Document vault
Policies, quotes, applications, ACORDs, certificates, loss runs, carrier correspondence and proposals, filed against the client and policy they belong to. Full-text search, folders, versions and permissions.
← All platform
Vault · PrismArchive · extraction
Documents for today's certificate and the twenty years of scans behind it.
Documents in PrismAMS is a document vault with an archive engine behind it. Every agency has two document problems. The live one: today's certificate has to be findable in five seconds. And the buried one: two decades of scanned client files sitting in a folder as thousand-page PDFs nobody can search. PrismAMS handles both, and it does the second without asking you to hand us your client files.
- Page-level
- OCR, handwriting included
- Auto-split
- One scan into its real documents
- Zero-copy
- SharePoint source mode
- REST
- Extraction leaves the app
What's included
Everything in Documents, shipped today.
PrismArchive bulk OCR
AIPoint it at the scanned paper. It reads every page natively, handwriting included, and returns searchable text plus structured fields.
Document boundary detection
AIThe hard part of scanned paper isn't OCR. It's that a scanned client file is one continuous stack of many separate documents. PrismArchive finds the boundaries and splits a nineteen-page scan back into the fourteen or sixteen real documents inside it.
Entity index
Search the archive by person, email, phone or date. Emails, phones and dates are pulled out deterministically, so they're exact and cost nothing to run.
Document intelligence & extraction
AIClassification and structured field extraction across document types, with the schema shaped per document type instead of one generic blob.
Model comparison with real cost metering
Run the same file through different models and diff the results side by side, with the actual token spend recorded per run. You see what a given accuracy level really costs before you commit to it at volume.
Extraction over REST
The structured output leaves the platform over a tenant-scoped, key-authenticated REST API. For an integrator that JSON is the product, so we treat it as a contract.
eSign & PDF tools
Send for signature, prepare fields, split, merge and edit PDFs without leaving the system or buying a second product.
How it works
Four steps, no export in the middle.
- 01
Your scanner writes to your own folder
Staff scan into a folder in your SharePoint, exactly as they do now. Nothing about the front of the process changes.
- 02
PrismArchive reads it where it sits
The file is read transiently over Microsoft Graph, scoped to that one site. We never copy the PDF into our storage.
- 03
Only the index is kept
We keep the OCR text, the extracted fields, the entities and a pointer back to the file where it already lives. The document itself never moves.
- 04
Your staff search the index
Find the document by person, phone, date or content, then open the original in place in your own SharePoint.
Why it's different
Your client files never enter our storage.
- Access is granted per site. We can open the one scan folder you point us at and nothing else in your Microsoft estate.
- The PDF is read in memory and never copied. Switch the connection off and your documents are exactly where they always were.
- What we hold is an index and a pointer, so offboarding isn't a data-extraction project.
- A re-scan of the same file is detected and doesn't duplicate, so the folder can stay a working folder.
- Most vendors want your client files uploaded into their tenant. That's a different risk posture, and it's worth asking them about.
FAQ
Common questions.
+What does my Microsoft admin actually have to grant?
An app registration in your tenant with the Graph application permission Sites.Selected, admin-consented, then read access granted on the single scan site. Under that permission model the app can open that one site and has no path to anything else.
+Does this work if we don't use SharePoint?
Yes. Direct upload into the vault is the original path and it's fully supported. SharePoint mode exists for agencies that don't want client files leaving their own tenant.
+How accurate is it on bad scans and handwriting?
Pages are read natively rather than rasterized, which holds up better on poor scans, and handwriting is read instead of skipped. The model is a setting with per-run cost metering, so you can measure accuracy against cost on your own documents before rolling it out.
+It split one file into fifteen documents. Is that a bug?
Usually not. A scanned client file really is a stack of many separate documents, most of them one page. Over-splitting looks wrong until you check it against the page images, and there's a comparison view for exactly that.
Off-the-shelf is for cereal.
Give us 30 minutes and we'll walk through your real workflows in a tenant set up for your agency. No slides.