In short: AMD helps your dialer decide what answered a call. The quality of that decision can affect agent time, customer experience, and the way a campaign performs.
What is answering machine detection?
Answering machine detection (AMD) is the step an outbound dialer uses after a call is answered. It estimates whether the answer came from a person, a voicemail system, or another source so the call can be routed appropriately.
Why can traditional AMD get it wrong?
Many systems infer the answer from timing—silence length, pauses, and how quickly speech begins. Real people do not all answer the same way: a hesitant greeting, a noisy environment, or a short response can resemble a recording. A short voicemail greeting can create the opposite confusion.
What should a team measure?
Track both kinds of error: live people classified as machines, and machines classified as live answers. Also monitor time to decision, agent handoff quality, and unclear or fallback outcomes. Compare results on your own call mix instead of relying on a single headline accuracy figure.
How should you evaluate an AMD change?
Start with a representative sample of answered calls. Label the ground truth, agree on how uncertain cases should behave, then compare the candidate system with your existing dialer settings. Run a limited rollout, review edge cases with operators, and expand only when your own measurements support it.
Ask these questions before you switch.
- Is the test set representative of our real call mix?
- Are both false positives and false negatives reported?
- How are unclear calls and service outages handled?
- Can we review labels and edge cases after the test?