A live working concept: teachers submit preferred dates for Primary 3–6 museum-based learning visits, and an AI scheduling agent consolidates everyone's preferences against real museum capacity once submissions close.
01Teacher Booking Window
02NHB Agent Control Room
Sign in to your school's booking window
For this pilot, enter your school name to open a booking form. (Production would use the school's existing MOE / SingPass-linked login.)
Preferred visit dates —
Each level visits its designated museum (set by the curriculum team). Rank up to 3 preferred date & time slots per level — the agent will try your top choice first, and will try to keep your levels close together where possible. Sessions run back-to-back: 10:00–12:00, 13:00–15:00, 15:00–17:00.
Your school's submissions so far
Museum capacity (classes per session)
How many classes each museum's learning team can host in the same time slot. Adjust to test "what-if" capacity scenarios live.
Consolidated requests
Every preference submitted from the Teacher Booking Window lands here, in submission order.
Run the scheduling agent
Once the submission deadline closes, the agent works through schools in the order they applied — checking remaining capacity, locking in the best available slot per level, clustering each school's levels close together, and flagging anything it can't resolve.
Schedule — agent output
Confirmed visits are ready for calendar invites. Flagged levels need a manual decision below.
Demo note: this prototype runs the agent live against Claude (Sonnet) in your browser, school by school, so you can watch its reasoning happen in the log above. In production this loop would run inside an n8n workflow triggered on the submission deadline, reading from the booking database and writing confirmed slots to Google Calendar / Outlook with .ics invites sent via email automatically.