Document AI platform, built for outcomes.
Trusted by leading startups & enterprises
Generic AI wasn’t built to read regulated documents
AI not optimized for your documents
Generic AI tools can read a document, but they don't come with templates for KYC packets or claims files, review queues for exceptions.
Manual review doesn't scale with volume
Every document that falls out of automation becomes a queue, a reviewer, and a delay you don't get back.
Legacy IDP wasn't built for this decade
Rule-based extraction breaks on the exception — and the exception is most of what a regulated document actually contains.
Loved for our accuracy, speed, and ease of use
“We tested foundation models, and newer vendors, and evaluated options on accuracy, compliance and the user experience. Proof Perimeter was the only one that could meet our unique requirements, with the tooling we needed to scale.”
Chief Compliance Officer, #1 UK Home Mortgage Provider
One workspace for your high-risk documents
Get access to an end-to-end platform to run intelligent document processing workflows.
Pre-built and custom extraction templates for KYC, claims, LC, loan, and policy documents
Multi-step pipelines — classification → extraction → review → export
Run thousands of documents in a single job, with no per-document ceiling
Full programmatic access, for integration into your CBS, LOS, or claims system
Human-in-the-loop correction queues, with confidence scoring to route only what needs a person
Structured output to the formats your downstream systems already expect
Field-level lineage on every extracted value — what the model saw, what it decided, and why
Get 20% more accurate results with 50% fewer tokens.
Document AI inference is built for regulated document structures like recurring fields, standard clauses, complex tables, form-based layouts.
Every extraction call runs on models tuned for regulated document structures like KYC fields, claims tables, policy clauses.
Prompt orchestration, context windowing, and confidence-based routing wrap every inference call, keeping extraction accurate across templates, edge cases, and multi-page documents.
Full platform handling caching, batching, scale, and concurrency to handle enterprise volume.
Built for teams handling high-risk documents.
Every stalled application is a customer waiting on a document someone still has to read by hand.
One missed field in a beneficial ownership document is a compliance gap.
Invoice line items that don't match the PO get caught in a spreadsheet, days after the payment should've gone out.
A claims file with an inconsistency buried on page 40 is the one that gets approved anyway, until an audit finds it.
Which path works for you?
Top-grade compliance ships with Enterprise.
AI governance and controls offered with Enterprise plan.
- Zero egress — nothing leaves your chosen boundary
- SSO, RBAC & maker-checker on every workflow
- Independently assessed, not self-reported
Common questions
Is Bring Your Own Key actually free?+
Yes. The full workspace — templates, workflows, batch processing, API, review, export, provenance — has no document limit and no platform fee. You pay your model provider directly, at their published rates.
What happens to my documents on the Bring Your Own Key path?+
They're processed under your model provider's data handling terms — the same terms already governing any other use of that API key inside your organization. Where that boundary isn't acceptable for a given document type, that's what the Enterprise on-premise option is for.
Can we move from Bring Your Own Key to Enterprise later?+
Yes. Templates, workflows, and review configurations carry over — you're swapping the model underneath the workspace, not rebuilding it.
What's actually different about Document AI versus a general-purpose model?+
It's fine-tuned specifically on regulated document types — KYC, claims, LC, loan, and policy documents — rather than adapted from a general-purpose model. In our benchmarking on standard document-extraction tasks, that delivers 20% higher accuracy and 50% lower token consumption than general-purpose frontier models on the same task.
Which Enterprise deployment should we pick?+
Cloud-hosted if speed to production matters most. Your-infrastructure if you need the model inside your own tenancy for governance reasons. On-premise if the data cannot leave the building under any circumstance.
Which model providers work with Bring Your Own Key?+
OpenAI, Anthropic, and Google at launch, with more being added.