, stop treating AI adoption as a technology program. AI adoption is not just about tools, platforms and workflows. It is about changing how people think, decide and create value. If your AI strategy does not include a human capability strategy, it is incomplete.
Second, define the boundary between scaffold and substitute. Organizations need clearer norms around when AI should support thinking and when humans must think first. The same tool can either develop capability or erode it. Leaders need to make that distinction explicit before the default drifts toward substitution.
Third, train judgement, not just prompting. We have nothing against prompt training … it is useful, but insufficient. The more important skill is knowing whether the output is good. That requires domain knowledge, critical thinking, contextual awareness and the confidence to challenge fluent answers.
Fourth, protect the apprenticeship pathway. Do not automate junior work without asking how people will now learn. Some tasks may need to be preserved, redesigned or deliberately practised because they build the judgement future experts will need.
Finally, reframe value around orchestration and accountability. In a world of abundant AI-generated output, value moves to the human capabilities around it: framing, synthesis, judgement, communication, ethics and action. That is what clients, customers and organizations will increasingly pay for.
AI is forcing a difficult but necessary conversation about the future of knowledge work. For years, many organizations have treated human capability as something vague, secondary or difficult to measure. The hard stuff was technology, process and data. The soft stuff was communication, creativity, curiosity and judgement. That distinction is breaking down. In an AI-shaped workplace, the so-called soft skills become the hard edge of advantage.
The machine side of the equation is developing at extraordinary speed. The human side now needs the same level of seriousness, investment and ambition. Because the future will not be won by organizations that simply use more AI. It will be won by organizations that become more intelligent because of it.
Executive summary
AI is not simply being used as a productivity tool. It is becoming part of the thinking process itself.
Consultants are using AI as a cognitive scaffold: a way to structure problems, create starting points, provide perspective and accelerate delivery.
But the same technology also creates discomfort: concerns about over-reliance, weakened judgement, cognitive atrophy, professional guilt and the hollowing out of expertise.
The most important shift is from analysis to orchestration. As AI takes on more analytical work, human value moves toward framing, judgement, sensemaking, contextualization and accountability.
This matters far beyond consulting. Any knowledge-based function that depends on analysis, synthesis, expertise or communication will face the same challenge.
The commercial implication is clear: organizations cannot invest only in AI capability. They must also invest in the human capabilities needed to work intelligently with AI.