Extracting data from commercial auto declarations
Commercial auto insurers print declarations on proprietary or ISO-based forms, from a one-page schedule of covered auto symbols to a fleet listing with a block per vehicle. Brokers check every policy at renewal, and a fleet schedules can span many vehicles. Underwriters get the expiring declarations with submissions, as PDFs from portals or email.
Policy checking needs each vehicle's VIN, its coverages with their limits and premiums, and the total policy premium. An agency management system compares those values with the expiring policy and the quote, and a rating system prices the renewal from the vehicle schedule.
What makes commercial auto declarations hard
Forms differ from carrier to carrier
Some carriers define covered autos with numbered symbols on a single page. Others print a separate block for every vehicle, with its own coverage rows.
Coverages nest inside each vehicle
Every scheduled vehicle carries its own table of coverages and premiums, and a trailer lists fewer coverage rows than the trucks beside it.
Limits are printed as text
One row reads $1Million ea occ, the next a split per person and per occurrence, and physical damage shows actual cash value with a deductible.
Parse, extract, verify
Parse
Parse turns the document into reading-order Markdown and a hierarchy of blocks (text, tables, figures), and grounds each block to its exact location on the page.
Extract
You define the fields in a JSON schema, and Extract returns their values in a consistent, structured format, each linked back to where it appears in the Markdown.
Verify
Each value links back to where it appears, down to the character range in the Markdown, and the page and bounding box in the source document, so every result can be verified.
One commercial auto policy's declarations, end to end
About this document: The declarations of an Owners Insurance Company renewal, from the Auto-Owners Insurance Group, for a small excavation contractor. Six portrait pages schedule ten trucks and trailers, each with its own coverage rows and total. Page 6 rolls them up into totals by coverage and the term premium. Page 1 is featured. The insured, agency, policy number and VINs were replaced with fictitious equivalents, and the drivers list was removed.
Processing objective: The objective is to read one row of the fleet schedule the way a policy checker does, tying a vehicle to the liability it carries and its cost. The schedule reconciles: the ten vehicle totals plus one endorsement premium add up to the $11,558.40 policy premium on page 6.
Demonstration fields: Policy number, Scheduled vehicle, Combined liability limit, Combined liability premium, Vehicle total premium
The document
4 of 6 pages
Page 1shown below
Page 4
Page 5
Page 6
Where each value came from
policy_number
{
"value": "95-342-160-75",
"page": 1,
"range": {
"end": 309,
"start": 281
},
"box": {
"xmin": 0.62902,
"ymin": 0.12319,
"xmax": 0.94083,
"ymax": 0.13655
}
}vehicles[0].description
{
"value": "2001 DODG RAM 3500 ST/SLT/SPT",
"page": 1,
"range": {
"end": 845,
"start": 787
},
"box": {
"xmin": 0.05719,
"ymin": 0.26699,
"xmax": 0.60756,
"ymax": 0.30321
}
}vehicles[0].coverages[0].limit
{
"value": "$1Million ea occ",
"page": 1,
"range": {
"end": 1036,
"start": 1020
},
"box": {
"xmin": 0.30245,
"ymin": 0.31979,
"xmax": 0.60727,
"ymax": 0.3336
}
}vehicles[0].coverages[0].premium
{
"value": 718.69,
"page": 1,
"range": {
"end": 1061,
"start": 1054
},
"box": {
"xmin": 0.71843,
"ymin": 0.31977,
"xmax": 0.8346,
"ymax": 0.33356
}
}vehicles[0].total_premium
{
"value": 860.51,
"page": 1,
"range": {
"end": 1680,
"start": 1673
},
"box": {
"xmin": 0.71833,
"ymin": 0.40214,
"xmax": 0.83489,
"ymax": 0.41962
}
}How to try it yourself
Everything on this page is reproducible. The sample, its schema, and the Parse and Extract API outputs are published. You can repeat the same two calls from whichever surface you already build on.
- Claude Skilldescribe the job; the agent writes it
- CLIno code at all
- MCPexpose ADE docs as an agent tool
- REST APIany language
- Pythonlandingai-ade
- TypeScriptlandingai-ade
Documents into data
Accurate
The best accuracy on the difficult documents other tools get wrong, such as messy scans, complicated tables, checkboxes and handwriting.
Auditable
Every answer cites its sources. Values come back with an audit trail (page, bounding box, character range) so a human can review them.
Agentic
Applies multi-step reasoning, and plans before it acts. Handles documents it has never seen before and adapts to variations.
API-first
Composable APIs that drop into any tech stack and can be called from anywhere. Easy to build on and embed in your pipelines and products.
Affordable
Cost-competitive rates with complete price transparency. Spend controls before you run, and every job reports its exact cost.
Enterprise-grade
Built for production and scale from day one. Fully observable and monitored, running securely in our cloud or yours.
Built for high-volume, high-throughput production
- Zero data retention
- Priority rate limits
- Enhanced support
- VPC and on-prem deployment
Available on Team and Enterprise plans. VPC and on-prem deployment require an Enterprise plan.
