---
title: When the chatbot should stop and hand off to a person
locale: en
url: https://bitora.es/en/blog/cuando-escalar-chatbot-a-humano/
---

# When the chatbot should stop and hand off to a person

*2026-09-06*

Operational escalation signals: frustration, high stakes, sensitive data and after-hours — without leaving the customer in an empty loop.

A chatbot that tries to solve everything frustrates people. One that escalates too early saves no time. The middle ground is **written rules**, not model vibes.

## Signals that need a human now

1. **Repeated frustration** — the user rephrases the same complaint three times or uses angry language.
2. **High stakes** — cancellations, claims, large quotes.
3. **Sensitive data** — health, minors, bank details outside the allowed flow.
4. **Explicit ask** — “I want to talk to someone” is not a debate.
5. **Out of coverage** — the bot lacks the policy or stock; hand off rather than invent.

## What the hand-off must carry

Whoever picks up needs: a summary, data already captured (name, order, channel) and the reason for escalation. If the human starts from zero, the customer feels the bot only wasted time.

> **Watch out:** “someone will be with you shortly” with no real queue or SLA is worse than an honest response window.

## After hours

Honest options: a structured message, a callback slot, or escalation to a channel that is actually watched (email, ticket). Faking “an agent is typing…” destroys trust.

## How we design it at Bitora

On a [business chatbot](/en/solutions/business-chatbot/) escalation is in scope from day one. Starting prices and timelines (under two weeks with a closed scope) live on [pricing](/en/pricing/) and [chatbots for business](/en/chatbots-for-business/).