Case study · Regulated insurance
Polly, the policy agent
Born from a leaky ceiling — and a coverage question.
It started with water coming through a ceiling, and no easy answer to a simple question: what does this policy actually cover?
A rough prototype — policy documents plus an AI model — showed the potential within a day. Engineered properly inside a regulated insurer, it became Polly: a policy agent giving staff straight answers about cover, grounded in the actual policy wording.
The engineering that mattered wasn't the chat. It was retrieval grounded in the documents, clear escalation to humans, and firm boundaries — Polly won't speculate, won't chat about the rugby, and certainly can't pay your claim.
The point: useful AI in a regulated business isn't a chatbot. It's knowledge design, guardrails and human review points — wrapped around one good idea.
In the lab · Personal finance
Flynn, the personal CFO
A financial guide for life's biggest decisions.
What if we buy this house? What if I take six months off work? What changes if we have a child?
Most financial tools report what has already happened. Flynn is a prototype exploring the decisions that come first — combining AI, scenario modelling and open banking data so people can see possible futures side by side, in plain language, before they commit.
It isn't a replacement for regulated advice — it's designed to help people understand trade-offs, prepare better questions, and engage a human adviser at the right moment.
The point: good financial tools shouldn't just describe where you are. They should help you decide what comes next — a pattern that applies to decisions inside any organisation.