Will the AI say something silly to my customer?
Where the answer comes from
There's one difference that decides everything: the AI can answer from the model's general knowledge or from your business's content. The first sounds good and gets price, deadlines and terms wrong, because it has no way of knowing. The second only states what you registered.
In practice, this means feeding a base with what the team already answers every day, and marking what the AI can say. Items without that mark are visible only to the team.
When it should stop answering
An AI that answers everything is more dangerous than one that answers little. The three situations where it has to pass the ball:
- the question goes beyond what was registered
- the customer asks for a person, explicitly or implicitly
- the matter requires judgment: negotiation, complaint, exception
Either way, what matters is that the conversation reaches the agent with the history. A handoff that makes the customer repeat everything undoes the gain.
The detail that separates a bot from customer service
Customers write in pieces. An AI that replies to each fragment turns into a ping-pong nobody can stand: three customer messages, three mismatched replies.
Waiting for the person to finish typing before answering is a small decision, and it's what makes the conversation feel human the most, more than any tone adjustment.
Frequently asked questions
Can I turn off the AI and use only the inbox?
Yes. Automatic replies are optional per channel and per account: you can use the platform just as a shared team inbox.
Does the customer know they're talking to an AI?
It answers in the tone you configure and signs with the name you choose. When the topic goes beyond what it knows, it hands off to an agent by name instead of insisting.
Can the AI book appointments on its own?
It pre-books: it records the agreed time and notifies the team. Confirmation is still done by a person, nothing goes on the schedule without someone reviewing it.
Updated on