Pipelinev0.4.0-next.5
Evidence-linked claims operations

Data to decisions.

Turn incoming demand files into source-linked, reviewer-controlled claim decisions.

Compare document versions, verify material facts against their source, control response dates and exact settlement conditions, and preserve every recorded reviewer decision.

For carrier claims teams, examiners, and TPAs handling represented bodily-injury demands.

01 Intake02 Reconcile03 Decide04 Record
Product film · 36 seconds

See the control loop in motion.

From document intake to source-linked review and a recorded human decision. Illustrative scenes only—no customer or claimant data.

01 · Intake02 · Verify03 · Control04 · Record
Control 01

Deadline control

Keep response dates and exact settlement conditions visible while exceptions move through review.

Control 02

Evidence-linked review

Move from a material fact to its source and keep missing or conflicting evidence review-required.

California bodily-injury claim scale

Small handling gains compound across a large claim stream.

California Department of Insurance private-passenger-auto data for 2022 reports bodily-injury liability claim counts and incurred losses. The figures below are calculated from that official workbook; they are not settlement forecasts.

285,954California bodily-injury liability claims in 2022
1.8 minAverage interval between reported claims
$21,187Average incurred loss per claim
14,298Claims of annual capacity represented by a 5% handling-efficiency improvement

Method: 285,954 claims, 28,941,960 earned exposures, and $6,058,416,962 incurred losses. “Capacity” is 5% of annual claim count, not a savings or outcome guarantee. California Department of Insurance source and workbook.

The plaintiff technology curve

Plaintiff firms are already building an AI-enabled operating advantage.

Purpose-built platforms now help plaintiff teams structure medical records, build chronologies, draft demands, identify missing material, and prepare cases. Carrier teams cannot treat comparable evidence controls and workflow modernization as a distant roadmap item.

EvenUp · vendor reported$235M total funding

EvenUp reported more than 1,000 personal-injury law-firm customers when it announced its 2024 Series D and expanded its AI product suite.

Read the funding announcement
Supio · vendor reported$91M total funding

Supio describes a plaintiff-side platform for personal-injury and mass-tort firms spanning document intelligence, medical data, chronologies, and case preparation.

Read the funding announcement
Carrier responseComparable speed needs stronger controls.

Insurers need source-linked facts, explicit uncertainty, deadline ownership, human approval, and a durable decision record—not ungoverned automation.

Evaluate the workflow
Illustrative product view

One file. Recorded changes. Recorded decisions.

A fictional claim demonstrates version comparison, completeness review, source verification, and examiner-controlled action. It prepares internal records only—no live carrier dispatch or claim-system writeback. No customer or claimant data is shown.

Pipeline ClaimsClaim CLM-2026-1042Saved
Response due12 daysDocument setVersion 3Missing evidence2 itemsSource review8 / 10
What changed · 3 facts affectedSupervisor exception · 1 open
Evidence graphWhat changedDemand conditionsTreatmentEvaluation
Version comparison

Follow-up treatment added in document set v3

The affected treatment fact remains review-required until the examiner confirms the new source page.

Source page · 18
Illustrative interface using a fictional claim. Product configuration varies by deployment.
One controlled workflow

From incoming document set to documented action

  1. 01
    PDF

    Version intake

    Preserve submitted PDFs, detect duplicates, compare later documents, and capture deadlines and conditions.

  2. 02

    Reconcile the file

    Connect facts to evidence, identify what changed, and keep missing or conflicting material visible.

  3. 03
    !

    Control the response

    Confirm source-read values, assign mandatory conditions, and route operational exceptions.

  4. 04
    Approved

    Approve the action

    Prepare authority and administrative records with human approval and immutable lifecycle events.

Proof without invented promises

Inspect a complete fictional file

The evaluation workspace includes fictional demand fixtures with police-report, treatment, billing, and exact-settlement-term evidence. Use them to evaluate version-aware intake, field coverage, source navigation, and reviewer controls without using claimant data.

Review a demand package
Fictional
demand fixtures
10
source verification checks
1
versioned review trail
Security posture

Controls before automation.

Tenant boundaries, source fingerprints, retention controls, and a local deidentification boundary are designed to keep claim evidence organization-scoped and reviewers in charge.

Tenant and role boundaries

Claims APIs enforce organization membership and permission checks for review, billing, analytics, exports, and lifecycle actions.

Source integrity

Private source documents are bound to claims with SHA-256 fingerprints; reattached evidence must match the stored source.

Retention and audit controls

Case retention, legal hold, reviewer changes, and workflow events are designed to remain durable and organization scoped.

Human-controlled actions

Authority, correspondence, and audit records are prepared with human approval and immutable lifecycle events—no live carrier dispatch or claim-system writeback.

ClaimSafe model boundary

Internal, authenticated provider-bound text is deidentified locally into request-scoped tokens.

  • NER must load before every provider call, and only the local restoration map is AES-256-GCM encrypted; providers do not receive the map or key.
  • PDFs use bounded local text extraction and OCR, while raw PDFs and page images are blocked from provider egress.
  • Unreadable or low-confidence pages fail closed, and visual-only evidence remains human-review-only.
  • The public ChatGPT file tool is disabled because ChatGPT would receive an original first; production private ChatGPT claim tools remain disabled pending opaque capability handles.
  • The production deployment uses verified database transport, encrypted private storage and temporary volumes, allowlisted provider routing, access controls, monitoring, incident procedures, and organization-scoped approval.
  • No SOC 2 certification, HIPAA compliance, or production fitness is claimed; no BAA coverage or zero-retention claim is made by the software itself.

NAIC AI governance packet

A procurement-ready vendor packet maps model inventory, human oversight, validation and drift monitoring, change controls, incident support, and proposed audit-cooperation terms to the NAIC Model Bulletin. It supports—not replaces—an insurer’s AIS Program.

Review the governance approach
Claims pricing

Contract plans matched to the operation

Examiner

For an individual examiner working demands end to end.

Contact sales
  • Individual demand review
  • Evidence-linked PDF controls
  • Claim-file exports
Discuss Examiner

Enterprise

For carriers with procurement and security-review requirements.

Contact sales
  • SSO and deployment planning
  • Retention and legal-hold controls
  • Volume and support terms
Discuss Enterprise

Pricing and document-processing credits are provisioned by contract; the workspace does not present an unannounced self-service subscription. No public price or ROI promise is represented as validated. Pipeline provides review assistance and does not guarantee claim outcomes, savings, or regulatory compliance.

Talk to Pipeline

Bring one workflow you want to improve.

Tell us about your claims operation, current demand volume, and the controls your security team requires. We’ll use a fictional demand in the demonstration.

Book a product conversation

Email the product owner with your organization, role, and preferred meeting windows.

Request a demo by emailNo claim documents or sensitive claimant information, please.