-
-
Dashboard
-
Menu Management
-
Order List
-
Knowledge Base
-
Customer Using Voice Agent to Place Menu Orders
Inspiration
Ordering food should be simple and natural, but today’s experiences are often slow, screen-heavy, and frustrating — from long drive-thru queues to confusing self-service kiosks. We were inspired to explore how voice-first AI could remove friction from restaurant ordering, reduce operational costs, and create a more human experience. With recent advances in conversational AI and voice synthesis, we saw an opportunity to turn voice into a reliable ordering interface rather than just a novelty.
What it does
Chattable is a voice-first AI ordering platform that allows restaurants to replace kiosks, drive-thru staff, or QR-based menus with a natural conversational AI.
Restaurant owners can:
- Create and manage multiple restaurants
- Configure menus, pricing, and availability
- Customize the AI agent’s voice, language, opening message, and behavior
- Add restaurant-specific knowledge via text or documents
Customers interact entirely through voice using ElevenLabs Conversational Agents — asking questions about the menu, placing orders naturally, and receiving confirmations. Orders are delivered in real time to the restaurant dashboard and kitchen workflow through a public deployment URL that can be used on kiosks, drive-thru screens, or QR codes.
How we built it
Chattable is built using a modern, scalable AI stack:
- Next.js for the web application and dashboards
- Supabase for authentication, multi-tenant data modeling, and real-time order updates
- pgvector to store embeddings for restaurant knowledge bases
- Google Gemini for reasoning, intent extraction, and embeddings
- LangChain to orchestrate LLM workflows and retrieval-augmented generation (RAG)
- ElevenLabs JavaScript SDK to power real-time, natural voice interactions
The system combines structured menu data with unstructured restaurant knowledge, allowing the AI agent to understand customer intent, answer questions accurately, and reliably convert conversations into actionable orders.
Challenges we ran into
One of the biggest challenges was ensuring reliable order accuracy in a free-form voice conversation. We needed to carefully structure prompts, validate extracted intents, and confirm orders clearly without making the interaction feel robotic.
Another challenge was building a multi-tenant architecture where each restaurant has its own menus, knowledge base, voice personality, and deployment URL — while keeping the system performant and easy to manage.
Accomplishments that we're proud of
- Building a fully working, end-to-end voice ordering system — not just a demo
- Successfully integrating ElevenLabs voice agents with Google Gemini for real-time reasoning
- Supporting multi-restaurant management with customizable AI agents
- Implementing a knowledge base that allows restaurants to “train” their AI using documents and text
- Delivering a clean, production-like deployment flow suitable for real-world use
What we learned
We learned that voice UX requires a fundamentally different design mindset compared to traditional UI. Small changes in phrasing, confirmation logic, and tone can dramatically affect user trust and usability.
We also gained valuable experience orchestrating multiple AI systems — combining embeddings, retrieval, reasoning, and real-time voice interaction into a single cohesive product.
What's next for Chattable
Next, we plan to expand Chattable with:
- POS and payment integrations
- Advanced analytics on ordering behavior and upsell performance
- Franchise-level dashboards for large restaurant groups
- Deeper personalization of AI agent personalities and brand voices
Our long-term vision is to make voice the default interface for ordering food, anywhere.
Log in or sign up for Devpost to join the conversation.