Amazon Leadership Principles — Behavioral Stories (STAR)

Automating timetable creation using AI

• Amazon LPs — Invent and Simplify · Ownership · Deliver Results · Customer Obsession · Think Big
• Situation — At Toddle, creating a school's timetable is one of the most critical onboarding tasks. Schools shared timetables in many formats — PDFs, Excel sheets, scanned images — and with no automated import, support had to manually recreate them.
• Situation — This meant manually creating bell schedules, periods, teachers, classes, and routines — five to six hours per school. Complex ones were escalated to engineering to write backend scripts or DB queries. As we grew, I saw this was a recurring operational bottleneck, not an isolated issue.
• Task — Although it wasn't on our roadmap, I believed we could greatly improve onboarding by automating the import. I took ownership of exploring whether AI could interpret the formats and produce a structured timetable users could review before importing.
• Action — I first spoke with the support team to learn the most common formats and their frequent manual mistakes, and I collected real customer timetable examples so I could design around actual data, not assumptions.
• Action — I proposed an AI-assisted import tool: users upload any supported format → AI extracts classes, periods, teachers, rooms, schedules → the system generates a structured preview → users review, correct, and import instead of building everything manually.
• Action — After approval, I built it independently within available bandwidth, tested extensively on historical files, and validated the output in demo schools to ensure imported timetables behaved correctly.
• Action — I didn't just hand it over — I ran a one-hour training with support, demoed use cases, answered questions, and documented the workflow. Two weeks later I ran a feedback session on real onboarding.
• Result — A task that took five to six hours could now be done in ten to fifteen minutes. Support mainly reviewed the AI output and fixed small issues (spelling, ambiguous values), accelerating onboarding and freeing them to help customers instead of doing data entry.
• What I Learned — Some of the highest-impact engineering comes from improving internal operational efficiency, not just customer-facing features. Involving users early, training after launch, and collecting structured feedback let me solve their real problem, not just ship something technically interesting.