Counterintuitively, the rise of AI makes people who know a little about everything harder to replace than those who know everything about one thing.
A senior financial analyst at a mid-size asset manager recently described a meeting that stuck with her. A portfolio manager asked whether a proposed acquisition would trigger any material HR compliance issues in the target company’s German operations. The analyst knew the numbers cold. She had no idea about the HR question. Neither did the lawyer on the call, who handled securities, not employment. The HR lead wasn’t in the room.
The meeting stalled. Not because the information was unavailable — it was, somewhere — but because no one present had enough context across enough domains to even know what the right question was, let alone who should answer it.
This is the gap that AI is quietly exposing. Not the gap between what we know and what machines know. The gap between how work is actually structured — in silos, by specialty — and how problems actually arrive — messy, cross-functional, and impatient.
There is a reasonable fear that AI will hollow out specialist roles. Feed a contract into a large language model and it will flag the problematic clauses. Ask it about accounting standards and it will cite them accurately. Run your marketing copy through it and it will score readability, suggest A/B variants, and check brand tone.
That fear is not entirely wrong. Routine specialist work — the kind that involves retrieving, applying, and formatting domain knowledge — is genuinely under pressure. A junior lawyer spending six hours reviewing boilerplate NDAs is doing work that AI can accelerate by an order of magnitude. A financial analyst building a standard DCF model from a template is operating in territory that AI handles competently and quickly.
But here is what that observation misses: AI is a tool that executes within a defined frame. Someone has to set the frame. Someone has to know that the German HR question belongs in the acquisition conversation at all. Someone has to recognise that a marketing campaign, however legally compliant, carries reputational risk the legal team won’t flag because it’s not their job to think about brand.
That someone is, almost always, a generalist.
As AI raises the output capacity of individual specialists, organisations face a less obvious problem: more high-quality work being produced in parallel, with less human bandwidth to integrate it.
An operations director at a logistics firm described this precisely. Her company adopted AI tools across procurement, finance, and customer service within eighteen months. Each team became measurably faster. But the number of cross-functional conflicts — procurement decisions that created customer service problems, finance constraints that hadn’t been flagged to ops until too late — increased, not decreased. Speed without coordination produced faster mistakes.
The people who became most valuable in her organisation weren’t the ones who mastered the AI tools fastest. They were the ones who could walk between teams and translate. Who understood enough about procurement to spot the downstream implication, enough about finance to know what the constraint actually meant, enough about customer expectations to ask the right question before a decision calcified.
This is not a new observation about generalists. It is a newly urgent one.
Consider a few scenarios that are already playing out in professional environments.
An HR business partner at a technology company uses AI to generate compensation benchmarking reports that previously took a specialist analyst two days. She now has that data in an hour. The question is what she does with the next seven hours. If she is only an HR specialist, she interprets the data through an HR lens and writes an HR recommendation. If she understands enough about the business unit’s commercial pressures — headcount constraints, revenue targets, the product roadmap — she produces something different: a recommendation that finance will actually accept and that the business lead will actually implement. Same data, different value.
An IT project manager overseeing an ERP rollout uses AI to summarise vendor documentation and flag integration risks. But the AI cannot tell him which risks the CFO will care about versus the ones the operations team will deprioritise. That requires knowing enough about how each stakeholder thinks to translate technical risk into the language of business consequence. That translation is not a technical skill. It is a generalist skill.
A marketing strategist uses AI to generate ten campaign concepts in an afternoon. But selecting the right one requires judgment about brand positioning, legal exposure, sales team capacity, and timing relative to a product launch. The AI can model; it cannot weigh. Weighing across domains is what generalists do.
None of this means that depth is obsolete or that being a generalist is automatically an advantage. A shallow generalist — someone who knows just enough about everything to be confidently wrong — is not more valuable in an AI-augmented environment. They are more dangerous. The breadth has to be real enough to ask good questions, spot misfits, and know the limits of your own knowledge.
The useful generalist is not someone who avoids depth. They are someone who has developed genuine understanding in two or three domains and has built the habit of thinking across them. A lawyer who genuinely understands finance. An HR professional who has spent real time in operations. A marketer who can read a P&L without panicking.
That combination — domain literacy plus connective thinking — is precisely what AI cannot replicate, because AI has no stake in the outcome and no map of the organisation’s actual politics, history, and constraints.
Here is the insight worth sitting with: AI doesn’t make knowledge less valuable. It makes isolated knowledge less valuable. What increases in value is the ability to make knowledge useful across contexts — to carry it into rooms where it wasn’t expected, to connect it to something the specialist in the corner didn’t think to mention.
The generalist’s core skill was always synthesis. AI has just made synthesis the scarcest thing in the room.