If delivery time was a matter of headcount, the party with the most people on the floor won. Once planning largely does itself, headcount is no longer the lever that result turns on. The comparison between you and your competitors shifts not because AI "does something to the market", but because the dimension on which you used to win moves somewhere else. Anyone who does not work that through keeps comparing themselves on a scale that no longer determines the outcome.
This does not happen equally everywhere. In some companies, AI already takes over most of a task, with a human approving or rejecting. In other companies, the same task is still entirely human work, not because the company is behind, but because the task does not lend itself to it, or because the oversight required is more expensive than the time it saves. The difference between those two companies lies in the task, not in management's intention.
Three categories run through every competitive position: work that AI can take over, work that AI partly does under human oversight, and work that remains human work. A competitive dimension that falls entirely into the first category stops being a distinguishing capability once the technology becomes widely available: everyone can copy it. A dimension in the second category remains distinguishing, but then on the quality of the oversight, not on the speed of execution. And a dimension that remains human work becomes relatively heavier in the comparison, precisely because the rest catches up faster.
That does not mean AI delivers the same amount everywhere. What it comes down to depends on how repeatable the task is, how well structured the input is, and how much risk a wrong decision carries. A task with a lot of variation and high stakes remains human work longer than a task with fixed steps and low stakes. What you cannot copy from a competitor deploying AI addresses exactly that distinction: not every lead is transferable, even when the technology is.
Commercial decision-makers who lose deals often point to price or delivery time, because those are the dimensions on which loss is visible. But visible loss and actual cause do not always sit in the same place. A customer you lose to a competitor who responds faster may not be lost to speed but to the consistency with which that competitor now responds equally fast everywhere, something that only becomes possible once part of that work has been taken over. Why your customers see the difference before you do explains how that delay in your own perception arises.
The competitive benchmark scores your position and that of your peer group on the dimensions that actually decide deals in your market, with an evidence matrix in which every score is traceable to a concrete observation: a win-loss conversation, a price list, a delivery time, a response time. No score without a source, and no source without a note on how solid it is. Some scores rest on hard data, others on a limited number of conversations, and that difference in certainty is stated, not smoothed over.
A benchmark says something about the dimensions that were included, not about dimensions no one has named. If a new entrant wins on something that is not yet on the list of measured dimensions, the benchmark only sees that after it has already translated into lost deals. That is a reason to ask the question again periodically, not to distrust the outcome of a one-time measurement. How often a comparison with competitors needs repeating discusses that frequency, and what a shift in the market signals for it.
A second limitation: a score on a dimension says something about the current position, not about the cause. A low score on delivery time can be due to headcount, to an outdated system, or to a process that has not yet been redesigned around what AI can already handle within it. The benchmark points to the dimension; it does not automatically explain why. That requires additional work, and that work sometimes touches on the question of which tasks an organization wants to set up differently. Where that question touches on personnel decisions, separate statutory requirements apply; the benchmark is not a basis for such decisions and makes no statement about them.
A third limitation concerns entrants without an existing workforce, who sometimes set up a dimension from scratch around what AI can already handle, without the detour of an existing process. How to recognize an entrant that starts without staff describes which signals point to that before that entrant shows up in your win-loss overview.
The benchmark guarantees no won deal and no correct prediction of what a competitor will do tomorrow. It delivers a traceable state of affairs on the dimensions that make the difference today, with an indication of how certain each score is. Whether a task within your own company can truly be taken over by AI is a different question from what you compete on; that question is answered per task in the work scan of FTE TO AI, separate from the benchmark itself.
How you know what you truly win on and the question of how often a competitive analysis deserves repeating, as worked out in how often you should repeat a competitive analysis, are both questions that can be tested before you settle them internally. The free dimension check is the starting point for that: you name the dimensions on which you think you win, and see which of those claims can currently be defended with evidence. The full benchmark, with the peer group and the evidence matrix, is under construction.