Extracting data from retirement statements
A workplace retirement plan sends a statement each quarter, and every recordkeeper lays its own out differently. A mortgage underwriter reads one to count an applicant's reserves, and a divorce or benefits case turns on the same figures. Large lenders see thousands a quarter, arriving as portal downloads and scans.
An underwriter needs the ending balance, the contributions on each side and the vested portion. A loan origination system counts the vested balance toward reserves, and discounts what the employer could still take back if the participant left.
What makes retirement statements hard
Five plans, one statement
A participant can hold several workplace plans at once, and each repeats the same sections with a different set of lines filled in.
Two sides of every contribution
Money the participant put in and money the employer put in are reported separately, per plan and again in total.
Vesting can be withheld
One plan prints a vested percent and the next says it is not displayed, which a reader has to tell apart from zero.
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 retirement statement, end to end
About this document: TIAA's own sample quarterly statement, marked SAMPLE on every page with the participant's name and plan numbers left blank. Fifteen portrait pages: a cover with the balance and a projected income chart, then five workplace plans each with holdings, changes in value and vesting, and disclosures. Page 2 is featured, carrying the portfolio summary and the allocation.
Processing objective: The objective is to lift the quarter's portfolio activity, the five figures that account for the change in balance. They reconcile exactly: the opening balance plus both sides' contributions plus gains gives the closing balance to the cent.
Demonstration fields: Beginning balance, Your contributions, Employer contributions, Gains and losses, Ending balance
The document
4 of 15 pages
Page 1
Page 2shown below
Page 3
Page 9
Where each value came from
portfolio_summary.beginning_balance
{
"value": 460806.88,
"page": 2,
"range": {
"end": 2627,
"start": 2616
},
"box": {
"xmin": 0.25328,
"ymin": 0.15647,
"xmax": 0.44792,
"ymax": 0.18315
}
}portfolio_summary.employee_contributions
{
"value": 8250.02,
"page": 2,
"range": {
"end": 2701,
"start": 2693
},
"box": {
"xmin": 0.25329,
"ymin": 0.18314,
"xmax": 0.44796,
"ymax": 0.20096
}
}portfolio_summary.employer_contributions
{
"value": 7425.03,
"page": 2,
"range": {
"end": 2776,
"start": 2768
},
"box": {
"xmin": 0.2533,
"ymin": 0.20092,
"xmax": 0.44801,
"ymax": 0.22749
}
}portfolio_summary.gains_losses
{
"value": 25492.73,
"page": 2,
"range": {
"end": 2840,
"start": 2831
},
"box": {
"xmin": 0.25331,
"ymin": 0.22748,
"xmax": 0.44805,
"ymax": 0.25241
}
}portfolio_summary.ending_balance
{
"value": 501974.66,
"page": 2,
"range": {
"end": 2911,
"start": 2900
},
"box": {
"xmin": 0.25333,
"ymin": 0.25241,
"xmax": 0.44809,
"ymax": 0.27112
}
}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.
