One agent did all the QA. Now AI does.

The situation

We'd worked with this insurtech for years — product sprints, real builds. So when their leadership decided they needed an AI strategy, they called us instead of a consultancy. They expected a strategy. We proposed something better.

Rather than hand over a deck, we ran the strategy session in parallel with building two working AI prototypes — so the team could see, touch, and feel how AI actually lands inside an organization: the workflows it changes, the roles it reshapes, the way you build with it now.

So we went in. Two weeks on-site, in person. We interviewed their call-centre people, sat behind them while they worked, listened in on live calls, and took a full inventory of their systems.

What we found

One shared inbox. A "knowledge base" that was a PDF of FAQs. And one very senior, very capable agent spending his days running around answering everyone else's questions — then, once a month, carving out time to monitor calls for quality and compliance.

That was the bottleneck, and the opportunity. Automating it would do two things at once: tighten governance, and give their most valuable person his time back — time he could spend training a bench of young leaders who'd been starved of it while the company scaled. The COO and the head of the call centre agreed. That became the scope.

What we built

Two prototypes, in two weeks, both grounded in what we'd watched them do.

• Automated call monitoring & QA. Every call is transcribed and scored against the criteria that matter — call quality and UK insurance compliance (FCA, data protection, Treating Customers Fairly, complaints handling). Breaches get flagged automatically; agents get specific coaching notes. Monthly spot-checks became continuous coverage.
• A knowledge assistant. We turned that PDF into a real, searchable knowledge base — 1,441 answers across every policy — so any agent can find the right answer in seconds, instead of shouting across the floor.

What changed

• Every call gets reviewed now — not the 5% a human could get to once a month. Compliance risk stops hiding in the calls no one listened to.
• Their senior expert got his month back. Instead of QA and question-answering, he's coaching the next generation of team leads.
• The strategy landed because they could feel it. Two working prototypes told them more about their AI future than any slide could — and gave them a running start instead of a plan.