Contained Is Not Resolved: Why AI Demands New Customer Service Metrics
Malte Kosub, is cofounder and CEO of Parloa, the company behind an agentic AI platform reimagining how enterprises connect with customers.
gettyEveryone has been in this situation: You reach out to a customer support line with a problem, but the bot doesn’t understand you and every path leads you somewhere other than the solution. After a few minutes, you give up.
For you, as the customer, the outcome is a total loss. Your problem remains unresolved, you’re frustrated and you have a reason to think twice before reaching out for help again.
But the company sees none of this. Since the automation kept the call, the system counts it as handled. The industry even has a word for this: contained. It doesn’t matter what the customer experience was. On the dashboard, contained looks like success.
I’ve spent nearly a decade building AI for customer experience, and I’ve come to believe that this is one of the most expensive habits in business: the metrics our industry has historically run on don’t measure what we think they measure. And as AI transforms the customer service experience, we risk carrying that mistake into the next era on a much larger scale.
None of this happened by accident.
For years, the name of the game in customer service was call deflection. Human capacity was expensive, so the goal was to keep customers away from it: Route them to the FAQ, the phone menu or the chatbot instead. And the metrics were designed to grade the deflection: average handle time, cost per contact and containment rates.
That design explains almost every service experience people hate. Agents rushing you off the phone because the clock is running. Departments bouncing you around so each team’s numbers stay clean. Having to repeat your story to a fourth person. The system was doing exactly what it was designed to do.
But what started as a cost-saving measure is now costing organizations their customers. When Parloa surveyed 1,001 U.S. consumers about their service experiences, seven in ten said that if automated service didn’t resolve their issue in under two minutes, they would disengage from the system entirely.
Deflection doesn’t just lose customers, though. It also loses the diagnosis.
Every service conversation carries a signal about something upstream. A spike in billing calls might signal an error across invoices. Multiple complaints about the same feature might point to a product defect. When your metrics only tell you how quickly conversations ended, all of this insight stays buried.
But there’s an opportunity here. For the first time, the substance of every conversation is now measurable.
Across the millions of enterprise conversations happening every day, we can now leverage AI to answer questions like: Was the issue fully resolved, end to end? How hard did the customer work to get there? Did their mood improve or deteriorate over the call? What triggered their frustration?
We can look at these interactions individually, or holistically, across thousands of interactions, not just the small sample size of customers who decided to complete a survey. And we can use the insights to improve the experience for our customers, every time.
If we look ahead at the next few years, one thing is guaranteed: AI will take on more conversations, with fewer humans in the loop. And what companies gain in terms of scale, they’re going to lose in terms of visibility.
That means the only way to see what’s happening inside of these calls is to leverage AI to observe them, and to measure the right things. Like understanding what customers ask for, how they feel while the interaction unfolds and whether they received the help they reached out for.
Metrics that capture the full story of every interaction are what create better experiences, and better experiences are what turn one-time customers into lifetime ones.
Your customers are already telling you what to fix. The only question is whether you’re set up to hear them.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
