Document AI

About Document AI

Document AI helps you collect and analyze documents that are not government-issued IDs—such as SSN cards, W2s, IRS notices, utility bills—from individuals or companies.

With our new AI-powered features, you can:

• Securely collect and automatically classify documents
• Automatically extract fields from collected documents
• Review collected documents quickly and easily with AI-generated insights

You can also run a set of basic fraud checks on documents.

Use cases

You may use documents as proof of address, proof of SSN, proof of EIN, or other similar validations in your user diligence or business diligence process. Document AI can help your organization automate a large portion of what is often manual and time-consuming work for this diligence.

Get started

If you're interested in Document AI, please reach out to your customer success manager or contact the Persona support team.

Securely collect and auto classify documents

You can use Document AI to securely request and collect a variety of documents from your end users. Most types of documents can be automatically classified.

Collect documents

You can collect documents as part of your existing Persona identity verification flow, or as a standalone document request.

Popular Persona flows include:

• Document only (request one or more documents)
• Government ID Verification + Selfie Verification + document
• Database Verification with a fallback to Government ID Verification + Selfie Verification + document

You can incorporate Document AI via API or by integrating via Persona's hosted or embedded UI flow. If you integrate our UI, you can customize the copy on the screens where you request a document to outline which documents you are requesting and which documents are acceptable.

Secure Document Collection can be used without any other Document AI features, simply as a way to securely collect and store documents.

You can link collected documents back to an Account within Persona via an Account Identifier. You can then view all documents collected for an end user, along with the rest of that user's Persona record—including Verifications, Reports, and more—within the Account.

Automatically classify document type

For most standard documents, Persona can automatically classify the type of document. Classification is performed by using a combination of machine learning, machine vision, and rule-based techniques.

Supported types include:

• Government forms (e.g. W-2, W-9, 1040)
• Vehicle-related documents (e.g. vehicle insurance card, title, registration)
• Statements from financial institutions (e.g. bank statements, credit card statements)

For less common documents, Document AI may still be able to support your use case. Reach out to your customer success manager or contact our support team.

Automatically extract fields

Persona can extract specific information from a document by leveraging optical character recognition (OCR). Note that extraction accuracy is dependent on the quality of the images provided.

Extract exact values, or by type of field

Persona offers AI extraction as the recommended option to accurately extract information. To configure this extraction method, we provide a list of extraction keys and instructions (a search prompt) to large language models.

This is particularly helpful when:

• The document layouts lack consistency
• The document lacks relevant search keys used with other extraction methods

Below are examples of fields we can extract:

• Address
• Business name
• Date
• EIN
• Individual name
• Phone number
• Price
• SSN

Take action based on document extraction

We recommend that you use Workflows and Cases to establish business logic for approving, declining, or manually reviewing documents based on the outcome of document extraction.

Review collected documents

You can view each collected document within the Document Viewer in the Persona Dashboard.

The Document Viewer displays the document alongside a list of each extracted value, which helps you quickly review the document. You can click on an extracted value to scroll to the place in the document where it was extracted.

To expedite the review process, we now offer an AI Insights feature that allows for more nuanced data interpretation and unlocks value in complex document analysis tasks. For example, AI Insights can provide information on the type of corporation (Sole proprietor, C corp, Partnership, etc), and the number of UBOs. When processing bank statements, it can return the number of debit transactions vs credit transactions.

Run fraud checks

Persona can run certain fraud checks on documents to look for instances where a document has been tampered with editing software, is publicly found online, or is missing required fields.

The following types of fraud checks can be run on documents:

• Detect compromised documents: Checks the submitted document against sets of known compromised documents and flags if there is a match.
• Flag document modification: Detects if a document has been tampered with by an editing tool after its creation.
• Verify extracted information: Verifies properties found on the document (e.g. EIN) against government and other databases.
• Check for recency: Ensures that a document (e.g. utility bill) has been issued within a period of X days (e.g. 30, 60, 90 days).

FAQ

How does Document AI work?

Document AI works just like Persona's Government ID flow. Your user will be prompted to upload their document. In a few seconds, Persona analyzes the document and, depending on your customized criteria, either accepts the document, fails it, or sends it to manual review.

What types of documents can Persona accept?

Persona accepts a variety of documents. To learn more, reach out to your customer success manager or contact our support team.

Can I require that certain properties exist in a document in order for it to be accepted?

Yes. For example, let's say your organization wants to only accept documents that contain an address. Persona can recognize whether a document has an address on it and if it does, allow the user to upload it successfully, or if not, prompt the user to try again.

What file types can Persona accept?

Persona accepts PDF, JPG/JPEG, PNG, HEIC/HEIF, and TIF/TIFF files. You can further configure which file types you want to accept in an Inquiry flow.

Is there a size limit for each file that is uploaded?

Each file can be up to 10MB.

Is there a page limit for PDFs that are uploaded?

Each PDF can have up to 20 pages.