Customer Stories
FM has delivered intelligence-backed commercial property insurance for almost 200 years, underpinned by its engineering risk assessments.
To surface operating standards data faster and better, FM worked with Microsoft and Spyglass MTG to develop an AI-powered search solution on Microsoft Azure.
Engineers can access information faster and more easily, enhancing the client experience.
In FM’s world, accuracy matters. Engineering‑led commercial property insurance depends on sound judgment, grounded in science‑based engineering expertise and deep technical knowledge. Underwriting decisions, risk assessments, and client recommendations rely on the precise and consistent application of that knowledge.
For almost 200 years, FM has built its competitive advantage on scientific research and engineering expertise backed by empirical data—captured and summarized in its engineering standards. These standards play a foundational role in underwriting decisions, risk assessments, and client recommendations. Their impact depends on how quickly and confidently engineers can access, assemble, and apply the right information at the moment it is needed.
To work faster without compromising rigor, FM partnered with Microsoft and Spyglass to build a secure, production‑grade AI‑powered search solution on Microsoft Azure. The goal was clear: help engineers surface relevant information and guidance for unique risk situations quickly, while maintaining the accuracy, transparency, and control required in high‑risk, complex industrial environments.
“The combination of Azure OpenAI in Foundry Models and Azure AI Search, together with reasoning built into the prompt engineering, stretched it as far as it can go—to maximum accuracy.”
Toby Denbow, Vice President of Analytics Technology and AI Engineering, FM
From the outset, FM set a clear standard for the project: accuracy is essential, because client trust is built on reliable, dependable answers.
In an organization where engineering recommendations directly affect client risk, accuracy is nonnegotiable.
The challenge was amplified by the scale and complexity of FM’s engineering content. The material spans tens of thousands of pages and includes intricate relationships, diagrams, flow charts, tables, and multi‑page technical guidance. Simply making the information searchable was not enough. An out‑of‑the‑box chatbot approach would not meet FM’s standards for reliability, transparency, or accuracy.
“We know that to reduce the risk of error, we need to give the AI only the information it needs to make the decision,” explains Toby Denbow, Vice President of Analytics Technology and AI Engineering, FM. “That meant being very precise about our search strategy, how information is assembled, and how content is retrieved.”
Working with Spyglass, FM designed the solution to reflect how engineers actually reason through problems. Information is carefully structured and chunked according to engineering logic, rather than by document format alone. Answers are validated against hundreds of known “ground truth” examples, and the system is continuously evaluated through built‑in governance and feedback loops.
This disciplined approach allows FM to maintain confidence in the results as standards evolve. The outcome is a precision solution designed to support engineering judgment—helping engineers be more efficient while preserving the rigor and trust their work demands.
For FM, AI does not replace expertise—it removes friction so expertise can shine.
Carlos Saenz is a Senior Engineering Specialist in FM’s Madrid office, where he focuses on assessing and improving risk for boiler and machinery assets. For much of his 19-year tenure, referencing standards meant navigating a library of more than 300 PDF files, some exceeding 100 pages. Searching required exact terminology and manual review, often under significant time pressure.
Today, AI powered search returns the answers Saenz needs immediately, along with the relevant context—often in a clearer and more usable form than before. During a recent multi-day visit to a mining facility, he had just five minutes to prepare a client presentation.
“Normally, that situation would be stressful,” says Saenz. “Now, I type in my query and quickly get back the right answer. I dropped it into PowerPoint and presented it to the client. It was perfect. It gives me confidence to know I have that AI-powered support when I need it.”
Like Saenz, FM field engineers have enthusiastically adopted the solution. In its first two months, it resolved more than 17,000 queries, with usage continuing to grow.
With 6–10 minutes saved per search, multiplied across more than 1,500 engineers, FM estimates the solution frees up thousands of engineering hours each year—time that can be reinvested in client engagement, risk improvement, and higher‑value work.
“It gives me confidence to know I have that AI-powered support when I need it.”
Carlos Saenz, Senior Engineering Specialist, FM
Solution supports more than 1,500 field engineers
When Andrew Bryson, now Senior Vice President of Engineering and Research at FM, began his career some 30 years ago, accessing standards looked very different. He carried 13 volumes of printed engineering guidance in the trunk of his car—a reflection of both the depth of FM’s knowledge and the effort required to apply it in the field.
“What we’re doing now is giving engineers focused access to the right information at the right moment,” says Bryson. “Using Azure AI Services this way puts a turbocharger on an engine that was already very powerful. It helps engineers apply their judgment more efficiently when they’re advising clients or making critical decisions in real time.”
Saenz agrees.
“Year after year, we give the best support to our clients so they know we are backing their progress,” he says. “We have to demonstrate every day that we can be trusted. This AI helps us do that.”
“Using Azure AI Services this way puts a turbocharger on an engine that was already very powerful.”
Andrew Bryson, Senior Vice President Engineering and Research, FM
FM’s ability to move quickly was grounded in decades of disciplined data stewardship. With deep datasets built by engineers for nearly 200 years, and more than 30 years of advanced analytics and predictive modeling, FM has the foundation to deploy AI with confidence.
The solution runs entirely within Azure, leveraging FM’s existing identity, security, and governance controls. “The combination of Azure OpenAI in Foundry Models and Azure AI Search, together with reasoning built into the prompt engineering, stretched it as far as it can go—to maximum accuracy,” adds Denbow. Centralized identity authentication, combined with Azure’s data protection and governance capabilities, helps ensure proprietary engineering standards remain protected and accessed only within approved environments.
“Our priority is ensuring data is handled in accordance with our security and governance model,” says Denbow. “With Azure, we have confidence that our data remains secure, governed, and auditable.”
17,000 queries made in first two months
Performance was a critical design consideration from the start. FM wanted answers returned in 15 seconds or less to ensure real‑world usability and strong adoption. Just as important, the solution had to deliver high availability and resilience, with no tolerance for downtime.
Built on Microsoft Azure, the architecture was designed for global scale and redundancy. Requests are routed to the least‑utilized regions to optimize performance, and multi‑region support ensures continuity in the event of an outage.
“Azure gives us the backbone to make it possible,” says Denbow. “When Azure experienced a localized production outage, our system automatically switched over to another Azure region and performed seamlessly—no user even knew it happened.”
FM worked closely with Spyglass MTG, leveraging its architectural expertise and experience delivering enterprise‑grade generative AI solutions. Together with Microsoft, the partners helped FM move quickly while maintaining rigor, governance, and trust.
Denbow adds, “Without the partnerships with Microsoft and Spyglass, I don’t think we would have been able to deliver this in the meaningful time that we did. Their guidance saved us time and improved reliability.”
“Without the partnerships with Microsoft and Spyglass, I don’t think we would have been able to deliver this in the meaningful time that we did. Their guidance saved us time and improved reliability.”
Toby Denbow, Vice President of Analytics Technology and AI Engineering, FM
Follow Microsoft