Extracting data from powers of attorney
Lenders and securitization trusts grant powers of attorney so a servicer or agent can sign assignments, endorsements and releases for them, each drafted in its own words. Title examiners and document custodians check one behind every document an attorney-in-fact signs, as recorded county copies or PDFs in the loan file.
Review needs the principal, the agent, the powers granted, the execution date and the notarial acknowledgment. Title review confirms the agent named signed each assignment, and that the grant covered the act on the day it was signed.
What makes powers of attorney hard
Values typed into legal prose
Names and dates fill blanks inside legal sentences, and an execution date can be split across three separate blanks on one line.
Scope set by listed acts
A limited power names the acts it covers, such as signing releases, and each signed document has to fall inside them.
Two signers and a notary
The principal's and agent's signature blocks can sit side by side, above a notarial acknowledgment with its own venue and dates.
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 power of attorney, end to end
About this document: A specimen corporate power of attorney published by a document AI vendor. A mortgage lender appoints a financial services company as its agent for its notes and loan documents. One portrait page: the grant in legal prose, two signature blocks side by side and a notarial acknowledgment. Page 1 is featured. Every name is invented, with signatures drawn to match.
Processing objective: The objective is to establish who granted the power, to whom, and when it was executed and acknowledged before a notary. Both dates are assembled from three separate blanks, and each grounds to the line that holds them.
Demonstration fields: Principal, Agent, Execution date, Notarized on, Notary commission expires
The document
1 page
Page 1shown below
Where each value came from
principal.name
{
"value": "Fake Mortgage Company",
"page": 1,
"range": {
"end": 130,
"start": 55
},
"box": {
"xmin": 0.11456,
"ymin": 0.13744,
"xmax": 0.8113,
"ymax": 0.15222
}
}agent.name
{
"value": "Sample Financial Services, Inc.",
"page": 1,
"range": {
"end": 448,
"start": 351
},
"box": {
"xmin": 0.11327,
"ymin": 0.19581,
"xmax": 0.79029,
"ymax": 0.21107
}
}execution_date
Printed on the page as terms. Executed this 01 day of 01 , 20 25 ., normalized on extraction.
{
"value": "2025-01-01",
"page": 1,
"range": {
"end": 3104,
"start": 3061
},
"box": {
"xmin": 0.11492,
"ymin": 0.60322,
"xmax": 0.56999,
"ymax": 0.61796
}
}notary.acknowledgment_date
Printed on the page as I certify that on 01 , 15 , 2025 personally came before, normalized on extraction.
{
"value": "2025-01-15",
"page": 1,
"range": {
"end": 3353,
"start": 3298
},
"box": {
"xmin": 0.11343,
"ymin": 0.73297,
"xmax": 0.8658,
"ymax": 0.74844
}
}notary.commission_expiration
Printed on the page as 12/14/2029, normalized on extraction.
{
"value": "2029-12-14",
"page": 1,
"range": {
"end": 3613,
"start": 3603
},
"box": {
"xmin": 0.13262,
"ymin": 0.79132,
"xmax": 0.22154,
"ymax": 0.80436
}
}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.
