ACTIGENGoverned document AI
Turn documents into trusted, structured data.
Bring your documents. Say what information you need. Check a sample, process at scale, and receive data where every value traces back to the page it came from.
A scanned newspaper page arrives. Before any AI touches it, the file is checked.
sn84022871_1874-03-09_p2.jpg
1800 × 2385 px · 1 page · Library of Congress
- Format and integrity checkedIngestGate
- Page condition measured: stained, noisyVisionMax
- Contrast and sharpness improved on a copy, original keptVisionMax
Bring files by upload, ZIP, API or a watched folder. Nothing is overwritten.
Walkthrough on a public-domain page: Orleans County Monitor, Barton, Vermont, 9 March 1874, page 2 (Library of Congress, Chronicling America). Stage names, timings, the layout regions and both review catches come from a real ACTIGEN run on this page. Field values were checked by hand against the page.
Configure once. Reuse at scale.
Your team gets the result. The engine handles the AI.
Nobody needs to choose models, write prompts or wire agents together. A project remembers the fields, checks and review rules you approved, so the next batch of the same documents runs the same way.
Templates, not tuning
Start from a workflow proven on similar documents; nobody picks models or writes prompts.
Evidence on every value
Select any field and see where it was read on the page.
People where it matters
Only uncertain and failed values go to review; the rest flows through.
Nothing silently changes
Corrections and approvals are recorded, and original files are never overwritten.
Start with one document workflow
Built for the documents you already have
Historical records
People, places, dates, events
Registers & certificates
Births, marriages, deaths, relationships
Newspapers
Articles, notices, names, dates
Legislative records
Bills, questions, members, rulings
Invoices
Supplier, totals, tax, line items
Contracts
Parties, dates, obligations, renewals
Catalogues
Products, attributes, prices
Identity / KYC
Name, document number, expiry
Print, handwriting and mixed scripts, including English, French, Russian and Indian languages. Languages depend on the template; ask about yours.
Trust you can inspect
Every exported value links back to its source and the setup that produced it
Checked before processing
File integrity and malware checks run before any document enters the pipeline.
Governed pipeline
Specialised stages run in a recorded order, with quality gates between them.
Human review with a record
Accept, correct or reject, with who did it and when kept for audit.
Keep data where it must be
On-premises keeps documents and processing on your own infrastructure.
Access by role
Business users, reviewers, auditors and administrators each see only what they need.
Standard outputs
CSV and JSON for business teams; IIIF, ALTO and PREMIS for archives.
One engine
Three ways in
Workspace
For business users: projects, review and export.
API
For your applications: send documents, receive accepted data.
Studio
For technical teams: templates, workflows, models and rules.
Deploy the way your data requires
Cloud
Start fast
- Create an account and start the same day
- Subscription plus usage
- Pilot or department
Pilot
Prove it on your own documents
A pilot starts with one repeat document workflow. Together we agree the fields, a reference set to measure against, and what an accepted result is worth to you.
- 1
Share a sample
A few hundred representative pages and the fields you need.
- 2
See measured results
Accuracy against your reference set, review effort and cost per accepted record.
- 3
Decide on volume
Continue in Cloud, or move to Private or On-premises.