Escalation to a human agent

Escalation with the full history attached, a live inbox for the team, working hours and topics that skip the bot entirely — the customer repeats nothing.

An automated assistant that does not know when to step aside generates more complaints than it resolves. 4chaty treats escalation as a core function, not an exception: when the bot cannot find an answer grounded in the knowledge base, when the customer explicitly asks for a person, or when the topic falls into a sensitive category, the conversation moves to the team inbox with the full history attached. The agent sees from the first message what has already been discussed and does not ask the customer to repeat it. Outside working hours the bot states transparently when a reply will come and collects the details needed to follow up.

Full history on handover — the agent sees the whole conversation, the customer repeats nothing.
Shared live inbox for the team, with assignment per colleague and conversation status.
Topics that skip the bot — words such as "complaint" or "refund" go straight to a human.
Configurable working hours — outside them, the bot states the reply window and collects contact details.
Escalation on request — the visitor can ask for an agent at any time, without hunting for a hidden button.
Internal notes on the conversation, visible only to the team.
No per-agent fee — add colleagues to the inbox without the subscription rising per seat.
Asynchronous follow-up — conversations started overnight are picked up in the morning, with context preserved.

Three escalation triggers

First: missing information. If the bot cannot find a sufficiently relevant fragment in the knowledge base, it does not improvise — it says it does not have the answer and offers an agent. Second: the explicit request. Any phrasing along the lines of "I want to speak to someone" triggers handover immediately, without the visitor hunting for a button. Third: the sensitive topic. You define the list — complaints, refunds, disputes, medical cases — and messages touching those subjects skip the bot entirely. Those three rules cover practically every situation where an automated answer would have done harm.

What the agent sees on takeover

The conversation arrives in the inbox with everything that came before: the visitor’s questions, the bot’s answers, the cited sources and the details already collected (name, phone, request type). The agent can add internal notes visible only to the team and assign the discussion to a specialist colleague. The single biggest gain is that the visitor does not start the story over — the number one frustration in bot-assisted support. From the customer’s point of view, the move from assistant to human is a continuation, not a restart.

Outside working hours

A chatbot promising "we will get back to you shortly" at 11pm makes a promise it cannot keep. 4chaty knows the team’s working hours and changes its behaviour accordingly: outside those hours it keeps answering from the knowledge base, but on escalation it states transparently when the discussion will be picked up and collects the details needed to follow up. The conversation stays open in the inbox and is taken over in the morning with the full context preserved. That is how you cover evenings and weekends without hiring a night shift.

Legal references

Regulamentul (UE) 2024/1689 (AI Act) — art. 50 privind informarea că interacțiunea inițială are loc cu un sistem AI
Regulamentul (UE) 2016/679 (GDPR) — art. 32 privind accesul controlat al operatorilor la datele din conversații

Frequently asked questions

Does the customer have to repeat what they told the bot?

No. The conversation reaches the agent with the full history attached: the questions, the bot’s answers, the cited sources and the details already collected. Takeover is a continuation, not a restart.

Can I force certain topics straight to a human?

Yes. You define the list of sensitive topics — complaints, refunds, disputes — and messages touching them skip the bot entirely and land directly in the inbox.

What happens outside working hours?

The bot keeps answering from the knowledge base, but on escalation it states the reply window transparently and collects contact details. The conversation is picked up first thing, with context preserved.

How many agents can I add to the inbox?

As many as you need. Usage is metered in AI credits for the conversations processed, not per agent, so adding colleagues does not raise the subscription per seat.

Does the visitor know they are talking to a bot?

Yes, they are informed from the start, as required by art. 50 of the AI Act. When a colleague takes over, the change is announced explicitly in the conversation.

Related resources

Chatbot AI pe site →Conversația de la care pornește escaladarea.Integrări WhatsApp & Messenger (în pregătire) →În pregătire: același inbox pentru toate canalele de mesagerie.Rapoarte & conversații →Cât la sută din conversații ajung la un om și de ce.Ghid: chatbot pentru site →

Automate what you can, hand the rest to a human

A live inbox for the team, full history on handover and working hours that are respected — with no per-agent fee.

See pricing Talk to a consultant