Until recently, the question of how mature a competitor is in AI was mostly a curiosity. Now it is a question with commercial weight, because AI in parts of the work actually takes over tasks: some completely, some with a human who approves or rejects, and some not at all yet. Where that happens changes not only a competitor's cost price. It changes what you still win on. Delivery time was long a function of headcount: more people in planning, shorter lead time. If at a competitor planning largely runs itself, delivery time is no longer a staffing question but a software question, and that changes who is ahead on that dimension. That is precisely why measuring from the outside makes sense: not to know how much AI a competitor uses, but to know which dimensions of the comparison are shifting as a result.
From the outside you see no systems, no prompts and no internal process descriptions. You see behavior and consequences: response times that drop while the team does not grow, job postings that shift from executing to reviewing, delivery times that get faster with unchanged job advertisements, pricing that becomes just slightly more flexible than the cost structure would explain. None of these signals proves AI use on its own. Together, across multiple dimensions, they form a pattern that does say something. The method is therefore not reading off an AI percentage for a competitor, that does not exist as an externally observable number. The method is scoring the dimensions on which your market is actually won, and asking for each score what the underlying evidence is and how solid that evidence is.
Every score in a competitive benchmark stands next to the evidence on which it is based: a job posting, a customer signal, an observed delivery time, a price change. That makes the score testable and, just as important, refutable. A score without a traceable source is an assumption with a number attached. In a management meeting that difference quickly becomes clear: the question "what is this based on" needs a concrete answer, otherwise the comparison is noise. That same test, incidentally, also applies to your own organization. Before comparing with the outside world it is useful to know what a peer group is and how you assemble one, because a score next to the wrong comparison group is just as misleading as a score without a source.
An external estimate of AI maturity has limits that cannot be wished away. A competitor may be running a pilot that has not yet produced any external signal, and therefore score unjustly low. A competitor may also make an AI claim in marketing material that is not reflected in the executing work, and therefore score unjustly high until that claim is tested. Staff numbers that remain stable say little if part of that staff is structurally hired in through third parties. And a fast response time can just as easily be the result of an extra service as of automation. The method therefore works with bandwidths and with explicit uncertainty margins per dimension, not with a single number that suggests a certainty that is not there. A score that rests on one weak signal is also flagged as such in the matrix.
The outcome says little if the dimension itself is no longer distinguishing. If AI deployment on a particular dimension has become commonplace across the entire market, then no one wins on it anymore, and scoring it is an exercise without commercial consequence. That is one of the reasons the benchmark starts with the question which dimensions become worthless once everyone deploys AI, before time goes into scoring dimensions that no longer affect the answer. The same applies to a lead that is real today: it lapses as soon as competitors take over the same work, and the question how long an AI lead holds up therefore belongs in the interpretation as standard, not as an afterthought.
The external estimate of a competitor is one half. The other half is the question of which work in your own company can actually be taken over by AI, and that question is answered per task with the work scan from FTE TO AI. That same distinction, what AI takes over, what happens under supervision and what remains human work, is also what you can still compete on once the executing work has largely been taken over, and that is a question that does not only play out externally: see what you still compete on when AI does the executing work. A competitor with fewer people on the payroll is not automatically the cheaper party, as worked out in why a competitor with fewer people is not automatically cheaper. Where any of these questions touches on personnel decisions within your own organization, its own legal requirements apply; this page does not provide substantiation for such decisions.
You can start without waiting for the full benchmark. Identify the dimensions on which you think you win, and test them with the free dimension check: you see which of those claims can be defended with evidence and which rest on assumption. Those who want to look beyond their own position can aim the same test at where room remains unused at competitors via a white space analysis. The full competitive benchmark is under construction; the dimension check is already available now.
Vraag maar waarop er in uw markt gewonnen wordt. Ik vergelijk liever dan dat ik uitleg.
Answers come from this site’s knowledge base. Not tailored advice, and not a scan of your company.