What Triggers Escalation

Three things cause a chatbot to escalate a conversation to your team. The first is a question that is not covered in the knowledge base at all. The visitor asks something the chatbot has no information about. The second is a question where the chatbot has partial information but not enough to give a confident, specific answer. The third is a visitor who explicitly asks to speak with a person.

A well-configured chatbot handles the first two gracefully. It does not simply say “I don’t know” and leave the visitor stranded. It acknowledges the limit clearly, offers to connect the visitor with your team, and collects their contact information to make that connection happen.

The phrasing matters. A chatbot that says “I am not able to answer that, but I can make sure the right person reaches out to you. What is the best email address for you?” converts the gap into a lead. A chatbot that just says “I don’t have that information” loses the visitor.

How the Handoff Works

When the chatbot escalates a conversation, it follows a configured handoff flow. The visitor is asked for their name and email (and optionally phone number). The chatbot confirms it will pass their question to a team member and sets an expectation for when they will hear back. This is closely related to how the chatbot builds its knowledge base — gaps in knowledge are what trigger most escalations.

Your team receives an email notification immediately with the visitor’s contact information and the full conversation transcript. They know exactly what was asked, what the chatbot said, and where the conversation ended. There is no ambiguity about what the visitor needs when your team reaches out.

For businesses with CRM integrations, the lead is also created automatically in the CRM tagged as an escalation, with the transcript attached as a note.

The best chatbot gaps I have seen are the ones that surface questions businesses never thought to answer on their website. When the chatbot cannot answer something 20 visitors asked in one month, that is a clear signal to add that content. The gap log becomes a content roadmap.

Scott Thomas, AgentScott

What Your Team Receives

Every escalation notification includes the same information, formatted consistently so your team can respond efficiently.

  • Visitor name and email: The contact information collected during the handoff.
  • Page the conversation started on: Where the visitor was on your website when they opened the chatbot.
  • Full transcript: Every message in the conversation, in order, including what the chatbot said and what the visitor asked.
  • Escalation reason: Whether the chatbot flagged a knowledge gap, low confidence, or a visitor request for a human.
  • Timestamp: When the conversation happened, so after-hours escalations are obvious.

How Gaps Get Filled Over Time

Every month I review the escalation log from your chatbot as part of the retainer. Recurring questions that the chatbot could not answer are the highest-priority knowledge base additions. If five visitors in a month asked about a specific topic not covered in the knowledge base, that topic gets added before the next month starts.

Within 60 to 90 days of launch, most chatbots have covered the majority of real visitor questions because the refinement process is driven by actual behavior rather than assumptions about what visitors might ask. Escalation rates typically drop 40 to 60 percent between launch and month three as gaps are systematically closed. For the full picture of how these systems work together, see how an AI chatbot for customer service works.

Chatbot EscalationAI ChatbotCustomer ServiceLead CaptureKnowledge Base

Frequently Asked Questions

No. The chatbot is specifically configured to stay within its knowledge base and to acknowledge uncertainty honestly rather than fabricate an answer. This is one of the most important configurations in the system prompt. A chatbot that makes up answers destroys customer trust. Honest uncertainty followed by a clear handoff is far better for your business than a confident wrong answer.
The visitor can decline and the chatbot will offer alternative ways to reach your team, such as your email address or phone number. Visitors are never blocked from getting help because they do not want to share their email. The goal is to serve the visitor, not to gate-keep information behind a contact form.
The faster the better, but the chatbot sets an expectation during the handoff. If your business responds within one business day, the chatbot tells visitors to expect a response within one business day. Matching the expectation you set is what matters. Visitors who get a follow-up within the promised timeframe are generally satisfied even if the initial chatbot could not answer their question.
Yes. The monthly performance report includes a summary of all escalations with the questions that triggered them. You can also review the full conversation log at any time through the chatbot dashboard. This visibility is part of why escalations improve over time: you can see exactly what your visitors are asking that the chatbot cannot yet answer.