The Same Work. Half the Hours. A Different Business.

Aug 23, 2026 · 3 min read

Case study — enterprise custom software.

A software company we worked with was doing the same work it had always done and earning less for it. Two things were squeezing at once. Clients had started asking to buy an outcome rather than a delivery team. And AI-assisted development was cutting the hours a build took, so a job that once billed for four months now billed for two.

Underneath both sat the same problem. The company understood itself through what it made. Capability statement led with languages and platforms. Account plans organised by product line. Sales conversations opening with what it could build.

That works while clients are buying build capacity. It stops working when they start buying an outcome, because a client who wants a problem solved cannot compare four vendors listing their tech stacks.

We ran a two-day strategy workshop with thirty of their people. The research went in first, and all of it was machine-assembled before a single person sat down. We ran agentic research across every sector the company sold into. Coordinated agents working in parallel, each one gathering public evidence on that sector’s buying patterns, competitors, regulatory pressure and procurement behaviour, cross-checking claims against each other and flagging where sources disagreed.

Each sector came back as a written point of view, not a pile of links. Then we put the eleven client interviews and the staff questionnaire through the same machinery, so what people inside the company believed could be set directly against what clients had actually said.

By hand that is weeks of desk research and it never gets done, which is why most workshops open with everyone in the room reciting their own version of the market. Every one of these packs was written and circulated before anyone walked in.

Ten limiting beliefs surfaced on the first day. Most were product-centred: that clients choose on technical capability, that platform expertise is the differentiator, that a broad build capability is safer than a narrow specialism. The evidence contradicted several outright.

The company came in describing what it made. It left describing what its clients were trying to get done. The AI did not produce that shift. Thirty people arguing for two days produced it. What the AI bought was the argument itself: nobody had to spend the first morning explaining the market, so the room could spend both days disagreeing about what to do with it. That is the trade worth making. Machines do the gathering and people do (some of) the deciding.