Every year, Connections gives our community a chance to step back from the day-to-day and look at where the insurance industry is heading. I’ve spent more than 30 years working alongside the insurance industry’s technology transformation, almost 13 of them at Guidewire, and Connections has long been a fixture in my calendar. I thought I knew what to expect. I didn’t. This year’s was, by some distance, the most extraordinary Connections I have experienced. Sitting in the main keynote, I had the sense of watching the next decade of P&C insurance take shape on stage.
One line from the opening has stayed with me ever since: the future is already here, it’s just not evenly distributed. For those of us working with insurers across Europe, the Middle East and Africa, that line lands with particular force. Our region spans dozens of markets, languages and regulatory regimes. Some insurers are already running AI in production; others are still deciding where to start. Our job at Guidewire is to help every one of them adopt that future at a pace that suits them, with confidence.
A year of unprecedented change
The main keynote opened with a bold observation: the past year has brought more technological change, especially in AI, than perhaps all previous years combined. AI systems can now carry out complex tasks accurately over extended periods of time. Coding agents, frontier models and a new wave of personal agents have moved from curiosity to everyday tools in a matter of months.
The message to the audience was simple and practical: use these tools yourself. Understand what they can do first-hand, then think about how they could reshape your enterprise and the wider industry.
Why does this matter so much now? The keynote pointed to a McKinsey analysis of the industry’s persistent challenges: slow growth, low productivity, high distribution costs and a slow pace of change. Anyone who works with European insurers will recognise that list. Add rising claims costs driven by inflation and more frequent extreme weather, an ageing workforce and a demanding regulatory environment, and the pressure on our customers is clear.
AI offers a fundamental opportunity to address these challenges. But the keynote was honest about the catch: it requires us to change the way we change. Technology is rarely what holds insurers back. Organisations are. Getting teams and governance to keep pace with the tools is now the real work.
What doesn’t change
In a period of so much disruption, I found it reassuring that the keynote started by being clear about what stays the same.
Our mission is unchanged: to power the insurance industry with a trusted platform that enables engagement, innovation and efficient growth. The fundamental insurance applications (product, pricing, distribution, policy, claims and billing) remain essential. How they are deployed, and how people and systems interact with them, will evolve. But the need for a reliable system of record does not go away in an agentic world. If anything, it becomes more important.
The big shift is that our platform will treat AI agents as first-party consumers. Agents will work alongside people, using the same data, APIs and business rules, under the same controls. And we remain fully committed to our partner ecosystem. In EMEA, where so much depends on local market knowledge, that commitment matters.
This is where the modern core comes in. The keynote described six characteristics of a modern core, all embodied in Guidewire Cloud Platform:
- A domain data model built for insurance
- An API-first architecture
- Comprehensive data collection
- Runtime configurability
- Continuous updates
- Built-in security and governance
For European insurers, that last point deserves emphasis. With frameworks such as DORA, the EU AI Act and GDPR shaping how technology is adopted, agentic AI cannot be bolted onto fragmented legacy systems. It needs a governed, secure foundation. A modern core is what makes trustworthy AI possible at scale.
Agentic underwriting: from mundane tasks to differentiated work
The headline announcement for many in the room, and certainly for me, was Guidewire’s agreement to acquire hyperexponential, a company specialising in agentic underwriting. The planned acquisition reflects a shared ambition to help transform underwriting so that human underwriters spend less time on mundane tasks and more on high-leverage, differentiated work. With hyperexponential’s strong roots in the London market, this is news with a distinctly European flavour.
The keynote then showed what agentic underwriting looks like across the product lifecycle.
Advanced Product Designer (APD). AI agents can now create new product structures, such as a cyber insurance product, directly from unstructured inputs like broker brochures. They can also implement complex regulatory changes weeks faster by generating the required rules and attributes automatically. The example on stage was California’s wildfire measures, but the parallel for EMEA is obvious: think of the steady stream of consumer protection, climate and pricing rules our customers must translate into products, market by market. CNA shared that it launched nine specialty products in a year with APD, including a complex line in just four months.
PricingCenter. When products change, AI agents adjust pricing plans automatically and precisely, for example by applying wildfire mitigation discounts. PricingCenter also offers AI-generated explanations of complex formulas, automated version naming and guided deployment to live pricing engines.
UnderwritingCenter. AI-powered agents triage submissions, clear risks, provide recommendations, summarise exposures and answer questions using all linked sources: emails, attachments, account history and guidelines. They populate risk data directly from submissions. The new Risk Accumulation capability uses interactive maps and geospatial insights to identify concentrated exposure.
Agentic claims in action
Claims is where insurers keep their promise, and the keynote showed how AI agents can help them keep it faster and more consistently.
Concierge, a new AI-native product, automates and scales policyholder interactions at First Notice of Loss. The live demo was one of the highlights of the week. An AI agent held a detailed, natural conversation with a policyholder, gathered all the information required and filed a complete, straight-through-processing-ready claim directly into ClaimCenter in real time. It also handled multi-language input. Plenty of European insurers serve customers in five or more languages, and a single storm can bring thousands of claims in a few days. An agent that copes with both is no small thing.
