.png?width=300&name=progressive-fleet%20-%20black%20(1).png)
Bevaya’s AI agents reduced our claims indexing from 5 days to under an hour. The accuracy was better than our most experienced staff on day one.
AI Models
The insurance-native AI inside Bevaya.
InsurGPT™ is a mosaic of specialized models, trained on proprietary insurance data general AI labs can't access. It's the intelligence behind every AI Agent on the Bevaya platform.
Reading:
Demand letter
Trusted by leading P&C carriers, brokers, and TPAs
Production reality
Built and proven at scale
Insurance documents trained on
300M+
Accuracy on insurance tasks
98%+
Straight-through processing
70%+
Production deployments
115+
Why specialized AI
General AI takes you only so far.
Insurance needs more.
General AI is good at many things. Understanding insurance — the documents, the language, the way claims and underwriting actually work — isn't one of them. That takes models trained for the work, by people who do it every day.
Trained on documents the rest of the world can't see.
300M+ proprietary insurance documents — supplemental applications, loss runs, demand letters, claim forms. The data Google, OpenAI, and Anthropic don't have access to.
Specialists for specialist work.
One model for each insurance job. The Loss Run Model reads loss runs. The All ACORDs Model reads ACORDs. The GutenOCR Model reads the messy stuff. Each is the best at one thing.
Engineered for high-stakes accuracy.
Every answer is checked by a separate Verifier model, grounded to its exact spot in the source, and scored for confidence. Auditable by design.
“
99% straight-through processing and 246% ROI in just 6 months.
F5
Fortune 500 Carrier Property & Casualty
“
We evaluated six AI vendors. Bevaya was the only one that understood our underwriting workflows from day one — no six-month education period.
CU
Chief Underwriting Officer Specialty Carrier · $2B+ GDP
“
The confidence scoring changed everything. Our reviewers know exactly which items need attention and which can go straight through. We trust it.
DO
Director of Operations Workers’ Compensation Insurer · 74 NSP
“
Cut COI turnaround from 24 hours to minutes and saved millions.
T5
Top-5 Broker National
MEET THE MODELS
A powerful engine of
insurance-trained models
Each model is trained for a different part of insurance work.
The engine combines them as the job requires.
Reads loss history
Across hundreds of carrier formats.
Extracts ACORDs
Every variant with 99%+ accuracy.
Traces every answer
Shows where each answer came from.
Separates documents
Splits merged PDFs into the right files.
Reads the unreadable
OCR that reads what others can't.
Extracts any form
Any structured form, including medical.
The Benchmarks
When put to the test, InsurGPT™ beats general AI on every insurance task.
Specialized models outperform general-purpose ones on the work your team actually handles. Here's the head-to-head.
On claims accuracy, InsurGPT™ scored 99%. The strongest general model managed 62%.
A 37-point gap on the work your team handles every day — claims indexing, FNOL, demand letters, medical bills.
Source: Roots benchmark tests, December 2025. View methodology →
Claims Accuracy
- InsurGPT™
- 0%
- Mistral AI
- 62%
- GPT-5.0
- 58%
- Gemini 3.0 Pro
- 55%
On underwriting accuracy, InsurGPT™ scored 93%. The strongest general model reached 84%.
9 to 13 points ahead of GPT-5.0, Gemini 3.0 Pro, and GPT-4.1 — across submission intake, loss runs, and exposure schedules.
Source: Roots benchmark tests, December 2025. View methodology →
Underwriting Accuracy
- InsurGPT™
- 0%
- Gemini 3.0 Pro
- 84%
- GPT-4.1
- 81%
- GPT-5.0
- 80%
Insurgpt
Why insurance-native AI is
so much more accurate
Specialized training is the start.
What surrounds the models is what makes them production-grade.
Architecture | Composite
Proprietary insurance models combined with the best of frontier AI. The platform picks whichever delivers the highest accuracy for each job — not whichever one the vendor happens to sell.
Reliability | Verified
One model produces the answer. A second model audits it against the source. Hallucinations get flagged before they reach your team — not after.
Vision | Grounded
GutenOCR reads logos, checkboxes, stamps, and complex forms — with exact spatial coordinates kept intact for full traceability. Reviewers can click any element and jump to where it lives on the page.
Confidence | Calibrated
Confidence scores you can actually trust. Set thresholds for straight-through processing — and reach 70%+ STP within 90 days.
Speed | Dedicated
3–5× faster than GPT-5 and Gemini 3. Dedicated infrastructure means predictable SLAs and no shared-API latency. Faster time to quote. Faster time to settle.
Human-in-the-Loop | Platform
A real platform for exceptions — not a queue. Reviewers see the source, the conflict, and the right action in one place. Every correction also trains the models, so accuracy compounds on your team's work.
Powered by InsurGPT™
The model behind every Bevaya agent
Insurance documents trained on
300M+
Insurance-specific fields the model understands
1,200+
Document formats handled
ACORDs
SOVs & schedules
Loss runs
Declarations pages
Endorsements
Adjuster notes
Fields understood across the policy lifecycle
Named insured
Risk location
Limits & TIV
Coverages & forms
Effective dates
COPE attributes
Claim cause & status
Loss history & reserves
+ more
InsurGPT™
Purpose-built for insurance. From submission to claim close, InsurGPT™ reads the documents your teams already work from — ACORDs, SOVs, declarations, endorsements, loss runs, adjuster notes — and turns them into structured, traceable data.
