Consulting firms: the window to productise your expertise is closing

Inside consulting, HR, legal, accounting, marketing and finance firms, the conversation shifted in 2025. Partners no longer ask whether AI will hit their model — only how fast. And above all: should we try to productise what we know how to do, before our clients do it themselves?

The pattern is the same everywhere. A share of repetitive work — document synthesis, preliminary due diligence, standard-contract review, recurring reporting, first-line qualification — is being absorbed by the tools clients install themselves. The ratio of billable person-days per engagement is dropping quietly. Not violently, but durably.

The productisation bet, and why it usually fails

Turning a firm's know-how into a SaaS platform isn't a new idea. It comes back in cycles, with every technology wave. Internet, mobile, cloud, AI. Every time, a fraction of firms try. Every time, most give up after twelve to eighteen months, with the same lucid verdict: we didn't ship, we burned cash, and we distracted our best partners from the core business.

Three traps recur. First, scope drift: a firm productising its method tends to want to reproduce in software the full richness of what it delivers in engagements, instead of isolating the repetitive, codifiable layer. The result is an ambitious product, slow to ship, hard to sell. Second, the under-staffed internal team: hiring a CTO, two engineers and a PM takes six to nine months, and by the time the team ships, the market has moved. Third, the confusion between consulting and product: a firm that bills hours knows how to sell expert time; selling a £199-a-month subscription requires a different commercial culture.

What changed in 2026

Two things make productisation more accessible today than it was two years ago, and a third one makes it more urgent.

More accessible: generative-AI building blocks now let you encode business behaviour that previously demanded years of R&D. A firm that knows from experience how to qualify a case or draft a first synthesis can now transmit that logic to an AI agent in weeks — provided there is a clean piece of software to host it. Domain modelling — once outsourced to expensive external consultants — can be industrialised in code from version one.

More urgent: clients themselves are adopting ChatGPT, Claude, Mistral. What used to be a competitive edge on internal productivity — being able to handle a case in two days rather than five — becomes invisible when the client expects the same speed. The moment a client says "I'll just do this myself with ChatGPT" is the moment the commercial window closes.

Productising without breaking the firm

The operational question is rarely "can we productise?" but "how do we productise without damaging the firm during the effort?". Partners who pull this off share a few principles.

Isolate the repetitive layer, not the entire business

A firm never productises its full core practice. It productises the lower layer — the repetitive, codifiable work, low incremental value per engagement but high volume. That layer is precisely the one clients are starting to automate themselves. Productising it means taking that part of the value chain back, and freeing partners for the upper layer where human expertise stays irreplaceable.

Sell v1 to existing clients, don't chase new ones

A firm launching a SaaS has an asset nobody else does: a portfolio of clients already convinced of the consulting quality. Version one shouldn't chase new markets — it should be offered to existing clients as a complement to consulting work, often at cost in the first year. That brings fast feedback, social proof, and predictable revenue ahead of acquisition.

Cap the bet at twelve weeks, not eighteen months

A firm cannot afford a long R&D cycle. Not for lack of cash — many have it — but because partner attention is the scarcest resource. Capping version one at an eight-to-twelve-week scope forces discipline: pick the most profitable sub-problem, implement it cleanly, put it in front of a real client, learn. The rest follows or doesn't. What you avoid is the eighteen-month project that drains energy without generating revenue.

The invisible technical layer that makes the difference

On paper, a consulting-firm SaaS looks like any other B2B SaaS: interface, database, authentication, subscriptions. The difference plays out in domain modelling. A firm that productises without modelling its domain produces software that demos well but starts diverging by the third client: business rules scattered across the code, edge cases piling up, and the AI agent eventually hallucinating on situations the firm itself has handled in engagements for ten years.

Conversely, software that explicitly models the concepts of the business — the participants' roles, the states of a case, the transition rules, the invariants — becomes a substrate on which agents can be plugged that behave like a serious junior of the firm. The agent doesn't guess; it operates on a model that contains the knowledge.

What decision for 2026

Not every firm is meant to become a SaaS vendor, and some expertise doesn't productise — the kind whose value lies precisely in the relationship, the contextual judgement, the interpersonal trust. For the rest, the decision isn't binary between "productise everything" and "do nothing". It is: which fraction of our lower layer can be turned into a product within twelve months, with which partner, and with what level of human commitment.

What will no longer be available in 2027, however, is the option "let's watch and wait". The motion is already underway among the clients themselves, and waiting is paid for in margin.