Use cases
Use cases start from a concrete insurance operations problem and show which Bevaya pattern solves it — the template to start from, the nodes that do the work, the integrations involved, how a person stays in the loop, and how to track the work once it's running.
Every use case follows the same shape under the hood. A flow is triggered, documents are read and prepared, an AI step classifies and extracts what matters, the flow routes or matches the result, a person reviews exceptions, and the item is finalized. None of the templates ship wired to a particular system of record — if you want a result looked up or written back to your system, you add that connection yourself. The building blocks are the same across use cases; what changes is how they're composed and configured. See Platform concepts for the underlying model and Build AI Agents for how flows are built.
How to choose a starting pattern
Pick a use case by matching four things to your situation:
- The operational problem — what work are you trying to take off a person's plate?
- The source documents — what arrives, and in what shape?
- A system of record, if you want one — do you want to connect the flow to look something up or write a result back, and to which system?
- The review needs — how much human confirmation does the work require before it's trusted?
The patterns documented today are below. Each links to its full guide; start there for the template details, the released nodes it uses, the integrations to configure, the review model, and how to track runs.
Underwriting
Submission Ingestion & Clearance — A multi-line commercial submission arrives as separately-filed documents (application, coverage-line supplements, schedules, loss runs, SOV). The flow classifies and extracts each one, then synthesizes package-level clearance, appetite, completeness, and exposure findings. Every submission routes to an underwriter for review and approval.
Triage and Routing — A commercial P&C broker submission package arrives as an email with attachments, a zip, or one large PDF. The flow classifies and extracts every document, reconciles one account view, then applies a deterministic triage engine covering clearance (a placeholder in this version), completeness, appetite across nine lines of business, complexity tier, and priority score.
Loss Run Processing — A loss run report needs its policy and claim history turned into structured, trustworthy data before anyone can price or reserve against it. The flow extracts policy- and claim-level tables, resolves calculated figures, checks the result against financial and business rules, and uses an AI step to distinguish genuine errors from known carrier-specific reporting conventions.
Rating & Quoting — A cleared submission still needs to be rated, its loss history stratified, and a quote drafted before an underwriter can decide. The flow reconciles the account, rates every requested line, checks the result against authority, and drafts a two-option quote. Standard quotes and referrals route to separate review queues; every quote is reviewed before release.
Policy Renewal Handling — An expiring policy needs the same re-underwriting pass every year, against a statutory notice clock. The flow reconciles expiring against renewal exposures, checks the deadline, prices the renewal, and segments the account into Auto-Renew, Low-Touch, or High-Touch. Only clean, confident accounts complete without review.
Policy-to-Policy Comparison — A renewal or replacement policy needs to be checked against the one it's replacing, attribute by attribute, before anyone signs off. The flow extracts both policies, compares every attribute, and classifies each change as a reduction, an improvement, or no change. Every comparison goes to a reviewer before it's finalized.
Claims
Claim-to-Policy Comparison — A new loss is reported as an FNOL package. The flow classifies and extracts every document — including any policy documents you provide — and assembles an AI coverage comparison and a recommended reserve as an adjuster briefing. It works entirely from the documents in the package; connecting it to your carrier system to create the claim or write back is something you add yourself. The adjuster approves before anything is finalized — the comparison is decision support, not a determination.
Claim Indexing — Inbound claim mail (medical bills, legal correspondence, reports, photos) needs to be sorted, identified, and captured before anyone can act on it. The flow classifies each document, extracts the fields that identify it, and surfaces a package briefing with any urgent flags. It runs entirely on the uploaded content; filing the package against a claim in your system is something you add yourself. Every package goes to a person for confirmation.
Claim File Summarization — Someone needs to come up to speed on a claim quickly — an adjuster handoff, a supervisor review, an exposure audit. The flow reads the whole claim file and produces a one-page briefing (loss event, investigation, financials, coverage, open actions). No system match; the summary lives on the item and can optionally be written back. A reviewer confirms it.
Medical Bills Identification and Extraction — Providers often send several bills bundled into one document, and downstream systems need each one split out and filed to its own claim. The flow uses Split Documents to separate the bundle, then classifies, extracts, and reconciles each bill's charges against its stated total. Integrate it with your claim system to search for and index each bill to the right claim, or with your policy system to assess coverage. The agent pauses for review when a bill's reconciliation doesn't match or a required field is missing.
