What is containment rate?
Containment rate is the share of conversations an AI agent or chatbot handles from start to finish without passing them to a person. A contained conversation never reached your team.
Also called: bot containment rate, self-service rate, automation rate.
By Santhul Joseph, AI EngineerUpdated
How to calculate it
Containment rate is the number of conversations that stayed with the bot, divided by all bot conversations, times 100.
Example: a hotel's chat has 500 conversations in a month, and 60 are handed to reception. 440 stayed with the agent, so containment is 440 divided by 500, or 88%.
Why containment alone can mislead
- A customer who gives up and closes the chat counts as contained.
- A bot that makes it hard to reach a person scores high and leaves customers annoyed.
- Some conversations should leave the bot, like complaints, refunds and legal questions. Handing those over is the right result.
- Read containment next to resolution rate and the questions the agent couldn't answer.
Containment in SimplyBoost
SimplyBoost aims for good handovers, not the highest possible containment. The agent hands over when the customer asks for a person, for complaints, refunds, billing disputes, cancellations, legal or data-protection requests and security problems, when someone is upset and the issue isn't solved, or when it couldn't answer twice in a row.
Performance shows how many conversations are resolved, active and escalated, plus the knowledge gap, so you can see which handovers came from missing knowledge.
Questions people ask
Is a higher containment rate always better?
No. Some conversations belong with a person. A good AI agent handles the routine questions and hands over the rest quickly, with the full conversation attached.
How is containment different from resolution rate?
Containment asks whether a person got involved. Resolution asks whether the problem was solved. You want both: questions solved by the agent, and the rest handed over cleanly.
How do I raise containment the right way?
Fill the gaps. Look at conversations that were handed over because the agent couldn't answer, add that information to its knowledge, and the next customer gets the answer straight away.