video1.0<iframe src="https://www.loom.com/embed/faef42755b90412190f050a0358ea680" frameborder="0" width="1693" height="1270" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>12701693Loomhttps://www.loom.com12701693https://cdn.loom.com/sessions/thumbnails/faef42755b90412190f050a0358ea680-f71ac24e11974666.gif918.749This Loom explains the governance and practical framework Terry Coll uses to apply AI in a non-profit for homeless young adults. He argues that while generative AI can speed creative work, the real challenge is privacy, predictable pricing, and API stability, since consumer AI offers little privacy. He describes building about 15 to 16 single-pass LLM workflows that classify and summarize incoming messages across more than 100 data streams, producing case files at pennies per entry and typically costing about one to two dollars per month. He also notes that Google’s nonprofit pricing guarantees the production API for at least 12 months, supports commercial data privacy, and can be configured for zero data retention. The system has been running for nine months with stable results, verified by format checks and human review before action.