10,000 people a day are retiring with your operating manual

There is a number that ought to be reshaping every operations budget in the country, and mostly is not. Every single day, roughly 10,000 experienced professionals retire. They leave with decades of operational judgment that exists in no system, no manual and no dataset; only in their heads. Tomorrow another 10,000 go.

Meanwhile, in the US alone, 1.9 million operational roles sit unfilled. The cost of that gap to legacy industries runs to an estimated $1.4 trillion a year in operational inefficiency. Those two facts are usually discussed separately, as a hiring problem and a productivity problem. They are one problem, and it has one honest conclusion: no amount of recruiting closes it.

Why hiring cannot close the gap

Consider the arithmetic. To fill 1.9 million roles you need 1.9 million people who want operational work in legacy industries, and you need them at a wage those industries can carry. Then you need to train each of them, which in practice means pairing them with an experienced operator, the exact resource that is disappearing at 10,000 a day. The constraint is not headcount. It is the supply of expertise available to train headcount.

This is why the shortage behaves differently from a normal labour cycle. A cyclical shortage resolves when wages rise. A shortage driven by the retirement of tacit knowledge does not resolve, because the thing in short supply cannot be manufactured quickly at any price.

What actually leaves when someone retires

Not the process. The process is documented, more or less. What leaves is the reasoning underneath it: which exceptions genuinely matter and which look urgent but are not, what “resolved” means for this customer as opposed to that one, which carrier will answer at 6pm, which rule exists for a real reason and which is merely a habit nobody has questioned.

Ask your best operator to write it down and you will get a thin sketch that misses everything load-bearing. That is not a failure of effort. It is the defining property of tacit knowledge: the people who have it cannot fully articulate it, because the rules sit below the level of language. Two decades of process documentation and RPA projects ran into this wall and mostly lost.

Why the current wave of AI is not catching it either

95% of GenAI pilots deliver zero P&L impact. The usual explanation is that the models are not good enough yet. That is not what the evidence suggests. Give a frontier model a real operations job and it will underperform a competent human who has done that job for a year — not because the human is more intelligent, but because the human knows things the model was never told.

Most teams building agentic systems for operations are working on how the work gets done: mapping tickets, scripting workflows, wiring integrations. That produces automation that breaks the first time a process changes. The harder and more valuable problem is why the work gets done that way — the decision-making behind each action. Capture that and you get an operator. Skip it and you get a faster dashboard.

The window is closing, and it is dated

Most strategic problems have no deadline. This one does, and it is set by a demographic curve rather than a competitor. Every operator who retires before their expertise is captured takes an irreplaceable asset with them. An organisation that captures it converts a retirement risk into something it owns permanently — inspectable, correctable, and still working at 3am.

That is the case for treating expertise capture as urgent operational work rather than a future IT project. The cost of waiting is not measured in software licences. It is measured in people who left.

The most experienced workforce in history

Here is the more optimistic reading of the same numbers. Expertise that has been codified does not retire, does not resign and does not need to be re-trained from scratch at the next company. It compounds. Every desk whose reasoning gets captured makes the next one faster to stand up.

Run that forward and you arrive somewhere genuinely new: the most experienced workforce in history, assembled from the accumulated judgment of the people who actually did the work, available to any operation that wants it. The 10,000 a day are a warning. They are also, if you move now, the source material.

Start with one desk

You do not need a transformation programme. Pick the desk where the most knowledge is closest to walking out, and capture that first. Evos deploys an autonomous operator onto a live workflow in under 24 hours, on the systems you already run. Book an assessment and we will tell you which desk to start with.