24-Hour Voice Agents Must Pass 12 Usability Tests

On November 7, 2026, at 09:33, DEV Community's AI section published an article titled 'A 24/7 Voice Agent Needs a 12-Case Availability Test.' The piece argues that claiming a service runs '24/7' only establishes when a phone number may accept calls; it does not prove that a caller can complete a useful and safe task at every hour. Because of that gap, developers evaluating an AI phone agent should test the entire caller path rather than relying on theoretical uptime. The author enumerates 12 areas that need to be exercised before deployment: carrier, SIP routing, media, speech services, model, approved data, text-to-speech, the transfer destination, and the people or fallback process sitting behind that transfer, along with related handoff points in the chain. The test's purpose is to confirm that a call keeps working when any single element degrades or fails, not merely that the endpoint answers. A central caution runs through the article: if one link in the chain has no tested alternative, the service should not be described as 24/7 at all, because an untested gap can still take the whole experience down during off-hours. Only this single report has been published so far, so the matter currently rests at the point where this position and checklist were first put forward.