Imagine announcing that your AI chatbot was doing the equivalent work of 700 full-time agents and projecting $40 million in savings¹.
Now, imagine finding yourself in headlines worldwide being called an AI failure story.
This is exactly what happened to Klarna, BUT the real narrative has been completely misunderstood.
We sat down with Dr. Philipp Streicher, Pomegranate's Lead Data & AI Scientist, to break down why this story reveals less failure and more good lessons learned.
Phil, the headlines are calling Klarna's move to rehire humans an AI failure. What's your take?
"Look, from what we have read that narrative completely misses what actually happened here.
Back at the start of 2024, reports showed that their bot was handling over 2 million conversations monthly, reducing resolution times by 5 minutes and projecting $40 million in profit improvements². The technical performance was brilliant.
And according to Klarna spokesperson, Klarna's AI chatbot hasn't stopped; it's still running at full capacity today, handling two-thirds of customer interactions and delivering those impressive metrics³. The story here is that they integrated human agents back into their service mix to drive strategic advancement."
So what was the actual problem then?
"Well, customer satisfaction dropped 22% after the AI rollout⁴. That's the metric that wasn't making headlines alongside all the cost savings.
From a behavioural standpoint, this makes perfect sense. The AI could process information efficiently, but it couldn't handle the emotional complexity of human frustration or provide the empathy people needed for sensitive financial situations."
So, how did they solve this?
"Brilliantly, actually. The lesson here, and what fundamentally came out of this, is Klarna's new philosophy: "AI gives us speed, talent gives us empathy."
They created what they call an "Uber-type setup"—remote, flexible human agents who handle complex situations while AI continues to manage routine queries⁵. It's not about choosing efficiency or satisfaction; it's about strategically combining both."
Is this approach working financially?
"The numbers suggest it is. They achieved the projected $40 million profit improvement⁶, with customer service costs per transaction decreasing by 40% since early 2023⁷.
More importantly, though, they're building sustainable customer relationships rather than just optimising for short-term efficiency."
What does this mean for other companies that use AI?
"The key insight is that optimal AI deployment isn't about replacement, it's about strategic integration. The most effective solutions combine AI efficiency with human expertise, where each adds unique value.
It's about building teams that leverage the strengths of both human and digital resource."
So, what's your advice for companies deploying AI in their customer service?
"Monitor holistic metrics, not just efficiency ones. Customer satisfaction and brand perception matter just as much as operational performance.
The most successful companies won't be those replacing humans with AI entirely, but instead those thoughtfully combining AI capabilities with human strengths."
Do you have any final thoughts on where this leaves us?
"This story perfectly illustrates why understanding customer behaviour is crucial in AI deployment. The question isn't whether AI can do the job, it's whether pure efficiency creates the kind of customer experience that builds lasting loyalty and aligns your brand values.
What Klarna did, evolving from an AI-only to an AI-plus-human approach, demonstrates sophisticated thinking that the industry should study, not dismiss."
Ready to build an AI strategy that optimises for long-term success? Speak with our Pomegranate team today.