98%+
extraction accuracy across the P&C document set
Transparent by design
Every answer shows its work.
Bevaya Labs
Insurance AI is its own science. Bevaya Labs is doing it.
Our applied research team — AI scientists and insurance veterans — builds the methods general AI labs don't: post-training for insurance reasoning, grounded vision for insurance documents, and the benchmarks that prove it. We publish what we learn.
Trust & Security
Trust by design
Built for an industry where data security isn't optional.
Data ownership
Your data stays yours.
Never shared with other customers or vendors. Bevaya doesn't train shared models on your data.
Visit the Trust Center
Your tenant
No training
Logical isolation
Role-based access
SSO + SCIM
Customer-managed keys
Your data · only your team sees it
Compliance
Encrypted end-to-end.
256-bit AES encryption, in transit and at rest. Independent third-party audits conducted annually.
Visit the Trust Center
SOC 2 Type 2
HIPAA
GDPR
CCPA
23 NYCRR 500
AES-256
Audited annually · independent third party
Deployment
Runs in Azure.
Enterprise-grade infrastructure, hosted where insurance organizations already trust their data.
Visit the Trust Center
Microsoft Azure
Azure Marketplace
AWS
Private VPC
Google Cloud
GCP-native
Azure Marketplace
Guidewire Marketplace
Deploy where your stack already lives
Oversight
Every decision audited.
Immutable audit logs. Confidence scoring. Human-in-the-Loop review on low-confidence items.
Visit the Trust Center
AI extracted limits from ACORD 125
98% conf.
Reviewer confirmed coverage
Approved
Endorsement flagged for review
62% · HITL
Policy match validated
95% conf.
Audit log written · immutable
Sealed
Immutable trail · every decision, every reviewer
More Capabilities
Explore the rest of the platform.
Designed, deployed, and governed together. Powered by InsurGPT™ and accessed through the AI Assistant.
Canvas
Workflow Canvas
Visual builder and production runtime for every automation.
Current page ReviewHuman-in-the-Loop
Configurable review queues with X-Ray verification and a patented feedback loop.
Current page DocumentsDocument Intelligence
Read any insurance document — hundreds of carrier formats, scanned or digital.
Current page GroundingGrounded Explainability
Every value traceable to its source. X-Ray Highlight Mode brings citations to reviewers.
Current page AnalyticsAnalytics Dashboard
Live accuracy, STP rates, reviewer SLA, and agent performance across every workflow.
Current page GovernanceGoverned Automation
Immutable audit trails, role-based access, flow versioning. Compliance is the architecture.
Current page
The AI behind the platform
Insurance-Native AI · Engine
AI Assistant · across the platform
InsurGPT™
A mosaic of specialized insurance AI models, trained on real insurance data.
Current pageAn AI agent across the platform
Builds your workflows. Executes work alongside you. Explains every step.
Current pageFAQ
Frequently asked questions.
InsurGPT™ is insurance-native AI. It combines proprietary models trained on 300M+ insurance documents with the best of frontier AI — selecting the right model for each task. On insurance benchmarks, InsurGPT™ beats GPT-5.0, Gemini 3.0, and Mistral on every task we've tested. Just as important, InsurGPT™ surrounds the models with verification, grounding, calibrated confidence, and continuous learning — layers general AI doesn't have.
Yes — for a demo. A general LLM can find a claim number when it's printed clearly on a clean document. That solves about 20% of claims indexing and 0% of FROI. Production needs more: a Human-in-the-Loop platform for exceptions, business rules for when claim numbers are missing or conflicting, page-splitting for multi-document PDFs, grounded extraction reviewers can audit in seconds, calibrated confidence with adjustable thresholds, and a Verifier model that audits the AI's own work. That's the difference between an extraction tool and a production-grade AI Agent — and what gets deployments to 70%+ STP.
InsurGPT™ is the AI engine inside the Bevaya platform. The platform turns the engine into AI Agents that read, route, decide, and write — automating work across claims, underwriting, and policy servicing. You design the agents in the Workflow Canvas, deploy them across the insurance value chain, and govern them with human-in-the-loop review, confidence thresholds, and audit trails. One place for all three.
Your data stays yours. InsurGPT™ runs in your cloud — AWS, Azure, or GCP — and we don't train shared models on your data. Bevaya is SOC 2 Type 2 certified and HIPAA, CCPA, GDPR, and 23 NYCRR 500 compliant. Federated learning improves the platform without ever exposing your specific documents.
Most customers go live in 8–12 weeks. Pre-built AI Agents — powered by InsurGPT™ — handle the most common claims, underwriting, and policy servicing tasks out of the box. Bevaya is available on Azure Marketplace and Guidewire Marketplace.
Yes. Beyond the pre-trained models, InsurGPT™ can be tuned on your carrier-specific formats and workflows. Our Human-in-the-Loop review experience captures expert corrections and feeds them back into the models — so accuracy improves the more your team uses it.
GET STARTED
Ready to design, deploy and govern
your AI workforce?
Bevaya AI Agents can help you triage, analyze, and recommend across underwriting, claims, and policy servicing. Let's connect and show you how it works.