# US immigration services split scanned applications into documents with a learning model, and send any page the tool cannot identify to a person

Page: https://flyvolo.ai/en/changes/ev-20241104-data-entry-clerk-8
Date: 2024-11-04 · Stage: Deployment
Can this record move a task judgement? yes
Occupation: https://flyvolo.ai/en/careers/data-entry-clerk

## Scope

United States. The agency's own entry in the federal AI use-case inventory, marked deployed and operational since 4 November 2024, says that before the system all pages of an I-539 application were scanned and stored as one document, delaying adjudication; that the tool uses a learning model to identify, classify and separate the individual documents; and that all pages are accounted for, with any page not identified by the tool referred to a human for document type resolution. One form at one agency, run under a contractor; it does not say how many pages go to people or what job title handles them.

## Source

- Office of Management and Budget — 2025 Federal Agency AI Use Case Inventory, individually reported use cases (Department of Homeland Security / USCIS entry DHS-2385, Intelligent Document Processing for I-539 Form Digitization) — https://raw.githubusercontent.com/ombegov/2025-Federal-Agency-AI-Use-Case-Inventory/3c225ba8438e48306ace7698c8c7feb9486cbc69/Data/2025_individually_reported_AI_use_cases.csv (primary source)

## What this means

In a deployed document system, the leftovers the machine cannot read are routed to a person by design.

## What it does not show yet

One form at one agency, with no count of the pages left to people.

## How to verify it yourself

Open the 2025 Federal Agency AI Use Case Inventory and find entry DHS-2385, "referred to a human for document type resolution".


More: https://flyvolo.ai/llms.txt