What is call containment rate?

Call containment rate is the percentage of inbound calls that are fully handled by an automated system — historically an IVR, now more often an AI voice agent — without ever being transferred to a human agent. It's the primary efficiency metric used to evaluate how well an automated front-door is actually doing its job.

A high containment rate means the AI is genuinely resolving calls, saving human labor, and providing good service. A low containment rate means the AI is mostly serving as a pre-step before human handling — which may or may not be the intended outcome.

How it's calculated

Containment rate = (calls fully handled without human transfer / total calls received) × 100

•Fully handled — the caller's goal was achieved and the call ended without escalation

•Without human transfer — no live agent picked up at any point

•Total calls — all inbound calls the automation received; excludes calls that went straight to voicemail or were abandoned before reaching the automation

Some organizations also exclude calls shorter than a threshold (e.g., 10 seconds) to filter out misdials and hang-ups. Others include these in the denominator.

Containment rate vs. call deflection

These terms are often used interchangeably, but there's a subtle difference:

•Call deflection — a broader term covering any call handled without a human, including calls resolved by knowledge base redirects, self-service web prompts, or SMS deflection

•Containment rate — specifically within a given system; how often calls that enter your AI (or IVR) stay inside it to resolution

In practice: deflection is a business outcome; containment is an AI performance metric.

Typical containment rates

•Traditional IVR: 10–30% (most callers eventually press 0 to reach a human)

•Basic chatbots: 20–40% (works for narrow question types; breaks on edge cases)

•Modern AI voice agents: 60–90% (depends heavily on call mix and agent configuration)

A well-configured AI voice agent handling well-defined call types (appointment booking, FAQ, message taking) can push containment above 90%. Agents handling complex or sensitive workflows (healthcare intake, disputed billing) will land lower — and that's often correct, because the human escalation is what callers need.

What drives containment rate up

•Clear, well-scoped agent instructions that anticipate the top call types

•Strong intent detection — the AI knows what the caller wants

•Tool access — the AI can actually book the appointment or look up the answer

•Knowledge base coverage — the AI can answer the questions callers ask

•Graceful handling of unusual requests — the AI asks clarifying questions instead of instantly escalating

What drives containment rate down

•Overly cautious escalation — AI configured to "when in doubt, transfer" misses resolvable calls

•Missing integrations — AI can't actually check the calendar or look up the order

•Narrow knowledge base — AI answers "I don't know, let me transfer you" too often

•Caller preference — some callers demand a human regardless, and that's reasonable

What containment rate does not measure

•Caller satisfaction — a contained call may have been an annoyed caller who gave up

•Correctness — the AI may have contained the call by giving the wrong answer

•Resolution — the caller may have needed a different outcome than what the AI delivered

Measure containment alongside caller satisfaction, follow-up rate (did they call back?), and first call resolution. Any one of these metrics alone is misleading.

Related concepts

•Call deflection — the broader business outcome

•First call resolution (FCR)

•Average handle time (AHT)

•Intent detection — the AI capability most responsible for containment

See it in action

The Receptionist Agent at 365agents reports containment rate in real time, broken down by call type and time of day. Most of our customers see containment climb from week one as the AI learns the call patterns specific to their business.