From 18% to 33%, fixed upstream
for [24]7.ai
Phillip Hunter · Founder, RedpointAn AI customer care system was stuck well below its performance target. Usage analysis showed the real problem wasn’t the AI.
18% → 33% resolution
A home warranty company, through [24]7.ai
Diagnostics
60 days
The situation
Two weeks after launch, a home warranty company was ready to pull the plug on its new phone assistant. The call center numbers said the old system did better. Agents were taking calls that came out of the new one, and the project was costing more than what it replaced.
What we found
Nobody had agreed on what “working” meant. Once we did, the goal was resolving the main reason people called, opening a service request. The old system resolved 20% of those calls, the goal was 30%, and the new one was at 18%. Whole calls and more than a hundred recordings showed three problems: authentication up front, missed reasons for calling, and too many steps to finish. The traffic math showed that fixing the end wouldn’t get much closer to 30%, because not enough calls reached it. The fix had to be at the front.
What we did
Redesigned most of the opening. The greeting got specific. The prompt asking why people called got about 60% shorter, rebuilt as Why + Action + Method. Two alternate versions and the redesign went live against the original. All three beat the original, and one beat the others by more than 10 points.
The result
More calls reached the part of the system that could resolve them. Resolution went from 18% to 33%, past the 30% goal, within 60 days.
The full story: Look upstream first ↗