Smart QA Model Upgrade 💫 Join the BETA

Your Smart QA scores are about to get a lot more honest. Want first access?

Smart QA already scores every teammate on every conversation, not just a sample. Our upgraded model makes that signal sharper: it reads nuance better, is more disciplined about what it scores, and is stricter about handing out 5s.

Too many 5s? Scores that shift when you reorder criteria? Newsletters graded like real replies? We’ve fixed that, and we want you to test it first.

What you’ll get

• Scores that mean something. A 5/5 is harder to earn, so real gaps show up.
• Reliable scoring. Reorder, add or remove criteria while maintaining scoring reliability.
• Less noise. Newsletters and ads show as N/A, not failed.
• Fairer tone and empathy, including in French and other languages.

Changes to the data: expect your averages to dip by about 0.6 points with the stricter scale. For future improvements, we will be exploring ongoing tuning of the scoring to your standards based on overrides and feedback. 

Want in? Comment YES below and we’ll switch it on for your workspace.