Consultancies are selling the build, not the deck. The client keeps the run risk.

The load-bearing word in this repositioning is "run". An advisory engagement ends with a document, and its worst outcome is that nobody acts on it. A build-and-run engagement ends with something switched on inside the buyer, and its worst outcome is a system the buyer cannot staff, operate, or unwind. Business Insider describes exactly that swap: not generalist teams producing research and strategy decks, but tools, systems, and ongoing support, delivered through multi-year transformation projects ^[2].

KPMG's own account of the change shows what happens to the contract. Rob Fisher, vice chairman of advisory, says the firm a decade ago would have called itself a time-and-materials business with "smart people doing smart things" ^[6]. Now, he says, clients want to consume its expertise as subscription-style products alongside advice and managed services ^[7]. Time and materials puts scope risk on the buyer and almost none on the seller. A subscription puts availability on the seller and dependency on the buyer. Neither is an outcome guarantee, and the move from one to the other is not a transfer of implementation risk toward the firm.

There is also a supply-side reason for the change of form that has nothing to do with what clients asked for. The largest consultancies have used multibillion-dollar partnerships with OpenAI, Nvidia, Anthropic and Microsoft to build internal tools that automate work once handled by junior staff, and are now rolling out agent networks internally ^[12]. Junior hours were the leverage in the old model. Remove enough of that base and hourly billing stops carrying the margin, whichever way buyer preference runs. Recurring products and managed services are what remains.

The hiring is the part that cannot be restated in a press release. Accenture's recent annual reports show nearly 40,000 AI and data professionals added in two years, and EY reports 61,000 technologists added since 2023 ^[8][9], roughly 101,000 technical staff between two firms ^[10]. Against that, Deloitte is retiring the titles "analyst" and "consultant" for its US employees ^[11], and PwC has rebuilt its training around 15 AI skills and 15 human ones ^[4]. One of those is capacity. The others are vocabulary, and a buyer can tell them apart by asking who will be on the delivery team.

What the reporting does not show is any matching movement on accountability: new delivery forms, new job tracks, new headcount, and no described change to who is liable when the build underperforms ^[19]. Consider the skills question underneath it. Allison Heithoff, five years at West Monroe, studied business administration and now uses AI to code and develop client solutions, work she describes as formerly hands-to-keyboard ^[5]. The person configuring a client's systems may increasingly be supervising a model rather than writing the thing. Supervision quality is now a delivery risk, and it sits on the buyer's side of the invoice.