Elsewhen

Fifteen years old, around 150 people, and a client list—Google, Spotify, Mastercard, Inmarsat, Zego—that is unusually easy to verify for a firm this size.

The AI work is published under NDA, and it is the most numerate anonymized work in this register. Semantic search built on vector matching for one of Europe's largest package-holiday providers: matching by meaning in under 250 milliseconds, tracking to £15 million in incremental annual booking revenue within two weeks of go-live, with a 1.3% lift in end-to-end booking conversion behind it. A multi-model AI toolchain pointed at a FTSE 100 forecasting platform, ingesting documentation, mapping the codebase and reading transcripts to compress discovery from weeks to days—about a quarter of the project timeline and spend reclaimed, and every user chose to switch.

That is the shape of the practice: AI applied to a product outcome someone is measuring, rather than a model delivered as the deliverable.

The weaker fit is a buyer who needs the reference call. The marquee names attach to the design and platform work; the AI cases name a sector and stop.