About

Building insurance document infrastructure for the agent era.

PolDex is built by Olamilekan Akinuli to replace manual insurance document operations with deterministic extraction infrastructure. The first wedge was Commercial P&C; the release-ready product now spans the full 56-schema insurance universe with 99%+ benchmarked accuracy.

The product

PolDex is the 99%+ accurate vertical API for insurance extraction: documents in, evidence-backed structured data out.

The wedge

Commercial P&C was the first wedge because it is document-heavy, painful, and full of COIs, policies, endorsements, schedules, submissions, loss runs, and SOVs.

The expansion

The same rail now spans consumer P&C, claims, life, health, benefits, reinsurance, specialty, ACORD, and document primitives.

The boundary

PolDex does not sell insurance or own the workflow. It supplies insurance truth to the systems and agents that already do the work.

System

Insurance documents enter as files, URLs, text, email, or agent calls.

PolDex returns deterministic JSON, evidence, conflicts, exports, and artifacts.

Insurance Universe

One extraction rail, many insurance lines.

PolDex starts where the live system already works, then adds schema families, evidence rules, and benchmark coverage line by line instead of creating separate products for every insurance category.

Release-ready today

56 release-ready schema contracts across 7 insurance families are callable today across all 56 schema contracts across the insurance universe.

Release gate status

56 of 56 schema contracts cleared the 100+ document release gate, activating the 99%+ accuracy posture across the full universe.

Commercial depth

Loss Runs, SOVs, ACORDs, schedules, endorsements, and broker submission variants continue expanding the commercial surface.

Consumer P&C

Personal Auto, Homeowners, Renters, Condo, Personal Umbrella, Life, Health, Disability, Travel, and Pet are release-ready.

Document primitives

COI, Policy Declaration, Endorsement, Schedule, Binder, Quote, Renewal, Cancellation, and Application are release-ready document primitives.

Claims

FNOL, loss runs, claim packets, adjuster reports, estimates, bills, EOBs, settlements, and subrogation are release-ready claim schemas.

Life / Health / Benefits

Life, health, disability, group health, group life, group disability, stop loss, and benefits evidence all cleared the same benchmark discipline.

Specialty / Reinsurance

Bordereaux, treaties, and facultative certificates are release-ready and covered by the same 99%+ benchmark discipline.

Origin

The problem was visible before the product existed.

Public contractor insurance packets showed the same pattern repeatedly: long insurance PDFs, manual review, structured fields re-entered by hand, and downstream systems waiting for reliable data.

Observed workflow

Insurance documents were acting like the database.

Certificates, policies, endorsements, schedules, and submissions contained the source of truth, but teams still had to read, re-key, compare, and resolve them manually before anything useful could happen in software.

packet.pdf
 certificate_of_insurance.pdf
 commercial_gl_policy.pdf
 additional_insured_endorsement.pdf
 schedule_of_named_insureds.pdf

 required downstream record:
 limits, named insureds, carriers, dates,
 endorsements, exclusions, evidence, conflicts

Product decision

Build the extraction rail, not another destination app.

PolDex is designed to sit behind brokerages, underwriting tools, compliance products, procurement systems, agent products, and internal ops. It turns documents into deterministic records and gives the workflow back to the company using it.

PolDex surface area
 API
 Processor
 Live Proof
 Playground
 MCP server
 CLI
 OpenAPI
 Webhooks
 JSON / CSV / XLSX / ZIP artifacts

Infrastructure, not workflow

PolDex does not ask buyers to adopt another underwriting desk. It supplies clean insurance data to the systems they already use.

Strict contracts, not summaries

The output is schema-constrained JSON with evidence, truth states, conflicts, and abstentions instead of free-form AI narrative.

Agent-ready by default

The same extraction rail is exposed through API, MCP, CLI, OpenAPI, llms.txt, schema discovery, and downloadable artifacts.

Neutral across the market

PolDex can serve brokers, MGAs, carriers, compliance teams, and insurance software vendors because it does not own their workflow.

Meet the team behind PolDex

PolDex is currently built by a solo founder: product, engineering, extraction behavior, proof surfaces, docs, deployment, and agent interfaces are being built from one infrastructure-first point of view.

Contact ->

Olamilekan Akinuli, founder of PolDex — **Olamilekan Akinuli**Founder

Why Now

AI can read documents. Insurance workflows still need truth infrastructure.

The next wave of software in insurance will be partly operated by agents and partly by existing systems. Both need machine-readable, evidence-backed extraction instead of another manual review queue.

Insurance is document-heavy

COIs, policies, endorsements, schedules, claims packets, benefits forms, and broker packets still move as PDFs and attachments.

Generic extraction does not resolve truth

Insurance operations need field evidence, conflicts, abstentions, schema rules, and predictable exports, not just OCR or summaries.

Agents need callable infrastructure

AI agents can browse and reason, but they need machine-readable insurance extraction rails to complete real workflows.

Teams need proof before procurement

Public proof, playground, docs, benchmark, credits, and status make the product reviewable without a sales-led walkthrough.

What is live

The product is meant to be evaluated directly.

API extraction

POST /v1/extract, async jobs, estimates, credit ledger, and downloadable artifacts.

No-code processing

Processor supports batch operational review without changing the backend contract.

Agent interfaces

MCP server, CLI, OpenAPI, llms.txt, and schema discovery for AI-agent workflows.

Public review surfaces

Live proof, benchmark, docs, status, pricing, changelog, and roadmap remain public.

Position

PolDex sits below the workflow layer.

The adjacent categories are real. The difference is where PolDex sits: it sells insurance-native extraction itself, starting with commercial P&C and expanding line by line, so many workflows can use the same rail.

Dimension	PolDex	Adjacent category
Horizontal document AI	Insurance truth infrastructure with schema families, evidence, conflicts, abstentions, exports, signed delivery, and ledgered jobs.	Broad parsing and extraction platforms work across many industries, but they are not insurance-native systems of truth.
Workflow AI platforms	Extraction is the product: API, processor, strict schemas, evidence, exports, webhooks, and credit accounting.	Workflow products often extract as one capability inside intake, audit, broker automation, or compliance workflows.
AI-native brokerages	Neutral infrastructure any broker, MGA, carrier, agent company, or software platform can use.	Brokerages place risk and own distribution. PolDex does not sell insurance or compete with broker workflows.
Manual operations	Self-serve API and processor with repeatable outputs, named downloads, evidence, and audit trail.	Human review teams re-key fields into spreadsheets and internal systems, creating delay and inconsistent QA.