Extracting data from Form 5500 filings
Hundreds of thousands of benefit plans file a Form 5500 each year, each with schedules that vary by plan size and type. Retirement plan advisers and acquirers review them, downloading PDFs from the Labor Department's EFAST2 search or a data room.
Reviewers need the participant counts on lines 5 and 6, the plan characteristic codes, the attached schedules and the Schedule H asset totals. A benefits model tracks headcount and assets year over year, and diligence checks whether the participant count required an independent audit.
What makes Form 5500 filings hard
Different schedules for each plan
Which schedules attach depends on plan size and type, and each one restarts its own line numbering beneath its own header.
Answers are checkboxes and codes
Plan characteristics are entered as codes like 2G and 2L, and attached schedules are marked in checkbox grids on the main form.
Provider blocks repeat on Schedule C
Schedule C repeats a block for every service provider, each carrying its own EIN and compensation beside a list of service codes.
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 Form 5500, end to end
About this document: The Form 5500 for the St. Olaf College 403(b) Retirement Plan, posted by the college on its finance site. Twenty-four portrait pages: the three-page main form, then Schedules A, C, D, H and R. Page 2 carries the participant counts and the attached-schedule checkboxes. The name of the person who signed the filing was replaced with a fictitious one before publication.
Processing objective: The objective is to read the plan's headcount as it moved through the year, and to confirm which schedules the filing carries. Each count sits on its own numbered line, and the checked Schedule H box shows the plan files as a large plan.
Demonstration fields: Participants at start of year (line 5), Active participants at year end (6a(2)), Total participants (6f), Participants with account balances (6g(2)), Schedule H attached (10b)
The document
4 of 24 pages
Page 1
Page 2shown below
Page 10
Page 18
Where each value came from
participants.total_beginning_of_year
{
"value": 2080,
"page": 2,
"range": {
"end": 5029,
"start": 5025
},
"box": {
"xmin": 0.82473,
"ymin": 0.2486,
"xmax": 0.97893,
"ymax": 0.26683
}
}participants.active_end_of_year
{
"value": 798,
"page": 2,
"range": {
"end": 5480,
"start": 5477
},
"box": {
"xmin": 0.82458,
"ymin": 0.3166,
"xmax": 0.97894,
"ymax": 0.33672
}
}participants.total_end_of_year
{
"value": 2096,
"page": 2,
"range": {
"end": 6028,
"start": 6024
},
"box": {
"xmin": 0.82438,
"ymin": 0.40949,
"xmax": 0.97895,
"ymax": 0.42777
}
}participants.with_account_balances_end_of_year
{
"value": 2094,
"page": 2,
"range": {
"end": 6414,
"start": 6410
},
"box": {
"xmin": 0.82426,
"ymin": 0.45412,
"xmax": 0.97896,
"ymax": 0.48242
}
}schedules_attached[1]
{
"value": "H",
"page": 2,
"range": {
"end": 8384,
"start": 8354
},
"box": {
"xmin": 0.55024,
"ymin": 0.72216,
"xmax": 0.78096,
"ymax": 0.73959
}
}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.
