Acceptance testing a small salon app with fake data across 61 scenarios

Before a real salon used the app, a separate AI session acted as the owner on a test computer with fake data, running 61 scenarios over four rounds. The first round uncovered 19 defects, some serious, which were all fixed, and later rounds came back clean. The scenarios targeted difficult cases such as forced password changes, lockouts after wrong passwords, messy Italian-language spreadsheets, duplicate imports, wrong time zones, data isolation between salons, large customer lists, unreliable networks, and mobile and keyboard use. The report also noted what could not be verified, such as the real booking-software export and WhatsApp messages, and left these for a supervised first session with the owner. The author now runs such an acceptance round before handing over any app, since it catches a different kind of mistake than automated tests do.