From Bircher muesli to machines: Zurich’s AI offensive in insurance (Part 1)

Zurich Insurance Group used its hometown as a stage for a live demonstration of how artificial intelligence is mutating from hype to hard numbers. By reshaping insurance with imagery, co-pilots and agent-based systems, AI is uncovering risks, shortening cycle times and boosting growth. As every year, the invited guests entered the Innovation Championship Festival with great excitement.

The evening of October 16, 2025 began with walking through a slightly creepy, long and dark corridor full of smoke (dry ice). The Zurich Development Center is located high up on Zurich’s hills overlooking the city and lake. This is the place where, around a hundred years ago, the Swiss doctor Maximilian Bircher-Benner ran a sanatorium where he combined nutrition and science and invented the famous Birchermuesli. According to the hosts, this spirit of experimentation never left the city and Zurich Insurance. Innovation, they reminded the audience, was “firmly anchored in the corporate structure”.

AI as a strategic accelerator

Paolo Mantero, Group Chief Strategy Officer of Zurich Group, explained that artificial intelligence is not a passing trend, but a strategic accelerator. Zurich sees AI as a means of being there for customers around the clock, improving the frequency and quality of every interaction, and building a collective ‘digital memory’ for the company. This intelligence, shared by teams and generations of employees, will make underwriting more accurate and collaboration with the company faster and easier.

However, as competitors also have the same technology, Zurich’s success depends on deploying it faster and more widely and working with partners rather than going it alone. This approach has resulted in over fifty collaborations, supported by the Zurich Innovation Championship and this year’s new Agentic AI Hyper Challenge.

Recognizing risks from the air

The collaboration with Nearmap highlighted how external partnerships are already transforming the business. Chief Technology Officer Tom Celinski described how the company’s aerial imagery captures details that satellites overlook, such as roof corrosion or flood risk. The images are processed by AI, which extracts up to 130 property features and creates a detailed risk profile that underwriters can use immediately. Nearmap’s platform, currently in use at Zurich North America and expanding to Australia, has already uncovered hundreds of thousands of hidden risks for other insurers.

The digital brain of underwriting

Penny Seach, Group Chief Underwriting Officer at Zurich Insurance, described underwriting as the heart of insurance and the area most ready for change. Zurich’s goal is to create a “digital brain” that learns from every decision and supports underwriters rather than replacing them. Interloom ‘s winning solution showed how scattered corporate knowledge can be transformed into instant, actionable answers for brokers, reducing decision times from minutes to moments.

A second partner, Sixfold, has developed an AI “underwriting brain” that reads vast amounts of risk data and claims reports and summarises them in line with Zurich’s risk appetite. The system has been rolled out to all underwriters in the mid-market segment in the US and has halved the time it takes to prepare quotes and increased the premiums generated by each underwriter by around 30 per cent.

Learning from within

Zurich employees themselves are contributing to this transformation. Jamie Ward and Luke Rutley from Zurich Insurance UK are the winners of the internal hackathon. They presented a multi-agent tool that analyses accident reports, compares them with traffic regulations and relevant case law, and makes consistent, documented liability decisions. In a pilot case, the AI’s conclusion matched a decision that was later revised, demonstrating the potential of internal expertise in intelligent systems.

A learning advantage

What emerged on stage in Zurich was a vision of competitive learning. The insurer of the future, according to Paolo Mantaro, will be the one that combines proprietary data, external insights and human judgment in a feedback loop that improves with every transaction. Excellent underwriting, once a question of experience, will increasingly become a question of collective intelligence.

Binci Heeb

A second part of this article will be published in the next few days.

See and read also: InsurTech as a bridge builder: Kobi Bendelak on Israel, Switzerland and new hubs