Enterprise knowledge AI grounded on 2.4M policy documents at a national insurer
Underwriters and claims staff were working from conflicting versions of the same procedure. pronix.ai built a governed enterprise knowledge layer over 2.4M documents with ownership, freshness and citation enforced — 94% answer accuracy on the evaluation set and a citation on every single response.
National insurer, 14,000 employees · Azure AI Search · Azure AI Foundry · SharePoint · Guidewire
- Client
- National insurer, 14,000 employees
- Industry
- Insurance
- Platform
- Azure AI Search · Azure AI Foundry · SharePoint · Guidewire
*Representative outcome; results vary by client, scope and platform configuration.
The challenge
2.4M documents across SharePoint, a legacy DMS and team drives, with no authoritative version, no retirement process and state-specific variations buried inside national procedure files. Staff answered from memory or from whichever document search returned first, and audit had flagged inconsistent procedure application twice in three years.
Our approach
Governance before retrieval
Every source was classified as authoritative, reference or retire, with a named owner and review interval. 380,000 documents were retired before a single one was indexed.
Jurisdiction-aware chunking
State and line-of-business metadata is extracted and enforced at query time, so a Florida claims question cannot be answered with a Texas procedure.
Citations are mandatory
No citation, no answer. Responses link to the exact clause and show the document's effective date and owner.
Freshness as a first-class signal
Stale documents are down-ranked and their owners are notified automatically — the corpus improves without a content project.
A real evaluation set
1,600 questions written by underwriters and claims leads, scored each release by LLM-as-judge with human spot-checks. Nothing promotes below threshold.
“We stopped arguing about which document was right. That was worth more than the time savings.”
Take this case study into your next steering committee.
One page: the business problem, the solution, the technical architecture and the measured outcomes — sized to drop straight into a board pack. Share your work email to unlock it; one form unlocks every gated download on this site.
Board-ready briefs covering the platform and industry behind this program. Free with a work email.
Illustrative case study. Scenarios, metrics, quotes and client details are representative composites based on Pronix engagements and industry benchmarks unless a named client is shown with written consent. Outcomes vary by client, scope, data quality and platform configuration. Nothing on this page is a guarantee, warranty or professional advice. See our Terms of Use for the full disclaimer.
AWS, Salesforce, Microsoft, NICE, Genesys, Kore.ai, Google, ServiceNow, Amazon Connect and logos referenced on this site are trademarks of their respective owners. References are for descriptive purposes only and do not imply endorsement, sponsorship or partnership beyond stated partner relationships. See our Disclosures.
Talk to the team that shipped it.
Book a working session with a pronix.ai lead on this practice — we'll map the approach to your platform, industry and constraints.