Professional services — law firms, accountancy, consultancy, HR services, legal and tax advice — essentially sells something that is hard to automate without discussion: judgment. Where a factory optimizes a process and a shop manages an assortment, this sector sells the expertise of people who know something the client does not know. That makes the relationship between ambition and readiness different than elsewhere. Interest in AI is high, often higher than in sectors with less margin on knowledge. But the question of whether the organization can carry that interest often stalls on parts that have nothing to do with AI.
Many firms in professional services have grown around the partner, the senior advisor, the person with the network and the memory. Knowledge sits in heads, in email exchanges, in files built up per client or per case, not in shared systems that connect to one another. That works fine as long as the people stay. It becomes a problem the moment you want to build something that must be able to rely on that knowledge. AI that has to work with files, precedents, or client correspondence needs something to run on: a structure in which information is findable and comparable. That structure is still missing in parts of the sector, not out of unwillingness, but because the work itself never asked for that structure.
The five levels of the measurement — baseline, foundation, activation, insight, intelligence — show a varying picture per dimension in this sector, and that picture is rarely even. Some firms have their data management in order: files are structured, access rights are arranged, there is a sense of what is stored where. But the IT infrastructure behind it is outdated or fragmented across standalone applications that do not talk to each other. At other firms it is the exact opposite: modern systems, but organizationally no clarity about who decides on what, who may approve an AI application, or who is responsible if something goes wrong. Both situations lead to the same result — a pilot that works well on its own but does not scale because the fundamental dimensions are not at the same level. That is the order the measurement reveals: organization, infrastructure, and data management must first reach a certain level before the dependent dimensions, such as automated testing or continuous optimization, can benefit from it.
The plotting round, in which several people score separately, often reveals a specific pattern in professional services: the partner or director scores the organization higher than the advisor who works with the files on a daily basis. That difference says something. It may mean that decisions about AI are made at management level without reaching the work floor, or that the work floor has long been struggling with a system that is judged sufficient from the top. Both interpretations are useful, and both can only be made visible by having different people score separately, not by asking a single impression from whoever draws up the report.
What sets this sector apart from, for example, retail or hospitality is the sensitivity of the data itself. A law firm or accountancy firm works with information that falls under a duty of confidentiality, under disciplinary law, under rules that do not apply to a shop's inventory or a menu. That directly affects the data management dimension: not just whether data is findable, but whether that data is stored and shielded in a way that allows AI use without breaching confidentiality. That calls for a different pace than sectors where data is less sensitive, and that difference in pace is visible in the scores on this dimension.
The pattern of high ambition alongside an incomplete foundation is not unique to professional services. In the IT sector the opposite problem often plays out — strong infrastructure without organizational clarity — and in financial services the delay often lies with regulation that changes faster than the systems can keep up with. The comparison is not meant to make one sector better or worse than another, but to show that the order — first the fundamental dimensions, then the dependent ones — is the same everywhere, even though the details look different per sector.
This measurement says nothing about which tasks within a firm can be taken over by AI, and how much time that would free up. That is a different question, one that calls for a different lens: not at the organization as a whole, but at the work itself, task by task. Anyone who, after this measurement, wants to know what work, given the current readiness, is actually eligible for AI, will find that in the work scan from FTE TO AI, which calculates per task what portion can be taken over.
The tool that carries out the maturity measurement and the plotting round is still under construction. Anyone who wants to use it once it becomes available can sign up for the waiting list.
Vraag maar wat er moet staan voordat AI in uw organisatie kan landen.
Answers come from this site’s knowledge base. Not tailored advice, and not a scan of your company.