Norma

Inspiration

Concierge services today rely on slow, manual workflows that don’t scale, where a small number of users consume most operational resources. We wanted to automate this entire coordination layer with AI.

What it does

Norma is a voice-first AI concierge that turns natural language into structured plans and calendar-ready bookings. It detects conflicts, proposes alternatives, and confirms actions before generating real calendar events.

How we built it

We built a Node.js + TypeScript backend using MiniMax for reasoning, DeepInfra for embeddings, and ElevenLabs for voice, with a Next.js frontend designed in Tailwind and Vercel v0. Norma uses retrieval, structured planning, and human-in-the-loop confirmation to move from intent to execution.

Challenges we ran into

Getting reliable structured outputs from LLMs and handling natural-language time conflicts were our biggest challenges. We also had to tightly scope features to stay demo-ready.

Accomplishments that we're proud of

We built a fully working voice-enabled concierge that produces real calendar files and resolves scheduling conflicts. Everything runs end-to-end in a polished product experience.

What we learned

Users care more about outcomes than conversation, and structured planning dramatically improves reliability. Voice adds magic, but confirmation builds trust.

What's next for Norma

We plan to add real integrations for restaurants, travel, and tickets, plus persistent user preferences and multi-agent planning. Our goal is to make Norma a universal AI secretary for everyday life.