Avantia Group, the digital home insurance platform behind Homeprotect, has launched Lyra, a proprietary agentic AI orchestrator designed to coordinate multiple AI systems across the home insurance claims process. The tool builds on Holmes, Avantia's existing AI claims system launched in 2025, and introduces persistent memory that lets the system identify patterns and inconsistencies across claims over time.
Lyra sits above Holmes and other agentic tools, coordinating them the way a senior manager directs a team of specialists. Where Holmes processes individual interactions, Lyra connects information across the whole claims journey. That cross-referencing capability matters for fraud detection: the system can surface inconsistencies that would be invisible when viewing claims in isolation, helping prevent fraudulent payouts that drive up premiums for everyone.
Avantia says Lyra is the first live deployment of agentic AI in the UK home insurance market. The company currently serves 350,000 customers and has a stated path to scale past one million.
Human oversight remains central
Lyra does not replace human decision-makers. A dedicated operator continuously monitors Lyra's learning and decision-making, and final decisions on all claims stay with human experts. The model is explicit: AI handles precision and process, while claims handlers provide empathy and judgement.
Dan Huddart, CTO at Avantia Group, said: "Claims involve real people, who are sometimes dealing with genuinely difficult or distressing situations, and managing that within the landscape of AI tools requires a level of precision, memory and contextual understanding that previous technologies couldn't deliver. Lyra changes that."
Avantia is the only insurance business participating in the FCA's AI Live Testing initiative, which supports responsible AI deployment in UK financial services. The company has built a governance framework around Lyra aligned with FCA requirements and the Consumer Duty framework. Key principles include human accountability for all final decisions, dedicated human oversight of the system, transparency about AI's role, and auditability of the technology.
Why agentic AI in claims is different
Most insurers already use AI for pricing. Claims are a different challenge. They involve distressed customers, high stakes, and complex judgment calls. Lyra's persistent memory means the system can track a claim's full context rather than treating each interaction as a standalone event.
Huddart said: "Lyra's ability to remember and connect information across claims gives us a level of vigilance and oversight that simply wasn't possible before. That protects our customers, keeps premiums competitive and ensures that the people who genuinely need to make a claim get paid quickly and fairly."
The company is also investing in the skills mix to support the technology: AI specialists to build and govern the systems, alongside experienced claims professionals who ensure the technology is trained on real-world complexity. For insurance professionals tracking the shift toward AI Agents & Automation, Lyra offers a working example of how agentic systems can be deployed in a regulated, customer-facing environment.
Why this matters for insurance professionals
Lyra is a concrete case of agentic AI moving beyond pricing and back-office automation into the claims function, where regulatory scrutiny and customer sensitivity are highest. The governance model - human sign-off on every claim, dedicated monitoring, auditability - is worth studying because it shows how an insurer is trying to square AI efficiency with FCA expectations. For professionals in claims, underwriting, or broking, the practical takeaway is that orchestration tools with persistent memory are becoming viable for fraud detection and customer outcomes, but only when paired with clear human accountability structures. The broader shift toward AI for Insurance is no longer about isolated tools; it's about how those tools are coordinated.
Your membership also unlocks:
