The assumption that your knowledge lead persists: still valid or quietly expired

Where the assumption comes from

A strategy rarely rests on all available facts. It rests on a select number of assumptions that were sensible at the moment of writing and were not reopened afterward. One of these is often: our knowledge lead persists, because we have been in this trade longer, have seen more clients, have solved more edge cases than whoever is entering now. That assumption came from somewhere. Years of experience among specialists, files and precedents that could not be copied in half a year, a learning curve that newcomers simply had to go through. That was not overconfidence. That was a reasonable reading of how knowledge was built: slowly, in people, over years.

Why it held

The assumption held because knowledge and the person carrying it were barely separable. If a competitor wanted to match your lead, they had to hire people with the same years of experience, or invest those years themselves. Both were slow and visible. You saw a competitor recruiting or building up, and had time to respond. So the lead was not only substantive, it was also a difference in pace that you could track.

What has changed

AI is taking over work, and this does not happen in the same category everywhere. For part of the knowledge tasks, a language or analysis model can handle the first pass itself: recognizing patterns in files, drafting advice based on precedents, summarizing an analysis that used to cost a senior a day. For another part, it works with human oversight: the model produces the proposal, a specialist approves or rejects it and argues why. For yet another part, little changes, because the judgment is too context-dependent or too risky to automate. That division differs by company and by task, and that is exactly the point: it is not a general breakthrough, it is a shift that happens task by task, and companies that map this out per task are currently ahead of companies that still discuss the question at the strategic level.

Where that difference comes from is usually not substantive. It lies in who has already asked the question at the level of the work itself, and who is still asking it at the level of the sector. The same logic applies to the assumption that this work cannot be standardized: what once counted as too specific for automation often turns out, upon testing, to be partly breakable down into repeatable steps.

The consequence for knowledge lead specifically: the difference in pace that you used to be able to track is no longer guaranteed to be visible. A competitor does not need to hire ten senior people to reach your level on a partial task. They only need to recognize the right task as suitable for takeover by AI, and organize the oversight of it properly. They can do this without you seeing it happen, until the result is on the table.

How a management team notices that the assumption has expired

The signals usually do not sit in this quarter's figures. They sit in behavior that you attribute to something else. A quote from a smaller competitor that is substantively just as sharp as what only your senior people used to deliver, and faster. A client asking why an analysis that takes you a week is finished elsewhere in a day. New entrants who, without the years-long learning curve, already speak at your level. Each of these signals can be explained on its own, and that is exactly why the assumption remains unnoticed for so long: there is no single moment at which it visibly breaks, only a series of separate cases that only form a pattern in hindsight.

The underlying question is the same as with the assumption that scale offers protection: an advantage that was once slow to match only becomes unreliable once someone else skips the slow step. With knowledge lead, that slow step is the learning curve, and that step is skipped precisely where a task lends itself to takeover by a model.

What you should measure

To know whether the assumption still holds, it is not enough to ask whether your people are still good. They probably still are. The question is whether their lead is still the bottleneck for anyone who wants to match you. That requires tracking three things: per knowledge task, whether it falls into the category fully takeover-able, takeover-able with oversight, or human work, and whether that classification has been checked recently; how long ago it was last established that a competitor had not yet taken that step, and on what signal that was based; and whether the margin you attribute to that knowledge lead has been recorded as a separate assumption with a date, so that it is clear when it was last tested rather than merely repeated. Where the subject touches on personnel decisions, separate statutory requirements apply to those; these are not addressed here.

The underlying question — which work in this company can genuinely be taken over by AI — is answered by the FTE TO AI work scan per task, not at the level of the entire sector.