If you watch Ibrahim’s demo of the Gradient Labs voice agent carefully, you might spot a bunch of incredibly nuanced things that don’t typically feature in demos. These were all intentional, and can help you think about how to navigate the unfortunately common divide between a neat demo and a real, valuable AI system:

➡️ Start with an unclear intent- “I need some help.” The happy demo path usually has a pretty clear ask from the get-go. Real customers rarely so.

➡️ Expect empathy, but not at the expense of due process- “for your security I need to verify your identity.” Real human agents can’t proceed on calls without verifying who they are speaking to, and our AI agent does so too.

➡️ Make it messy! “Ohh wait I got the last three digits wrong.” This is your quickest take away. Interruptions and partial answers are one of the easiest way to cut through into a voice agent’s real quality.

➡️ Keep the customer in the loop- “let me double check.” All it takes is for a phone number to be off by one digit for them to be wrong. We’ve woven what human agents do back into the system’s design to mitigate these risks.