Inspiration
Small businesses lose valuable customers when calls arrive after hours, while staff often begin the next day sorting through voicemails, texts, and incomplete appointment requests. We built Veille to provide every business with a reliable front desk that remains available when the team cannot.
What it does
Veille is an AI-powered after-hours receptionist for service businesses. It answers phone calls and text messages, understands what customers need, identifies urgent situations, checks availability, books, reschedules and cancels appointments, and sends confirmations.
Business owners receive an organized dashboard containing calls, transcripts, summaries, bookings, and urgent escalations. Veille can also synchronize appointments with Google Calendar and notify external systems through webhooks.
How we built it
We built the dashboard with Next.js, React, TypeScript, and Supabase. The conversational agent uses the Strands Agents SDK and tool calling to safely perform actions such as checking availability, creating bookings, escalating urgent calls, and sending webhooks.
Twilio and Vapi handle voice and SMS communication, while Google Calendar provides calendar synchronization. Stripe powers subscription billing. We designed the interface as a responsive scheduling workspace inspired by professional tools such as Fresha and Square.
Challenges we ran into
The most difficult challenge was turning an AI conversation into reliable business actions. Handling appointments requires accurate timezone handling, availability checks, tenant isolation, database writes, calendar synchronization, and customer confirmation.
We also had to handle real-time voice interactions, webhook verification, failed integrations, duplicate events, and incomplete customer information. Another major challenge was designing one interface that remained useful across dense calendars, call transcripts, settings, and mobile screens.
Accomplishments that we're proud of
We are proud that Veille goes beyond answering questions. It can complete meaningful workflows during a live conversation, including booking appointments and escalating urgent situations.
We also built a native scheduling system, Google Calendar synchronization, real-time call updates, automated summaries, SMS confirmations, secure multi-tenant data access, and a responsive business dashboard. The result feels like a cohesive product rather than a collection of disconnected demos.
What we learned
We learned that trustworthy AI products require more than a capable model. Clear tool boundaries, structured data, validation, graceful failure handling, and security controls are just as important as conversational quality.
We also learned that small interface details matter in operational software. Business owners need information that is immediately scannable, especially when reviewing urgent calls and upcoming appointments at the beginning of the day.
What's next for Veille
Next, we plan to add live rescheduling and cancellations, support more calendar and booking providers, improve multilingual conversations, and introduce configurable workflows for different industries.
We also want to provide richer analytics, team assignments, customer history, automated follow-ups, and deeper integrations with CRM and practice-management platforms. Our goal is to make Veille a complete AI front desk that businesses can trust around the clock.
Built With
- amazon-web-services
- github
- next.js
- python
- supabase
- typescript
- vapi
- vercel
Log in or sign up for Devpost to join the conversation.