DTA uses an AI system called Specto, built by NeuroSoph, to automatically sort and categorize paper SNAP Interim Report forms submitted by recipients to report household changes. Before this system, DTA staff sorted these forms by hand; now the AI reads each form, identifies whether it is in English or Spanish, and classifies it as Change, No Change, or Needs Review. NeuroSoph reports the system processed over 86,000 forms in the period tracked from July 2023.

The system uses optical character recognition (OCR), machine learning, and natural language processing to classify SNAP Interim Reports into English and Spanish, then into three action categories: Change, No Change, and Needs Review. It runs on DTA's own servers and connects to the BEACON eligibility system via an API. NeuroSoph reports 100% accuracy on No Change classifications and over 86,000 forms processed in the period tracked from July 2023. Documents flagged as Needs Review are passed to human caseworkers for a final decision.