The principle
The AI answers first — on every channel, within seconds. It hands over to a person when it reaches its limit, or when the customer asks. The customer stays in the same conversation and the same channel; they notice nothing of the switch except the answer.
When a handover happens
- The AI does not know. It says so openly and puts the conversation in the team inbox instead of guessing. This is the most common case and the most important one.
- The customer wants a person. Including in spirit — "is anyone there", "I want to talk to someone". Not just the word agent.
- Money is involved. Cancellations, refunds, address changes after dispatch, exceptions to your terms. The AI never decides these alone.
- The customer is upset. Complaints and legal matters go to a person.
Urgency
Not every handover is equally urgent. Alpheris distinguishes three levels and acts accordingly:
- Calm. The conversation lands in the inbox and waits.
- Waiting. The customer is in the chat and expects an answer now — the row is highlighted.
- Urgent. If somebody is online, they are connected directly. If nobody is there, a high-priority ticket is created, plus a notification and an email with [URGENT] in the subject.
When nobody is available
Under Channels you set availability per channel. Outside those hours the AI keeps answering but does not connect — it creates a ticket and tells the customer when somebody will be back.
Taking over a conversation
In the inbox you assign the conversation to yourself and write. From that moment the AI stops answering there so you do not talk over each other. The rest of the team sees you are on it and stops getting notified for that conversation.
Handing back to the AI
When you end the live chat, the AI takes over again. The customer gets a short note so the switch does not go unexplained mid-conversation.
What you can configure
Each channel has its own rules — WhatsApp may be stricter than the chat window on your website, Instagram more cautious than both. You find the switches under Channels at each channel, not in one central place: the same question can have a different answer per channel.