Enabled the launch of a production-ready AI platform in just three months
Capable of improving the outcome of 50% of all insurance claims
Detects six times more potential signs of fraud
Using Vertex AI, BigQuery, and Gemini, Avantia boosts fraud detection sixfold while empowering agents to deliver faster, fairer service.
Using Vertex AI, BigQuery, and Gemini, Avantia boosts fraud detection sixfold while empowering agents to deliver faster, fairer service.
Tackling the growing cost of insurance fraud with AI
Not only are their tools incredibly accessible and easy to work with, Google's deep investment in generative AI — and the conviction behind its tooling — really stands out.
Dan Huddart
Chief Technology Officer, Avantia
In the United Kingdom, insurance fraud costs an estimated £1.1 billion every year. For insurers, detecting fraud effectively while preserving service quality for genuine customers is a long-standing challenge. "Agents support customers going through what can be highly emotional situations, such as burglaries or fires," says Dan Huddart, CTO at Avantia Group, "At the same time, they're processing forms, checking data, and copying information across multiple systems — all while keeping an eye out for red flags."
Avantia Group, a technology-first UK-based home insurance company, aims to make fair, affordable home insurance accessible to everyone.
The company currently insures more than a third of a million homes across the UK, has the largest digital underwriting footprint in the market, and aims to reach one million properties in the future. Avantia began exploring the potential of AI and machine learning (ML) in 2018, achieving strong results by moving its pricing models onto an ML platform. However, ML struggles to effectively automate areas that rely on the unstructured data central to insurance claims, such as call recordings, documents, and photos. The advent of generative AI, and its ability to effectively process and interpret that data, set Avantia on the path to developing the AI-powered fraud detection and claims analysis platform, 'Holmes.'
"We were really excited when gen AI came along and gave us the opportunity to extend what we'd done with ML into other sectors," says Huddart. The team took part in a 24-hour Google Cloud hackathon that enabled them to develop a working proof of concept for Holmes using Vertex AI, BigQuery, and Gemini 2.5 Flash. A production-ready version was then built and launched.
"When you're making a technology investment, you have to consider not just who's leading the market today, but who will still be leading in five years," says Huddart. "That's why we chose Google Cloud. Not only are their tools incredibly accessible and easy to work with, their deep investment in generative AI — and the conviction behind its tooling — really stands out."
From prototype to production-ready in just three months
Gemini is a market-leading model. We liked the combination with Vertex AI because it solved several challenges we'd always faced with large language models, namely change management and safe deployment features, which are essential in a regulated financial business. We also appreciate the speed of innovation and the integration with BigQuery.
Dan Huddart
Chief Technology Officer, Avantia
When a customer files a claim, Holmes instantly analyzes thousands of data points about the circumstances, the policy sold, and the history. It identifies what's covered under that customer's specific policy, flags any inconsistencies or potential signs of fraud, and presents a clear summary to the claims handler within seconds.
Holmes was built directly on the original hackathon prototype by Avantia's in-house data and engineering teams, which adapted Avantia's machine learning systems to use generative AI and ensured safe, accurate LLM deployment.
"Gemini is a market-leading model, and we liked the combination with Vertex AI because it solved several challenges we'd always faced with large language models, namely change management and safe deployment features, which are essential in a regulated financial business. We also appreciate the speed of innovation and the integration with BigQuery," says Huddart.
The platform uses BigQuery to consolidate information from multiple internal systems into a single, structured dataset, providing one unified context view for analysis. Vertex AI powers the AI orchestration and reasoning layer, while Gemini 2.5 Flash handles supporting tasks such as summarizing information and scoring data. Gemini 2.5 Pro is used for core analytical work that demands higher accuracy. "What we found genuinely different and refreshing about the Google Cloud toolset is the developer experience," adds Huddart. "Our teams want to build solutions quickly to solve real problems, not spend time configuring infrastructure and worrying about scaling."
The company estimates that Holmes will save Avantia around £1.2 million next year through a combination of efficiency savings. "Holmes allows our agents to really focus on the customer," Huddart continues. "It enables us to detect six times more potential signs of fraud than before, with tests showing the detection rate for fraud referrals rising from 2% to 12%. In half of those test cases, it identified something that could materially improve how the claim was handled, either by enabling faster resolution or saving money by avoiding unnecessary costs or delays."
From fraud detection to deeper transformation
Avantia sees Holmes as the beginning of a much larger transformation. "We have real conviction in using AI to run a better, more efficient, and more satisfying claims operation," says Huddart. "Holmes is just the first step. We have the ambition to be the first UK home insurer to redesign our entire operation around the capabilities of AI." The company will continue to expand Holmes' capabilities over time, ensuring it becomes more effective as AI models advance. Avantia is reimagining the design of its entire operation — including claims processes, decision making, and supplier management — so that every element is optimized around the use of AI wherever possible.
"We currently insure about a third of a million properties, and we're on track to insure one million homes in the future. That gives us a real sense of urgency around building systems and processes that scale efficiently," Huddart says, "Our aim is to use technology and AI tools so the business can stay small, nimble, and agile, serving three times as many customers with the same or even better level of service."
We're confident in the quality of Google's AI models and Vertex AI's ability to help us create increasingly intelligent and interconnected AI systems. Ultimately, this will enable our people to concentrate on what humans do best: providing empathy, communication, and customer experience.
Dan Huddart
Chief Technology Officer, Avantia
The Avantia team sees gen AI as a fundamental shift for the insurance industry, one comparable to the advent of the internet and online insurance aggregators. "Insurers that embrace AI and rethink how they operate around it will have a real advantage," Huddart concludes. "We're confident in the quality of the Google AI models and Vertex AI's ability to help us create increasingly intelligent and interconnected AI systems. Ultimately, this will enable our people to concentrate on what humans do best: providing empathy, communication, and customer experience."
Avantia Group is a high-growth home insurance platform and the parent company behind Homeprotect, an award-winning direct-to-consumer brand for home insurance.
Industry: Financial Services
Location: United Kingdom
Products: Google Cloud, BigQuery, Gemini 2.5 Flash, Gemini 2.5 Pro, Vertex AI
