Microsoft published a customer story on May 28, 2026, detailing how FM, a US-based commercial property insurer, harnessed Azure OpenAI to give more than 1,500 field engineers AI-assisted access to internal standards. The initiative—built in collaboration with Microsoft and consulting firm Spyglass MTG—showcases a governed, enterprise-grade retrieval-augmented generation (RAG) system that transforms how field teams retrieve critical technical documentation.

FM’s challenge was as sprawling as its operations. Field engineers routinely inspected industrial facilities, power plants, and commercial properties, each governed by thousands of pages of internal engineering standards. Finding the precise guideline for a specific piece of equipment or risk scenario meant sifting through unwieldy PDFs, jumping between legacy knowledge bases, or radioing back to headquarters. Delays in retrieving accurate standards didn’t just slow productivity—they introduced consistency risks across global inspections.

With Microsoft and Spyglass MTG, FM architected a retrieval system anchored by Azure OpenAI and Azure AI Search. The system ingests FM’s proprietary standards documents, indexes them with semantic search capabilities, and serves answers through a natural language interface. Field engineers can now type a query like, “What clearance is required around a low-pressure steam turbine?” and receive a concise, citation-backed answer pulled directly from approved internal documents. No scrolling through 300-page manuals. No version-control guesswork.

The Heart of the System: Governed Retrieval-Augmented Generation

At its core, the FM solution is a governed implementation of RAG—a pattern that combines large language models with a curated document store. Unlike public chatbots that draw on unverified web data, FM’s system restricts the model’s knowledge to authorized internal content. This design eliminates hallucinations rooted in irrelevant or outdated public information, a crucial requirement for an insurer whose recommendations carry legal and financial weight.

Azure AI Search acts as the retrieval engine, chunking documents and ranking results based on semantic relevance to the query. The Azure OpenAI model then synthesizes a concise response, appending direct citations to the source documents. Engineers can click those citations to open the original standard, verifying context and reinforcing trust. This auditability is a cornerstone of the governance framework.

Spyglass MTG, a firm specializing in AI-powered knowledge management for regulated industries, contributed the governance layer that ensures every response adheres to FM’s compliance and security policies. The system enforces role-based access controls—an engineer performing routine maintenance inspections sees only the standards applicable to their role and region, while a senior risk consultant might access the full global catalog. All queries and responses are logged for audit trails, and the platform automatically tracks document versions, so answers always reference the latest approved standard.

From Pilot to Production: Scaling Across 1,500 Engineers

FM’s rollout to over 1,500 field engineers didn’t happen overnight. The team first piloted the system with a small group of senior engineers, refining prompt design and response formatting. Early feedback drove several key enhancements: engineers wanted answers in a consistent template—problem statement, relevant standard clause, and practical interpretation. The system was tuned to deliver exactly that, using Azure OpenAI’s fine-tuning capabilities on historical question-answer pairs.

Connectivity was another hurdle. Field engineers often work in environments with limited internet access—deep inside boiler rooms or remote oil platforms. By deploying a subset of the system on edge devices using Azure Arc and caching frequently accessed standards locally, the team ensured offline resilience. This architecture keeps the AI responsive even when connectivity drops, syncing logs and updates once back online.

Governance That Meets Insurance-Grade Scrutiny

For a commercial property insurer, governance isn’t a checkbox—it’s a business imperative. FM’s recommendations influence underwriting decisions, liability assessments, and regulatory compliance. Any AI-generated answer must be as accurate and defensible as one provided by a senior engineer manually consulting a printed standard.

The governance framework operates on multiple layers. At the data layer, document ingestion pipelines validate source authenticity and enforce encryption at rest and in transit. At the search layer, Azure AI Search’s built-in security filters ensure engineers never see documents outside their clearance level. At the inference layer, Azure OpenAI’s content safety mechanisms screen inputs and outputs for inappropriate or off-policy content. All usage is monitored through Azure Monitor and Microsoft Sentinel, with dashboards alerting administrators to unusual query patterns that might indicate misuse or data exfiltration attempts.

Crucially, the system never trains on FM’s proprietary data. Azure OpenAI operates in a stateless mode, with no customer data used to fine-tune base models. Spyglass MTG designed the retrieval pipeline so that each query triggers a fresh search against the current document index, completely independent of previous interactions. This design choice preserves data residency requirements and aligns with FM’s stringent data governance policies.

Measurable Impact on Engineering Efficiency

Microsoft’s customer story reports several quantitative gains. Average time to locate a specific standard dropped from 12 minutes to under 90 seconds—an 87% reduction. First-response accuracy, measured by engineers marking answers as “fully resolved” without follow-up searches, climbed from 63% to 94%. Together, these metrics translate to an estimated 18,000 hours reclaimed per year across the field engineering organization.

Beyond efficiency, FM observed qualitative improvements. New engineers ramp up faster, learning standards through natural inquiry rather than rote memorization. Senior engineers spend less time answering repetitive reference questions, freeing them for higher-value tasks like complex risk assessments and mentoring. And with every query tied to a document citation, FM gains rich analytics on which standards are most frequently accessed—data that informs future standards development and training priorities.

Broader Implications for Regulated Industries

FM’s deployment signals a maturing approach to enterprise AI, moving beyond generic chatbots to domain-specific, fully governed knowledge systems. Insurers, pharmaceutical companies, energy firms, and other compliance-heavy sectors share a common need: making massive proprietary document sets instantly accessible without sacrificing accuracy or auditability. The FM pattern—combining Azure AI Search, Azure OpenAI, and a robust governance overlay—offers a repeatable blueprint.

Microsoft’s publicizing of this story also underscores Azure’s enterprise AI differentiation. By leaning on native services like Azure AI Search’s semantic ranking and Azure OpenAI’s secure deployment boundaries, organizations can build solutions that satisfy even the most risk-averse stakeholders. Spyglass MTG’s involvement highlights the growing ecosystem of niche consultants who can tailor these platforms to industry-specific governance requirements, accelerating time-to-value.

What’s Next for FM and Enterprise AI

FM’s roadmap includes expanding the system to support mobile and wearable devices, enabling engineers to ask questions hands-free while inspecting equipment. Plans also include integrating telemetry data from IoT sensors on insured properties, so an engineer can ask, “What standard applies given the vibration readings on this pump?” and receive a contextual answer that blends live data with document knowledge.

For Windows-centric enterprises observing this story, the integration with Azure services demonstrates how the Microsoft cloud stack can anchor mission-critical AI workloads. Engineers in the field often rely on Windows-based laptops, tablets, and ruggedized devices, and the solution’s alignment with Microsoft’s security and management ecosystem simplifies IT administration. As governed AI becomes a competitive necessity, stories like FM’s provide a practical reference for what works—and what governance really demands.

Microsoft’s customer story is available on the Azure blog and through the company’s AI customer success portal, offering additional technical deep dives for organizations considering similar implementations.