Q: How does Observe.AI calibrate for accuracy?
A: This is dependent on Moments set-up. If you have inaccurate data going in (into the Moment definition), you’re going to have inaccurate data getting out (Auto QA machine results).
Q: Can Auto QA be 100% accurate?
A: No, and neither can manual QA. As you go in and calibrate/audit the machine results, there will always be edge cases where you are not able to get the Moment definition to a stage where either the machine or the human QA will answer the question, in the same way, every single time. Even within your QA team, there might be differences of opinions on what the agent should say or do in a particular scenario - making your business definition subjective.
This is why Observe.AI shows you evidence of why the machine suggested a specific answer and what determining factors were used to arrive at the results. Users always have visibility into why a QA form is scored a certain way and they have control over calibration.
Q: Is it possible for only a portion of the form to have Auto QA questions mapped?A: Yes. Auto QA is mapped by question so in a single evaluation form you can have questions that have to be answered manually, questions where answers are recommended by the machine and questions that are auto-filled by the machine.
Q: Can answers that have been auto-filled by the machine be overwritten manually by a QA Analyst?
A: Yes. A QA analyst can select a different answer than what was auto-filled.