There is usually one dimension on which a company distinguishes itself. Faster delivery, more precise quotes, service that feels just a bit more personal. That edge was built with people: planners who are good at shuffling schedules, advisors who are good at listening, a back office that never makes a mistake. The problem is not that this edge disappears. The problem is that the reason it existed disappears.
When AI takes over the work behind a dimension, what that dimension measures changes. Delivery time was long a matter of staffing and experience: how many good planners do you have, and how good are they. If the planning largely does itself, with a human judging the exceptions, delivery time is no longer a staffing question but a software question. Whoever has that system, has the delivery time. Not whoever has the best people.
That is the shift. Not AI in general, but what happens to the comparison the moment a competitor can deliver the same dimension with less human work and the same outcome.
The takeover of work by AI falls into three categories, and these run across every sector. Part of the work can be done entirely by AI. Part goes partway, with a human who approves or rejects and can steer that. And part remains human work, because the judgment, the relationship, or the context cannot be captured in a system.
Where a company stands depends on what kind of work the dimension in question actually requires. A quoting process that runs on combining fixed rules and historical data lends itself more easily to takeover than a quoting process that runs on gauging a customer who doesn't say what he means. Two competitors in the same market can therefore differ sharply, not because one is more forward-looking, but because their underlying work is structured differently. For anyone wanting to distinguish this per company, there is the question of how you measure a competitor's AI maturity from the outside without access to their internal systems.
This says nothing about what a company should do with its staff. Whether and how capacity is redeployed is a decision for the employer, with its own legal requirements where it involves layoffs. What the benchmark delivers is something else: the number of hours that can be freed up by AI within a task, and the fte capacity that frees up for other work. Facts about work, not staffing advice.
The estimate of how much work within a dimension is susceptible to AI is never an exact figure. It depends on how the work is currently organized, which systems are already in use, and how much of the work consists of exceptions that cannot be automated. Two companies with comparable revenue and headcount can have a very different outcome on the same dimension, and that is precisely why fewer people does not automatically mean lower costs. That assumption is developed further in why a competitor with fewer people does not automatically operate more cheaply.
A score in the benchmark is therefore an estimate with a range, not a measurement to two decimal places. It is based on public signals: job postings, software in use, lead times observable from outside, statements made by the company itself. Where those signals are thin, the range is wide, and this is not hidden. An outcome mainly says something within the evidence matrix it sits in: every score can be traced back to the signals it rests on, and where those are missing, that is stated as well.
Equally important is when an outcome says nothing. An edge that holds true today may no longer exist in a year, not because the competitor became smarter, but because the dimension itself changed in nature. How long an edge holds up is therefore not a fixed term but a question that plays out differently per dimension, developed further in how long an AI edge holds up. And not every dimension that is distinguishing now stays that way: some become worthless precisely once everyone in the market can do the same, a question addressed separately in which dimensions become worthless once everyone uses AI.
If the dimension on which you now win becomes democratized, the question remains what you then still win on. That is not the same question as "what are we good at," because internal conviction and external distinguishing power often diverge. That question is addressed in how you know what you truly win on, separate from what the organization itself believes it offers.
Because the underlying technology keeps moving, a comparison is not a snapshot that remains valid for years. How often recalibration is needed depends on how quickly the dimensions in a specific market change in nature, something explored further in how often a competitive analysis should be repeated.
The question of which work within your own company can truly be taken over by AI is not answered by the benchmark but by the work scan from FTE TO AI, which works this out per task.
You can start with the free dimension check: you name what you believe you win on, and see which of those claims can be defended with evidence and which cannot. The full benchmark, with the evidence matrix per dimension, is under construction.