- Company
- Regional insurance provider
- Industry
- Insurance
- Process automated
- First Notice of Loss (FNOL) intake, claims categorization, risk detection, prioritization, routing, claimant communication, and adjuster assignment
- Core systems
- Zenphi, Gmail, Google Drive, Google Workspace, and Salesforce
- Volume
- Approximately 1,000 FNOL submissions per month
- Results
- 250+ hours of manual work saved monthly, $320,000 in annual savings, triage reduced from 15 minutes to 15 seconds per claim, and critical-claim response reduced from days to minutes
This insurance AI implementation focused on one operational bottleneck: high-volume FNOL intake and triage. FNOL submissions arrived through Gmail and included everything from form data to police reports and incident images. The team built the entire AI-powered claims workflow in Zenphi. It extracts claim details, enriches them with policy data, classifies urgency and risk, and routes each claim into the appropriate operational track.
1,000 monthly claims were entering a shared inbox faster than staff could triage them
The provider was managing roughly 1,000 First Notice of Loss submissions every month. Claims arrived in a shared Gmail inbox, where they could sit for days while staff worked through the existing queue.
The bottleneck was not simply message volume. The intake itself was difficult to standardize because claims contained unstructured information in multiple formats — including emails, PDFs, ACORD forms, incident photos, and handwritten police reports.
Administrative staff spent about 15 minutes manually reviewing and categorizing each claim before it could move forward.
The source case study reports manual errors in roughly 60% of entries, creating inconsistent categorization and routing.
The case study reports that 31% of policyholders were frustrated by slow triage, with settlement speed cited as a key pain point.
The team needed to extract critical details from police reports, incident photos, attached documents, and free-form email content — not only structured forms.
Claims arrived through Gmail and related documents were stored in Google Drive, while policy information was maintained in Salesforce.
The insurer wanted the workflow to absorb periods such as floods or wildfires without a pricing model tied directly to every additional task or user.
An AI claims workflow built in Zenphi for FNOL intake, categorization, and routing
The insurer’s Digital Transformation team built the complete claims workflow in Zenphi to process incoming FNOLs from the moment they arrive in Gmail. AI is used inside the workflow to interpret unstructured claim data, while Zenphi orchestrates the deterministic business logic, categorization, routing, notifications, assignments, and downstream actions.
1. Intake and unstructured data extraction
When an FNOL arrives in Gmail, the workflow reads the unstructured email body and supporting documentation. It extracts information such as claimant details, site IDs, and incident severity from sources including police reports and ACORD forms.
2. Salesforce enrichment
The claim is enriched by cross-referencing the claimant’s email against Salesforce to retrieve the relevant policy tier. This gives the workflow additional context before routing decisions are made.
3. Multi-stage categorization
The AI classifies claims by urgency — Critical, High, Medium, or Low — based on damage descriptions and mentions of injury. It also detects high-risk language such as structural compromise or litigation threats.
The claim enters through Gmail with its email content and supporting documents.
The AI identifies claimant details, site information, severity indicators, and other relevant facts from the submission.
The workflow looks up the claimant and pulls the policy tier needed for downstream handling.
The agent categorizes the claim and detects signals that should bypass normal queues.
Lower-complexity claims follow an efficiency path, while high-risk claims trigger priority handling.
The workflow communicates with claimants, assigns internal tasks, alerts senior adjusters, and schedules urgent field activity where required.
Low-complexity claims move fast; high-risk claims bypass the standard queue
AI interpretation combined with deep Google Workspace actions and predictable workflow pricing
The provider needed more than a model that could summarize text. The automation had to work inside the Google Workspace environment, connect to external systems, and continue into operational actions after the AI made a classification.
$320,000 in annual savings and 250+ hours of manual work recovered each month
| Metric | Before automation | With the Zenphi AI workflow |
|---|---|---|
| Administrative labor | 250+ hours per month spent on manual triage | Automated claims triage, representing $120,000 in annual savings |
| Triage time | About 15 minutes per claim | About 15 seconds per claim |
| Denial rework | $25–$117 per claim | 95%+ accuracy reported in the case study, representing $200,000 in annual savings |
| Critical-claim response | Claims could remain in the queue for days | Priority cases routed and responded to within minutes |
| Auditability | Manual handling and inconsistent prioritization | Permanent searchable audit trail showing why a claim was routed and who handled it |
| Total annual value | Manual triage and rework costs | $320,000 in annual savings |
From manual review to near-instant triage
By automating roughly 1,000 claims each month, administrative staff moved from approximately 15 minutes of manual triage per claim to about 15 seconds. Every claim now has a searchable audit trail documenting why it was routed and who handled it.
For critical claims, the response window dropped from days to minutes because high-risk signals can bypass the standard queue and trigger immediate escalation.
What this case study on AI implementation in insurance demonstrates
This is not a standalone AI model or chatbot. The AI is embedded inside a workflow built in Zenphi: AI interprets messy, unstructured inputs and identifies relevant facts, while Zenphi applies the business logic that determines what should happen next.
Want to see how this type of AI claims workflow is built?
Bring one intake, categorization, or routing process that still depends on manual review. We’ll show you how AI and deterministic workflow logic can work together in Zenphi.
Book a platform demo