Product: Data Validator prototype

[kuusinc]

Overview

As a Product Manager, I want to prototype an AI-assisted data validator so I can test whether AI can reduce heavy manual validation work by flagging listing mismatches with online resources, suggesting source-backed updates, and helping admins or volunteers prioritize listings with critical mismatches for further human review.

Action Items

• Review the current FOLA CSV export schema and identify which public food-seeker fields should be included in the prototype
• Define the minimum prototype workflow for validating a small sample of listings
• Build a free/low-cost lightweight prototype that can:
  • accept a sample CSV export
  • check listed website and social media sources
  • compare public source information against FOLA CSV listing data
  • flag mismatches
  • suggest replacement values when source evidence is available
  • include source links for reviewer follow-up
  • include source date where available for reviewer reference
• Test the prototype on a small sample of listings
• Document the prototype findings, limitations, and recommendation for whether this should move forward as a larger implementation task

Resources/Instructions

Related issue:

• Document the process of adding, validating an organization #2752

Prototype notes:

• Validate public food-seeker fields only
• Treat AI output as a recommendation requiring human review
• Prototype only reviews csv export data
• Start with a small sample, such as 25 listings