Why a Consultant in a GenAI World
Yes, I use AI. So does every consultant worth engaging.
The question isn't whether your advisor uses Claude or Copilot to help structure a document. Of course they do. The question is what they bring to the conversation before the AI touches it — and what judgment they apply after.
AI is genuinely good at drafting. It synthesises, structures, and produces coherent prose faster than any human. That part of consulting — the document production, the research compilation, the template population — is largely automated. Anyone telling you otherwise is either not using the tools available or not being straight with you.
The work doesn't disappear. It concentrates upstream and downstream of the AI.
Upstream is where the fee is earned. Knowing what question to actually ask. Understanding what the client's real problem is beneath what they've articulated. Providing the context the AI needs to produce something useful. AI cannot frame problems it doesn't understand exist. That requires pattern recognition built from years of domain experience.
In the middle sits judgment — the part that is hardest to see but most consequential. AI is good at summarising what's there. It is poor at knowing what's missing, what's contradictory, and what should make you uneasy. An experienced consultant reading a governance structure or a vendor contract can sense something is off before they can fully articulate why. That instinct is experience-encoded. It is not replicable by a model trained on generic text.
There is a layer of every engagement that never makes it into a brief. Navigating that requires presence, not processing power.
Clients rarely hand you the real problem. It surfaces in conversation — obliquely, gradually, sometimes only once trust is established. That kind of listening is not a skill you can prompt.
Downstream is accountability. Someone has to own the output. AI produces confident-sounding prose that can be subtly wrong, contextually inappropriate, or missing the thing that matters most. The human in the chain is the one whose reputation is on the line. That accountability cannot be delegated.
And underneath all of it is trust. Clients don't hire consultants for documents. They hire them for the confidence that someone capable and trustworthy is watching their back. That is entirely human.
The pyramid problem
The traditional large consultancy model — one partner, a team of graduates — is built on labour arbitrage. The partner sells on experience and relationships. The graduates do the production work. The economics depend on that pyramid.
GenAI just became the most cost-effective graduate in the room. It doesn't get tired, doesn't need onboarding, produces a solid first draft at 2am, and doesn't bill by the hour. The production layer of that pyramid is under pressure — not eventually, now. The large firms know it, which is why graduate intake is quietly being restructured across the industry.
The independent model is different. It was never built on a production pyramid. The economics were always based on delivering partner-level thinking directly, without the overhead.
That model doesn't just survive the AI transition. It gets stronger.
GenAI fills the analyst role without the cost, which improves the economics and sharpens the value proposition. The client gets senior judgment, backed by a research capacity that simply wasn't available at this scale before, delivered without paying for the layers in between.
The deliverable has always been a means to an end. What you are buying is the thinking behind it — and that has never been more clearly true than it is now.
The Future of Knowledge Workers — A Practical View
The conversation about AI and knowledge work tends toward one of two extremes — either breathless enthusiasm about productivity gains, or anxiety about displacement. Neither is particularly useful.
What is actually happening is more interesting and more gradual than either narrative suggests.
The mechanical layer of knowledge work is being automated. Drafting, summarising, reformatting, compiling — tasks that consumed real hours — are now largely handled by AI tools that most workers already have access to. That time is recovered. That is genuinely good.
What this does is concentrate human effort where it has always mattered most, but where it was previously diluted by the surrounding production work.
The knowledge worker of the near future spends less time producing and more time deciding. Less time structuring documents and more time determining what the document actually needs to say and why. Less time researching what is publicly knowable and more time interpreting what it means for this organisation, this situation, this decision.
This is not a trivial shift. It raises the cognitive bar. Workers who competed primarily on thoroughness — who could research harder and draft faster than their peers — will find that advantage compressed. Workers who compete on judgment, domain expertise, and the ability to operate in ambiguous situations will find their value amplified.
The workers most at risk are not those being replaced by AI. They are those who fail to move up the value chain — who continue to position themselves around output production rather than interpretive thinking.
The practical implication for organisations is equally significant. If your knowledge workers are primarily occupied with tasks AI can now perform, you have a workflow problem, not a headcount solution. The opportunity is to redirect that capacity toward higher-order work — the analysis, the judgment calls, the relationship management — that has always driven real outcomes but never had enough human attention.
AI raises the floor for everyone. What it cannot do is replace the ceiling.
The future of knowledge work is not fewer workers. It is workers operating at a consistently higher level — with less tolerance for those who don't.