Invoice Payment Processing — Vendor invoices from defense counsel, IMEs, investigators, restoration vendors, and repair shops arrive against open claims and need to be indexed to the right claim file, checked against the charges, and routed for payment. The flow extracts the invoice details and every line item, reconciles the line items against the invoice's stated total, and flags risk such as duplicate or out-of-scope charges. Integrate it with your claim and fee-agreement systems to add rate and authority checks; the agent pauses for review when the reconciliation doesn't match or a required field is missing.
Legal Demands Identification and Extraction — Demand letters and other legal correspondence arrive in the same claim mail queue as everything else, each with its own deadline, parties, and fields to capture. The flow triages the package, decides whether it's a demand, classifies demands into one of 17 document types, and extracts the fields specific to that type. Integrate it with your claim system to match the package to an open claim; the agent pauses for review when a field is missing or a value is low-confidence.
Policy Servicing
Endorsement Processing — A broker's policy change request arrives as a mixed package and has to be evaluated against the policy — in-force status, backdating, completeness, premium impact — before it's actioned. Clean requests complete straight through; anything else routes to a policy servicing rep with the specific issue already identified.
COI Issuance — A third party requests a certificate of insurance, and it has to be verified against the policy — limits, special status, in-force — before it can be issued. Clean requests issue straight through; anything else routes to a policy servicing rep with the specific gap already identified.
Premium Audit Processing — A workers' compensation premium audit submission arrives as payroll registers, audit schedules, owner/officer schedules, tax returns, and financials. The flow classifies and extracts every document, reconciles company identity, payroll, and class codes across the package, and surfaces a completeness check and audit exceptions. Every analysis goes to an auditor before it's finalized.
At a glance
| Use case | Source Document Examples* |
|---|---|
| Submission Ingestion & Clearance | Commercial submission (main application, ACORD-style coverage-line supplements, vehicle/driver schedules, loss runs, SOV) |
| Triage and Routing | Broker submission package (ACORD, exposure schedule, loss runs, broker email, supplemental, financials) |
| Loss Run Processing | Loss run report (policy and claim history by line of business) |
| Rating & Quoting | Cleared submission package, expiring declarations, loss control report |
| Policy Renewal Handling | Renewal file (expiring declarations, internal loss runs, premium audit, endorsement history, renewal application) |
| Policy-to-Policy Comparison | Two policies (expiring and renewal, proposed, or replacement) |
| Claim-to-Policy Comparison | FNOL package + policy declarations |
| Claim Indexing | Inbound claim correspondence |
| Claim File Summarization | A full claim file |
| Medical Bills Identification and Extraction | Medical bill package (CMS-1500, UB-04, itemized, pharmacy, dental) |
| Invoice Payment Processing | Vendor invoice (defense legal, IME, investigator, restoration, repair) |
| Legal Demands Identification and Extraction | Legal correspondence (demand letters, liens, court filings, and related document types) |
| Endorsement Processing | Broker change request package (request email, change form, vehicle/property/driver schedule, contract or lease, payroll or exposure update) |
| COI Issuance | Certificate request package (request email, contract or requirements schedule, prior certificate, declarations page, endorsement form) |
| Premium Audit Processing | Premium audit submission (payroll schedule, audit schedule, owner/officer schedule, tax return, financials) |
* These are the source documents each template ships configured for out of the box — configure a flow to work with any document types your use case needs.
This catalog grows as more templates are documented. If a workflow you need isn't listed here yet, you can still build it from scratch — see Create an AI Agent.
What's common to every use case
- Triggers. Each template ships with Manual File Upload; you can configure other triggers if you'd like, including a connection to an Outlook email inbox, and runs can be started programmatically through the Bevaya API. See Configure triggers.
- Nodes. Classification and extraction use InsurGPT nodes; routing and looping use Control nodes; file prep, validation, API calls, and human review use Utility nodes; and the item is closed with an Action node. If you add a connection to your own system, App integration nodes cover common systems like Outlook and Guidewire. The Node catalog is the full map.
- Human in the loop. Exceptions and approvals route to a person via Review; reviewers work the Insights and Review tabs on the item. See Human review.
- Tracking. Watch runs in Run history and counts by status in Item status reporting.
- Roles. Building and running flows requires builder permissions; reviewers work items and reviews but don't build. See Users and access.
Where to go next
- Build AI Agents — how flows are built, tested, published, and run.
- Create an AI Agent — start from a template or build from scratch.
- Monitor and Review — run history, human review, and item status.
- Bevaya API — drive any of these flows programmatically.
- Platform hierarchy and Users and access — set up your organization before you build.