Case Studies

A dental practice wanted recurring payments on lower-cost ACH without adding a second reconciliation job at the front desk. We built a posting path around the existing ledger instead of a migration project. The cloud service carries opaque payment information and nothing else. Patient matching, posting, and exception detail stay on the clinic workstation. An ambiguous match or a reversal stops and waits for staff. A posting completes only after Dentrix read-back confirms the expected ledger state. The system is deployed and running in production at the clinic.

An orthodontic clinic builds the same patient presentation for every case, from an intake PDF, clinical photos, and a template that keeps changing. We built a local-first pipeline that reads the active template, maps extracted values and photo views into explicit slots, and returns a deterministic deck with an error report and an audit. Missing inputs surface as warnings instead of invented clinical content. The operator confirms patient identity, reads the warnings, and decides whether the deck is ready. Working files are wiped after delivery. This record makes no outcome claim.

A semi-private airline asked how its teams could use AI inside work they already owned. We paired useful tools with guidance, guardrails, and human verification, so the people doing the work kept responsibility for reviewing it and making the final decision. The engagement is confidential. The client stays anonymous and the operational details stay withheld, so this record carries the principle rather than the implementation: AI should support responsibility rather than blur it.