Book a discovery call

16 CourtsApp · United States Shipped

One support agent that talks, types and hands players a checkout link

It searches court availability, in the widget on the site and inside ChatGPT.

Before
Support by hand on two disconnected channels
After
One agent answering voice and chat, searching availability and returning a checkout link, embeddable anywhere and usable inside ChatGPT
Voice + chat one widget, both channels
The work, before and after
Before and afterBefore: Player question, then Two separate channels, then Answered by hand. After: Voice or chat, then One agent, then Courts searched, then Checkout link. Result: Voice + chat, one widget, both channels.BEFORE2 OF 3 PAIN POINTSAFTER3 OF 4 AUTOMATEDPlayer questionTwo separate channelsAnswered by handVoice or chatOne agentCourts searchedCheckout linkVoice + chatone widget, both channelsBefore and afterBefore: Player question, then Two separate channels, then Answered by hand. After: Voice or chat, then One agent, then Courts searched, then Checkout link. Result: Voice + chat, one widget, both channels.BEFORE2 OF 3 PAIN POINTSAFTER3 OF 4 AUTOMATEDPlayer questionTwo separate channelsAnswered by handVoice or chatOne agentCourts searchedCheckout linkVoice + chatone widget, both channels
3
actions: search, facility, checkout
ChatGPT
works inside it, not only on the site
How the system runs
Input
  • Customer question
  • Voice or chat
ArrAI
Embedded support agent
Output
  • Answered, 24/7
  • One shared knowledge base

01 What changed

  • A player can ask, search, and get a checkout link without a human in the loop
  • Voice and chat give identical answers, because they share one brain
  • Keys stay server-side; the widget is safe to embed anywhere
The full story The problem, how we approached it, and what we built

A court-booking platform needed customer support that scaled past a human queue. We built an embeddable agent that handles live voice calls and chat from a single widget.

Built for
Marketplace founder and support lead
Engagement
CourtsApp
Sector
Sports-tech marketplace
Region
United States
Year
2025

02 The problem

A growing booking marketplace was handling customer support by hand. It did not scale.

A separate phone line and chat tool meant two systems, two experiences, and gaps between them where customers fell through.

03 The approach

We built one embeddable widget that drops into any page and handles both channels: real-time voice over WebRTC and text chat.

A backend-first design means voice and chat always use the same assistant, the same knowledge, the same answers. No split personality between channels.

04 What we built

  • Embeddable widget with voice and chat from one brain
  • Three actions behind the agent: search availability, show a facility, create a checkout
  • The same actions exposed to ChatGPT, so a player can book from there too
  • Keys kept server-side so the widget is safe to embed

Want this kind of result mapped for your business?

Book a discovery call