Multi-channel agent
A multi-channel agent is an AI agent that listens on more than one messaging surface and treats every conversation as one continuous thread. A voice turn at 7 a.m. references the SMS sent at 11 p.m. the night before. The agent has one identity, not one identity per channel.
Why it matters
Most AI assistants are single-channel: a chat window, a voice surface, an SMS bot. The user is multi-channel, so the user has to stitch the channels themselves. They start a thread on WhatsApp, finish it on the desktop, ask a follow-up to Alexa, and discover the agent has no idea what the previous chat said. That's not a stitch problem. That's the agent's problem.
A multi-channel agent solves it by treating every channel as a different entry point into the same conversation. The agent has one memory, one set of skills, one sense of who you are. When you message on WhatsApp, the agent loads the last few turns from any channel into context. When you ask Alexa what's on tomorrow, the agent's reply can reference the email you replied to from your laptop an hour ago.
The hard parts are identity (matching the same human across channels), preference (which channel to reply on by default), and transcript stitching (the agent's reply on Alexa might land as a "push to phone" SMS so the user has the link in their pocket). Done well, the result feels less like an assistant and more like a person who has been paying attention.
How Ajeris handles this
Ajeris ties identity to your phone number. The same E.164 number is your iMessage / SMS, your WhatsApp, and the routing key for your Alexa link, your Telegram bot session, and any inbound voice call. Conversation history is unified across channels: each turn writes a row tagged with its source surface, and the per-user agent loads the last 10 turns from any channel as a preamble. The agent picks its reply channel based on a "preferred channel" setting that auto-adjusts when one channel dominates 30 days of traffic. See the tour and the omnichannel architecture doc for more.