ProNavigator, Guidewire’s AI assistant, is embedded directly in core systems such as ClaimCenter. It works in three ways:
- Answers from the insurer’s own knowledge base, such as guidelines and procedures, with links to the source
- Suggestions with contextual recommendations and next best steps based on the claim or policy
- Actions through specialised agents that execute tasks on the user’s behalf
In the demo, ProNavigator identified a subrogation opportunity and automatically prepared a subrogation notice package, complete with supporting documentation, ready for an adjuster to review and send.
The results are already real, including here in EMEA. South Africa’s Hollard reported a 70% reduction in “dead time” on customer calls using claim summarisation.
Building, governing and powering agents
Agents are only as valuable as an insurer’s ability to build, test and trust them. That is why some of the announcements I am most excited about are the ones that sit underneath the business applications.
Agent Studio is a framework for building, improving, governing and monitoring AI agents. The demo generated an agent that flags discrepancies between claim reserves and service costs. It also showed non-deterministic reasoning in action, such as spotting suspicious front-grille damage on a rear-end collision. Skills are defined modularly and tested rigorously against data sets before go-live. In production, teams can monitor how often agents are invoked and how often they fail. That built-in testing and monitoring is what will get agents past the risk committee.
Around it sits a set of tools that speed up everything else. DevSpace lets developers spin up cloud development environments for ClaimCenter or PolicyCenter quickly. Developer Assistants have cut development time by 50% at early adopter RCIS. Explore Assistant lets analysts query and visualise data in plain language. And Kaleidoscope lets data scientists build predictive models on Guidewire’s anonymised Industry Intel data without sharing raw customer data.
Beyond the main stage: highlights from the other keynotes
The main keynote set the vision. The other keynotes were about what it takes to make it work.
USAA’s “1922 moment.” Randy Termier, President of the P&C Group at USAA, reminded us why we do this work. USAA was founded in 1922 by 25 army officers who were refused auto insurance. Today it serves more than 14 million members and, every 20 minutes, handles around 400 claims and 1,000 member interactions. USAA is now fully migrated to Guidewire ClaimCenter, and a key to its success was a “model office” where frontline adjusters and agents tested the technology before rollout. Randy’s challenge to the industry was about availability and affordability. After a catastrophe, customers want to know one thing: will you insure me again when I rebuild? That question is just as pressing for families affected by floods in Germany, Belgium, Spain or Italy. His closing words were memorable: if you cannot see the customer from where you sit, it’s time to move your chair.
Securing AI agents, with OpenAI. This session explained why agents need a new security paradigm compared with traditional APIs. Agents that access sensitive data, communicate externally and consume untrusted content, such as claim files, medical records and police reports, face the greatest risk from prompt injection. The answer is defence in depth: model training and oversight agents on one side, and sandboxing, least-privilege access and hard authorisation limits on the other. Two ideas stood out for me. First, prioritise protection against irreversible actions. Second, avoid “approval fatigue” by reserving human review for genuinely high-risk decisions.
ProNavigator in practice. TD Insurance and Gore Mutual shared how ProNavigator is tackling the “glue work” of finding information and moving it between systems. At Gore Mutual, tasks that took three to five minutes now take 20 to 30 seconds. The insurer scaled from a 27-user pilot to 210 users after reaching a 90% accuracy bar. TD Insurance described how newer employees now “ask Pro Navigator first,” freeing experienced colleagues to coach rather than simply answer questions.
The agentic SDLC. In the Developer Assistant keynote, the team fed the requirements for Australia’s General Insurance Code of Practice into the Developer Assistant. In 2023, a manual implementation of this code took around four months across nine Guidewire products and roughly 500 user stories. The Developer Assistant generated working code in six hours. A live demo then showed parallel sub-agents building a ClaimCenter change, a Jutro dashboard and a workflow from a single instruction. Six hours against four months is the number I keep coming back to.
The day closed with author David Epstein on range, breadth and the uniquely human skills that grow more valuable as AI spreads.
What this means for insurers in EMEA
Looking back on the week, four things stand out for insurers across the region.
- Start with the foundation. Agentic AI delivers most where data, processes and controls are already modern and governed. Getting the core right comes first.
- Keep people in charge. In every demo, a person made the decisions that mattered. Agents did the groundwork so that underwriters, actuaries and claims handlers could spend their time on judgement.
- Build trust by design. Explainability, auditability and security are what will allow AI to move from pilot to production under European regulation.
- Keep the customer in view. Faster claims, fairer pricing and continued availability of cover in a changing climate are what will ultimately define success.
The future may not yet be evenly distributed, but it is arriving faster than any of us expected. I suspect the insurers that thrive over the next decade will be the ones that rethink how work flows between people and agents, and build enough trust and governance to let agents act.
For me, that is what agentic insurance comes down to. Agents take on more of the work, and people get the time to apply their judgement where it counts: understanding emerging risks and standing by customers when they need us. As climate change and regulation reshape our region, that may be what keeps insurance available and affordable for the next generation.