As AI vendors release agentic tools capable of performing tasks once handled by entry-level workers in fields such as finance and law, some professional-services firms are rethinking how their businesses are structured. These industries have traditionally run on people: expertise is accumulated slowly, and junior staff climb a ladder of progressively harder work until they become the experts a firm sells. AI is unsettling that model, both by compressing how quickly expertise can be acquired and by changing how firms create and demonstrate value to clients.
Kantata, a cloud-based professional-services automation (PSA) vendor, argues the answer is not to replace people with agents but to make the two work together, supported by what it calls an “expertise engine.” In a recent interview, Sarah Edwards, Kantata’s chief product officer, set out how AI is pressuring service industries and the changes she believes they need to make.
The career ladder is being compressed
Service businesses have historically grown by adding people and depending on experts built up over years. Edwards notes that her own path to expert status as a consultant took around two decades. That timeline is collapsing. By her account, skills that once took five, ten, or fifteen years to develop can now become outdated in well under three, and people are increasingly expected to reach expert-level output far sooner, in part by working alongside AI agents that supply knowledge on demand. The traditional apprenticeship, in which juniors learn by doing low-level work, is being disrupted at exactly the point where that work is most easily automated.
The entry-level problem
That creates a genuine dilemma. If agents absorb much of the routine work that once trained newcomers, firms risk hollowing out the pipeline that produces their future experts. Edwards is candid that no one has a tidy solution. The opportunity she sees is to surface the tacit, hard-won knowledge of a firm’s best people and make it available to everyone, so that good advisers can become great and newcomers can step up faster than traditional training allowed. In that framing, AI is less a replacement for junior staff than a mechanism for spreading expertise that used to live in a few heads.
What an “expertise engine” is meant to do
Kantata’s proposed answer is to combine its operational data with a knowledge graph that models how a professional-services business actually works, spanning estimating, staffing, delivery, forecasting, and financial management, and to let that system learn continuously from every project and client interaction. The aim is to give AI agents enough business context to act usefully across those workflows rather than at a shallow, generic level. Edwards draws a distinction between using AI at the surface, which she suggests most organizations now do to some degree, and using it to transform how a services business is run and how it allocates resources.
Rethinking how services are valued
Underlying all of this is a shift in the unit of value. When a firm sells hours, headcount is the business. When agents can do some of that work, the calculation changes: a firm must allocate a mix of people and agents to projects, understand the cost and the value each brings, and increasingly price around outcomes rather than time. Edwards frames the central management question as how to deploy people and agents together, account for the cost of the agents, and capture the value they create, a different metrics framework from the hours-based model that has long defined professional services.
Practical implications for firms
For services leaders, the takeaway is less about adopting any single tool than about rethinking three things at once: how talent is developed, how work is staffed, and how engagements are priced. On talent, firms may need deliberate programs that give junior staff the judgment-building experience that routine work used to provide, because simply removing that work does not by itself create experts. On staffing, project planning increasingly has to treat human and AI capacity as a blended resource pool, with the cost and reliability of each accounted for. On pricing, a gradual move from billable hours toward value- or outcome-based models changes incentives across the firm, from how consultants are measured to how clients are sold. Most firms will adopt these shifts incrementally rather than all at once.
What to watch
The perspective comes from a vendor with software to sell, so its optimism about human-agent collaboration should be read with that in mind. Several open questions remain. Compressing the apprenticeship carries real risk if firms cut entry-level hiring faster than they build new ways to develop judgment, since some expertise still comes only from experience and accountability. Tacit knowledge is notoriously hard to capture in a knowledge graph, and an “expertise engine” is only as good as the data and governance behind it. Outcome-based pricing, while appealing, shifts risk onto the provider and is difficult to measure cleanly. The broader direction Edwards describes, AI reshaping both how expertise is built and how it is sold, is consistent with what is happening across the wider future of work, but the firms that navigate it well will likely be those that treat agents as a complement to human judgment rather than a substitute for it, and that recognise why so many agent projects stall without that discipline.
