Guidewire Connections 2026: What it takes to make AI work in claims
Our team spent four days at Guidewire Connections 2026 talking with claims leaders, technical experts, and the people who support them. AI was a prominent topic, from the opening Agentic Insurance keynote to workshops exploring AI across claims and other insurance workflows.
In conversations at the event, three themes kept coming up: whether carriers should build their own AI tools or buy existing solutions, what makes AI outputs trustworthy, and where people need to stay involved in the process.
Each of these themes points to the same idea: AI only delivers on its promise when people trust the results enough to use them. Here’s our take on what that means for claims teams and medical record review.
Carriers have more options than ever to build AI in-house, but complex tasks like medical record review take ongoing time and effort to get right. Claims leaders should consider where and how their teams’ time is best spent, and where a specialist tool can do the work more reliably.
Guidewire’s keynote session highlighted the growing range of AI capabilities available to carriers, including agents built into its applications and tools for building custom agents on Guidewire Cloud Platform. AI coding assistants are accelerating development, giving insurance teams more flexibility in how they build and source AI tools.
Building in-house gives teams more control, since models are trained and tested on their own claims, but it means investing the time and resources to develop, maintain, and improve the solution over time. Buying a solution provides access to specialized capabilities and ongoing support, reducing the effort of managing those demands in-house. With either option, claims teams still need to validate results and confirm the solution matches their workflow.
Medical record review is a good example. Relevant details can be buried across thousands of pages of handwritten notes, scanned forms and clinical records. Capturing the right information consistently is essential, because even a small missed detail can matter in the context of a claim.
In recent conversations, carriers with in-house AI work underway told us their early wins came on simpler tasks like document labeling, and that medical summarization raised harder questions about accuracy and consistency across very different document types.
These are the kinds of trade-offs we’re exploring in SiftMed’s upcoming build versus buy webinar. Our CEO, Holly Hill, and Chief AI Officer, Amir Tavafi, will cover what carriers are learning, what to watch for, and what to consider when deciding which route makes sense for your organization.
[inline-ad-1]
Claims teams trust AI outputs when they’re accurate and easy to verify. Guidewire made a related point in its keynote: agents need deep insurance context and trusted data to deliver results reliably. In our conversations, claims teams added a practical layer. They want to be able to check the output themselves.
Once people trust the output, the efficiency gains can follow. SiftMed helps claims teams save up to 70% of their medical record review time.
At Kenney Shelton Liptak Nowak, whose practice includes insurance defense, accuracy was a deciding factor in choosing SiftMed. Karl Daniel, a litigation partner, shared: “If we can review cases at a faster pace, without sacrificing accuracy, it’s a huge advantage for our clients.”
In claims review, adjusters need to know that diagnoses, treatment plans and provider details have been captured accurately before using them to inform claim decisions. If a summary flags a diagnosis or change in treatment, they need to trace it back to the original record, confirm the context is correct and ensure nothing important has been missed. Claims managers need that same visibility to review decisions and feel confident in the information their teams are working with.
Transparency matters when evaluating AI tools. Beyond the quality of a summary, claims teams need to see where the information came from, verify the details, and understand the surrounding context so nothing important gets overlooked.
Human judgment stays with the decisions. AI can do the reading and organizing, while adjusters assess the evidence and decide what happens next.
As more carriers are using AI in their claims work, the conversation has shifted towards where AI fits in existing workflows, which tasks it should handle, and when automation should stop and human judgment should take over.
Guidewire’s Designing the Future of AI-Native Claims Workflows workshop explored how teams can decide what AI should handle, when it should escalate, and how adjusters stay informed. A separate workshop on agent configuration addressed how teams set their confidence in automation and incorporate human oversight.
USAA’s session on Strengthening Human Connections Across 27 Million Claims focused on claims transformation. The goal was to make claims simpler for employees so they could spend more time working with members. Their approach kept technology in the background and people at the center of each claim.
Medical record review is another example of where balance matters. AI can help claims professionals work through extensive medical documentation and surface relevant findings, but it can’t replace the expertise needed to understand those findings in the context of a claim.
AI is evolving quickly across claims, and the next step for most carriers is deciding where it fits and what it takes to use it well. For me, one of the clearest takeaways from Connections was that trust is fundamental to adopting AI.
Claims professionals need to be able to verify AI-generated information before relying on it to inform decisions. The value of AI isn’t only in how quickly it produces an answer, but in how confidently claims teams can use that information in their work.
If you’re working through these questions for your claims team, I’d be glad to compare notes on what you’re considering and where specialist medical record review might fit.
Author: Rob Forsythe, GM, U.S. Insurance