Short answer
Containment rate is the percentage of customer contacts an automated system handles from start to finish without escalating to a human agent. It is a volume metric: it counts how many conversations stayed inside the bot, not whether the customer's problem was actually solved.That distinction matters. A conversation can be contained and still fail the customer, when the bot loops, stalls, or deflects without resolving. Containment that the customer didn't choose is not a win. So containment is best read alongside resolution and customer effort, never on its own.
Containment vs deflection vs resolution at a glance
| Dimension | Containment | Deflection | Resolution |
|---|---|---|---|
| What happened to the customer | stayed in automation start to finish | was steered away from an agent | got the problem solved |
| Problem solved? | not necessarily | often unknown | yes, by definition |
| How to read it | volume signal, pair with resolution | supply-side relief, check where customers went | the outcome the other two approximate |
Aide, the AI platform for customer service, treats containment as an output, not a target. The north-star number is the resolution rate buyers already understand: how many conversations were genuinely resolved. Intent Coverage Rate, the share of real customer intents with deployed, verified automation, sits underneath as the supporting diagnostic. When coverage is built intent by intent and tested before it goes live, containment rises as a consequence of resolving the right things well, not as a number chased for its own sake.
Read this way, containment stays honest: a conversation is contained only because its intent was verified first, so the contained share and the resolved share point the same way. And automated intents keep their place in the team's view of demand instead of vanishing into the bot.