The challenge
In-store kiosks had clunky touch menus that shoppers struggled with, so queues built up at the staffed counters.
The team wanted natural voice interaction but couldn't build kiosk voice infrastructure in-house.
The solution
The team embedded voice into their own app via the MyChatBot Web Voice SDK:
- Voice inside the app, the MyChatBot Web Voice SDK adds inbound (WebRTC) and outbound (Twilio/SIP) calls with a few lines of JS.
- Own voice and instructions, each AI voice agent gets its own voice, language and prompt, fully under the team's control.
- Client-side tools, the agent calls the app's own functions mid-call, navigating the UI, searching, booking, updating records.
- Persistent callers, defaultCallerId keeps a stable identity across sessions, so context follows the user.
- Scales with usage, minute bundles and per-minute overage mean the same SDK serves a prototype and a production app.
Voice was added to the kiosks in a week via the SDK, lifting kiosk usage 24% and cutting queues.
What was deployed
Agents used
Voice Agent (SDK)Client Tools
Channels connected
Web Voice SDKWebRTCTwilio/SIP
"Customers just talk to the kiosk now, and our counters finally breathe."
Oleh Savytskyi
Head of Innovation
Ship something like this
Set up your first AI agent in 10 minutes. No credit card needed.
Start free trial 