Extracting data from confidential information memoranda
A sell-side banker writes a confidential information memorandum for each company it takes to market, and each bank uses its own layout and chart style. Private equity deal teams and corporate buyers receive one after signing an NDA. A buyout fund can screen hundreds a year, downloaded from a data room or sent by email.
An associate pulls the revenue and EBITDA history, the management forecast, the customer mix, and the bid deadline. Those figures load into the deal CRM and a screening model, which scores the company against the fund's investment criteria.
What makes confidential information memoranda hard
Financials pasted in as images
Bankers often paste revenue charts and income statement tables into the deck as pictures, so the historical figures exist only as pixels.
One fact stated several times
Revenue and the bid deadline appear in the executive summary and again in the financial section or process letter, often rounded or worded differently.
Every bank labels figures differently
Each bank orders sections and names line items its own way, so one memo reports adjusted EBITDA and the next reports run-rate EBITDA.
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 confidential information memorandum, end to end
About this document: A memorandum issued by BDO Canada as court-appointed receiver, to sell a site services and equipment rental group in Newfoundland and Labrador. The receiver markets the business as a sell-side banker would, under a court-supervised sale. Its 19 portrait pages include an executive summary with charts and the historical and forecast financials. Page 3, the covering letter, is featured. Staff contacts were replaced with fictitious names.
Processing objective: The objective is to capture the terms of the sale from the one page that states them all. Each value grounds to its line in the letter. A bidder logs the court's authority and the offer deadline first, since a receivership sale runs to the court's timetable.
Demonstration fields: Company for sale, Seller capacity, Court, Receivership order date, Binding offer deadline
The document
4 of 19 pages
Page 1
Page 3shown below
Page 7
Page 11
Where each value came from
target_companies[0]
{
"value": "PTL Holdings Limited",
"page": 3,
"range": {
"end": 2484,
"start": 2368
},
"box": {
"xmin": 0.14013,
"ymin": 0.1654,
"xmax": 0.8684,
"ymax": 0.19679
}
}seller.capacity
{
"value": "Court-appointed Receiver",
"page": 3,
"range": {
"end": 2595,
"start": 2486
},
"box": {
"xmin": 0.13929,
"ymin": 0.22834,
"xmax": 0.86344,
"ymax": 0.24356
}
}insolvency_proceeding.court
{
"value": "Supreme Court of Newfoundland and Labrador",
"page": 3,
"range": {
"end": 2991,
"start": 2891
},
"box": {
"xmin": 0.13939,
"ymin": 0.30837,
"xmax": 0.8638,
"ymax": 0.32297
}
}insolvency_proceeding.appointment_date
{
"value": "May 11, 2018",
"page": 3,
"range": {
"end": 3095,
"start": 2992
},
"box": {
"xmin": 0.1392,
"ymin": 0.32397,
"xmax": 0.86391,
"ymax": 0.33931
}
}sale_process.binding_offer_deadline
{
"value": "2:00 p.m. (NDT) on Thursday, June 14, 2018",
"page": 3,
"range": {
"end": 4338,
"start": 4235
},
"box": {
"xmin": 0.1397,
"ymin": 0.57707,
"xmax": 0.86434,
"ymax": 0.59234
}
}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.
