AI review
AI That Checks the Work, Page by Page
DiligensHub reviews each document as it's uploaded. It finds totals that don't add up, missing signatures, risky clauses and exposed personal data, and shows you the page.
How the Review Works
- The same checks, every file
- Documents are reviewed one at a time, in a fixed order, through the same 11 checks.
- Findings show their source
- A finding appears only when the AI is confident and can point to where it is in the document.
- Scans count
- Scanned pages are read as images, so ink signatures and stamps are seen.
- Common formats
- PDFs, Word, Excel and text files are reviewed.
- Your documents stay yours
- Documents are never used to train AI models.
- You decide
- Switch AI review off for a deal, or turn individual groups of checks on and off.
Eleven Checks on Every Document
Every upload goes through eleven checks. Seven report on the file itself. Four pull out facts that Signals compares across the deal.
- Key details
The Facts of Each Document, Filled In
Every file gets a clean title, its type, a summary, key dates, and the people and companies it names. The stored file name never changes.
- Red flags
Deadlines and Disputes, Caught Early
Obligations coming due are listed with their dates, and expired ones are marked. Litigation, defaults and regulatory action are flagged with the page they're on.
- Sensitive information
Exposed Personal Data, Found and Masked
Social security numbers, bank and card numbers, passwords and API keys are flagged with where they appear. Numbers are validated first, and every match is stored masked.
- Key details
The Facts of Each Document, Filled In
Every file gets a clean title, its type, a summary, key dates, and the people and companies it names. The stored file name never changes.
- Red flags
Deadlines and Disputes, Caught Early
Obligations coming due are listed with their dates, and expired ones are marked. Litigation, defaults and regulatory action are flagged with the page they're on.
- Sensitive information
Exposed Personal Data, Found and Masked
Social security numbers, bank and card numbers, passwords and API keys are flagged with where they appear. Numbers are validated first, and every match is stored masked.
More Checks on Each File
Four more checks look at each document on its own.
- Categorization
- Files each document against your checklist: automatically when it's sure, as a suggestion when it isn't.
- Math check
- Totals and percentages that don't add up, in text and spreadsheets.
- Signatures and dates
- Missing signatures or counter-signatures, missing dates, drafts and blank placeholders.
- Legal clauses (Beta)
- Missing or unusual confidentiality, change-of-control and termination terms. Uncapped indemnity, unlimited liability, personal guarantees, exclusivity and most-favoured-nation terms.
Checks Across the Deal
Signals uses these four to compare each document with the rest of the room.
- Document links
- Amendments, exhibits and restatements, tied to the agreements they change.
- Financial figures
- Key figures such as revenue and EBITDA, pulled out so they can be compared across documents.
- Narrative claims
- What a CIM or presentation asserts, ready to check against the records.
- Customers and suppliers
- How concentrated revenue and supply are.
Signals
Every Finding, Across the Whole Deal
Signals brings each document's findings together with checks that look across documents. It catches what no single file shows. Forty kinds of signal, in five groups:
- Inside a document
- Missing signatures and dates, drafts, missing exhibits, math errors, risky or missing clauses, litigation and defaults, deadlines coming due, exposed personal data.
- Between documents
- Figures that differ from one document to the next, tax returns that don't match the accounts, two "final" versions of one agreement, gaps in a chain of amendments, company names that don't match.
- The story against the numbers
- A CIM that says "no litigation" when the records show a claim; growth claims the figures don't support. Also margin swings, falling revenue, cash that doesn't follow earnings, and forecasts that break from history.
- Gaps and concentration
- Expected documents that aren't there, such as debt agreements or insurance; customers or suppliers above a set share; key contracts missing or at risk.
- Room hygiene
- Checklist items closed without a supporting file, requests left unanswered, requests marked done with nothing attached.
Mark each signal fixed, or dismiss it with a reason, and export the open list to CSV.
Files Matched to Your Checklist
Each upload is matched against your diligence checklist. New checklist items look back through the room for files that already answer them. You confirm every match.
AI Assistant
Ask the Data Room a Question
Ask in plain English, for example: "Which contracts have a change-of-control clause?" The assistant searches the room, reads the documents, and answers with links to the exact pages. It only uses what the person asking is allowed to see.
- Every answer cites the document and page
- Limited to the files the person asking can see
- Switched on per deal