AI QUALITY SCORING
BY OWEN ZHANG, COMMSADVISOR. COM
ILLUSTRATION PROVIDED BY ADOBE STOCK
WHAT " 100 % QA COVERAGE " DOESN ' T TELL YOU
Ask any AI QA vendor what their product does and from what I ' ve found you may hear some version of the same sentence:
• Zendesk QA says its AutoQA feature " evaluates 100 % of interactions across digital and voice channels."
• Observe. AI promises to " accurately automate 100 % of QA evaluations."
• CallMiner contrasts its approach directly against " traditional sampling-based scorecards," pitching full coverage as the whole point of switching.
30 CONTACT CENTER PIPELINE
HOW YOU CAN ADJUST YOUR SCORECARDING ACCORDINGLY.
• EvaluAgent ' s pitch is nearly identical: " 100 % of conversations scored automatically."
None of that is false. These tools touch every interaction instead of the 1 %-2 % sample a human team can realistically review by hand. That ' s a significant capability, and it ' s the reason so many QA teams have adopted them in the first place.
But coverage and accuracy are two different claims, and vendors are careful to only make one of them with confidence.
... COVERAGE AND ACCURACY ARE TWO DIFFERENT CLAIMS, AND VEN- DORS ARE CAREFUL TO ONLY MAKE ONE OF THEM WITH CONFIDENCE.