A competitor introduces AI and delivery times drop, quotes go out faster, customer questions are answered within minutes. The reflex is then to ask which tool is behind it. That is the wrong question. Software can usually be replicated. What cannot be replicated is the reason why that tool works for that competitor and perhaps not for you: the quality of the underlying data, the extent to which the process was already standardized before AI was added, and the willingness of people in the organization to trust a system that takes over part of their work.
When AI takes over the work behind a dimension, it is not only speed that changes, but who can win on that dimension. Delivery time was long a matter of staffing: more people in planning, shorter lead times. Once planning is largely done by a system, staffing is no longer the lever being turned. The competitor with fewer staff can then deliver faster than the competitor with more. That changes what a customer sees as distinctive, and it changes what you should be comparing yourself on.
This same pattern plays out across multiple dimensions, always with three variants: a task that AI fully takes over, a task that AI prepares and that a human reviews with a reason for approval or rejection, and a task that remains human work because judgment, relationship, or context cannot be automated. Which dimension falls into which category differs by market and by company. That is precisely why a benchmark made two years ago may now have been overtaken on certain parts, and how often a comparison with competitors needs to be redone is then not a fixed term but a question that must be answered per dimension.
Two companies in the same sector, of comparable size, can reach a very different point at the same time. The difference rarely lies in the tool. It lies in whether the data was already in order, whether the process was already unambiguous enough to automate, and whether there are people in the organization who dare to overrule the outcome of a system when needed. A company where files, quotes, and customer contact have been recorded the same way for years can deploy an AI application on top of that relatively quickly. A company where that work still exists scattered, inconsistently, or only in people's heads must first organize it before taking over the task is realistic. That difference is measurable, not guesswork, and it explains why two competitors with the same ambition move at a different pace.
A benchmark that scores positions on these dimensions is only usable if it is honestly stated what the estimate relies on. A score on delivery time, based on public customer signals and what a competitor communicates itself, is an indication, not a measured fact. A score on something no outsider can see, such as the internal error margin of a process, cannot be given and therefore is not given; a score without evidence behind it is an opinion, not a benchmark. That is why every score in the evidence matrix states where it comes from, and every score has a weight indicating how certain it is. Some signals are strong, such as a consistent pattern in what a competitor's job postings reveal about where they are investing. Other signals are weak and are presented as weak as well.
An outcome also means nothing if the dimension being scored is not the dimension your customer actually decides on. Why customers see the difference before you do is about exactly that gap: internally, scoring is often done on what is measurable, not on what counts. And a score that turns out well on a dimension you were never going to win on anyway is a score without consequence; how you weigh a dimension you will never win on determines whether that dimension should remain in the comparison at all.
If a shift in AI use becomes visible in what a competitor's staff turnover says, that is a signal about the market, not a reason to base your own staffing decisions on this benchmark. What an employer does with its own staff falls under its own legal requirements and outside what this page or this method assesses. Here, only what is visibly shifting in the market is reported, in freed-up hours and capacity, not in who still does that work.
The question of which work in your own company can genuinely be taken over by AI, independent of what competitors do, is answered per task with the FTE TO AI work scan. For the comparison with the market, a simpler first step applies: name where you believe you are winning, and test whether that distinctive capability really exists based on evidence rather than assumption. That happens in the free dimension check: you indicate on which dimensions you believe you are ahead, and see which of those claims can be defended with evidence and which rest on assumption. The full benchmark, with scores for your entire peer group, is under construction.