ConfOS.

A conference is speakers, stages, sponsors, volunteers, vendors, invoices and refund requests, usually spread across a dozen spreadsheets and one overloaded inbox. ConfOS holds all of it in one structured system, so there is one schedule, one speaker list and one version of the truth.

The AI assistant sits on top of that data and can actually act on it. Ask it to reassign speakers and it creates the sessions and does the assignment. Ask which sponsor payments are outstanding and it checks the records and offers to draft the reminder emails. Every action runs through the same rules as the interface, so the assistant cannot put the data into a state a person couldn't.

The scheduling engine does the quiet work. Move a talk and it checks the speaker is not already booked on another stage, flags the conflict, and keeps draft and published versions separate so attendees, speakers and the tech team each see the schedule that is true. Speaker self-service and an attendee question queue keep the inbox out of it.

What we built

• A structured data model covering speakers, sessions, stages, sponsors, volunteers and finances
• An AI assistant that reads and writes that data through the same rules as the interface
• A scheduling engine with conflict detection and draft and published versions
• Speaker self-service so profiles and availability stay current
• A queue for attendee questions and refund requests
• Waitlist onboarding and subscription billing