A charity now verifies every class with photos and runs admin in under two hours a month

Teachers the charity does not employ run classes in buildings no funder visits, two countries away. The old answer was to hire people to check people. The new one is an operating system where a teacher logs each class from their phone with a photo, and the system verifies place, time and head count from the photo's own metadata.

Use case one: proving a class happened

Teachers log the date, time, class group and attendees, then upload a photo taken at the start or end. The system derives three signals and compares them with what the teacher typed: a head count and setting from the image, the EXIF timestamp, and the EXIF GPS converted to a distance from the orphanage. Matches confirm automatically. Mismatches go to a person. The AI never rejects a class or deducts anything.

Across classes analysed so far, 5.1% had photos from a different day, 4.0% fell outside the class window and 4.8% had head counts more than two off. Most turned out to be ordinary: a photo the next morning, a phone on the wrong time zone. Checking a photo costs about a cent, so every class gets checked.

Use case two: proving the kids are learning

A standardised quarterly assessment in four versions: vocabulary, listening, speaking (a short recording the teacher rates) and confidence (teacher and child each score out of five). Teachers run it from the admin panel or send a single-use link so a child can take it on any device without an account.

What automated itself

Once classes were structured and verified, invoicing became a query. Month-end invoices total hours per orphanage at the agreed rate; every billed hour traces to a class log with a photo. Bank accounts sync through APIs, so runway updates as classes and kids are added. Donors see exactly which classes their money funded.