Due diligence engine
A review package arrives as hundreds of documents answering a questionnaire nobody has the hours to check line by line. The engine maps every document to every question, returns a coverage status with the quotes behind it, and runs entirely on infrastructure you control.
How it works
Inside the system
11 componentsProjects and access
Each matter runs as a project with a counterparty, a description, and named participants whose permissions the team sets per project.
Reference libraries
Your reviewers maintain a reference questionnaire mapping which document types answer which questions, plus a risk base and checklists describing how to evaluate a document for encumbrances, expired permits, and other problem areas.
Questionnaire parsing
The DOCX questionnaire uploads and parses into machine-readable questions, with the preamble carried into context so the system understands the matter it is reviewing.
Ingestion and OCR
Documents are uploaded by file or by folder. Text PDFs convert straight through; scans route through OCR, and tables and formulas are recognised separately and reassembled into the document text.
Question-to-document analysis
Classification, summarisation, table-of-contents extraction, question routing, a relevance gate, verification, and aggregation run across tens of thousands of question-document pairs, distributed across workers with parallelism that adapts to GPU load.
Coverage statuses
Each question returns one of four states: fully covered, partially covered, requires review where the system declines to conclude on its own, and not covered where nothing was provided.
Cited answers
Every answer arrives with its reasoning, direct quotes, and links into the processed text that open at the exact passage. A finding is checkable in seconds.
Reviewer corrections
A comment on any question is taken into account when it is re-run, and statuses can be set manually where the review catches something the model read wrong.
Follow-up requests
Uncovered and partially covered questions compile into a follow-up document request, and the next round of documents runs through the same analysis.
Risk assessment
Documents are evaluated against your risk base and checklists to produce a risk list for the company or the asset under review.
Report generation
Your report template fills with the findings, the preamble data, the document assessments, and the risk results, then exports as a finished DOCX with citations back to sources.
Who it's for
What you can configure
What we need from you
Example use cases
Mapping 500 documents to a 60-question due diligence checklist
Reviewers spend weeks working out which of 60 checklist questions each document answers before analysis can begin. An AI review engine checks every question against every document and returns answers with quotes and a link to the exact passage.
Often deployed with
Document generation engine
Cut contract drafting from 5 days to 30 minutes.
In-editor document assistant
An AI assistant inside Word, on models you host yourself.