Channels are the different platforms or surfaces where your AI agent can communicate with customers.
| Channel Type | Examples |
|---|---|
| ๐ Voice/Phone | Phone calls |
| ๐ฌ Messaging | SMS, WhatsApp, Telegram |
| ๐ Web | Browser/web widget |
| ๐ซ Ticketing | Zendesk, Intercom, Freshdesk |
| ๐ฑ Social | Slack |
Simple analogy: Think of channels like different doors into the same store. Customers can enter through any door, but they reach the same place.
The core idea: Instead of building a separate agent for each platform, you build one agent and deploy it everywhere.
Your single agent carries:
Why this matters:
Even though it's one agent, it doesn't behave identically on every channel.
You can tune settings per surface, such as:
| Setting | Example |
|---|---|
| Verbosity | Shorter replies on SMS, longer on email |
| Speed | Faster responses for live chat |
| Format | Bullet points for ticketing, casual tone for Telegram |
Key insight: The personality and knowledge stay the same, but the communication style adapts to the channel.
Before going live, you can test your agent per channel using simulations.
How it works:
This ensures your SMS agent and your Intercom agent both perform correctly before real customers interact with them.
Here's how different teams can apply multi-channel agents:
E-commerce Brand
โโโ SMS/Telegram โ Resolve order tracking questions automatically
Support Team
โโโ Intercom/Freshdesk โ Auto-resolve routine support tickets via email
Build Once โ Tune Per Channel โ Test Per Channel โ Deploy Everywhere
| Step | What Happens |
|---|---|
| Build | Define knowledge, tools, guardrails |
| Tune | Adjust verbosity, speed, format per channel |
| Test | Run channel-scoped simulations |
| Deploy | Go live across all channels |
๐ Note: Telegram, Intercom, and Freshdesk are currently in Alpha, meaning they're available but still being refined.