Chatbot

What is Chatbot?

The word 'chatbot' is overloaded. In 2026, it refers simultaneously to legacy rule-based systems from 2015, basic LLM wrappers with no system integrations, and fully agentic systems that complete end-to-end workflows. This semantic inflation creates real confusion when enterprises evaluate AI vendors. The key diagnostic question is: what happens when the customer asks something outside the scripted flow? A rule-based chatbot deflects, escalates, or fails. An LLM-based chatbot generates a plausible-sounding answer that may be hallucinated. A properly designed agentic AI system retrieves the correct answer from your knowledge base and, if necessary, takes an action in your system to resolve the issue.

The commercial failure mode of legacy chatbots is well-documented: a 2023 McKinsey analysis found that most enterprise chatbot deployments achieved self-service resolution rates below 30%, with customers frequently abandoning or circumventing the bot to reach a human agent. The causes are consistent: too narrow in scope, too rigid in handling variation, unable to take actions in underlying systems. Agentic AI addresses all three root causes, which is why resolution rates from agentic deployments — typically 65-85% for in-scope queries — represent such a dramatic improvement.

Also known as: Bot, Chat Agent