Featured article

Klarna Chatbot Strategy Shift: Why Companies Are Rebalancing Human and AI Customer Service

The Pendulum Swings Back: Klarna Reemphasizes Human Agents

Almost a year after our team put Klarna’s AI customer service chatbot through its paces, the fintech giant has made a significant strategy adjustment. After initially championing AI chatbots to handle the majority of their customer service volume, Klarna is now reemphasizing the importance of human agent interactions with their customers.

This shift isn’t an abandonment of AI, but rather a recalibration of expectations and implementation. According to recent statements from Klarna, they’re moving toward a more balanced approach where AI handles routine inquiries while human agents tackle complex issues and high-value customer interactions.

klarna chatbot

Part of a Larger Trend: Skyhigh AI Hype Comes Down to Earth

The Klarna chatbot pivot is emblematic of a larger trend we’re seeing across industries. The initial excitement around AI, whether for content creation or customer service, is giving way to a more nuanced understanding of its capabilities and limitations. Companies are coming to terms with two fundamental realities:

Reality #1: Customers Still Want Human Support Options

Customers are getting more and more used to conversing with AI. And AI is getting better and better to talk to. That being said, according to an October 2024 study from Five9, “75% of consumers still prefer talking to a human for customer service”. There’s a lot to this, including trusting AI to handle their data – and years of poor experiences with subpar chatbots who can either answer a limited number of queries or just get stuck in an endless loop from which there is no escape. 

While customers appreciate the convenience and speed of AI chatbots for simple inquiries, they quickly become frustrated when these systems can’t resolve complex issues. Klarna is discovering what many companies are learning the hard way: if you don’t have people as core part of your process, customers will simply go to a service or product that can support them. This is particularly true for financial services, where issues often involve nuanced situations and high emotional stakes.

Learning from Klarna: Exposing the Myths & Facts of AI Agents

Join industry experts as they analyze Klarna’s strategic shift and what this means for broader AI agent adoption.

Reality #2: AI Agents Need Quality Monitoring Too

Another potential concern, that could very well be contributed to the first point, is the second reality companies are confronting: it’s hard to tell if the AI is any good. Without robust quality monitoring systems for AI interactions, major customer experience issues can develop with no clear way to identify them. And metrics like the number of conversations where a customer leaves a conversation and doesn’t come back could be resolution, but could just as easily be frustration leading to abandonment.

While Klarna has provided the number of chatbot handled conversations and estimated resolution rates, more qualitative metrics are not clear. One could guess that, given the substantial savings/profit from the chatbot that they mentioned in their announcement (“$40 million USD in profit”), it seems there’s more to this strategic 180 than just efficiency or cost savings. 

Almost a year to the date of the initial announcement, this about-face makes one question whether their AI chatbot was running into typical issues like:

  • Customers receiving incorrect information
  • Lack of empathetic responses to emotionally-charged situations
  • Circular conversation paths that customers abandon, but never actually reach resolution
  • Inconsistent policy enforcement (and/or hallucinating fictional policies)
  • No clear escalation paths when AI fails

Without including your AI chatbot in your quality monitoring program, these problems can persist and scale before companies even realize they exist.

Finding the Right Balance

Klarna’s strategic adjustment doesn’t signal an industry-wide abandonment of AI. But it does show that adding AI isn’t the shortcut that everyone thought it would be. This new approach emphasizes:

1. Strategic deployment: Using AI for well-defined, routine inquiries where success is easily measurable

2. Seamless handoffs: Creating smooth transitions from AI to human agents when needed

3. Quality monitoring: Implementing specialized tools to evaluate AI conversation quality, not just efficiency metrics

4. Human oversight: Maintaining human review of AI interactions to identify improvement opportunities

5. Customer choice: Offering clear options for customers to choose their preferred support channel

This balanced approach acknowledges both the promise of AI and its current limitations. 

Lessons for Companies Deploying AI Customer Service

For organizations looking to deploy or optimize their AI customer service strategy, Klarna’s experience offers valuable lessons:

1. Start with quality in mind: Implement quality monitoring systems for all your agents, both human and AI, from day one, not as an afterthought.

2. Measure what matters: Look beyond efficiency metrics to measure actual customer experience quality and satisfaction.

3. Provide clear escape hatches: Always offer straightforward paths to human assistance when AI can’t adequately resolve an issue.

4. Continuously train and refine: Use real customer interaction data to identify and address AI shortcomings.

5. Consider emotional intelligence: Recognize that some situations require human empathy that AI simply can’t provide.

Looking Forward: Understanding Interactions Helps You Improve Them

Rather than get caught up in AI vs. humans, it’s best to look at each as a unique tool fit for certain use cases. And looking at your conversations, it can be easy to tell which is best for each, if you’re analyzing conversations correctly. For example, if a conversation has lots of back and forths and variable customer sentiment, it’s likely a better candidates for a human agent. On the other hand, if you see conversations with relatively high or flat sentiment and very few interactions, these should be easy for an AI chatbot to handle.

Understanding these conversational paths by the specific use case or customer intent can give you a clear playbook for where to apply AI confidently. Klarna seems to have flipped the switch on without this visibility and now we’re seeing them have to backtrack. With this perspective, there may not have been as big a media splash, but it also would have avoided a strategic reversal. 

As AI technology continues to evolve, AI will be taking more and more of these conversations. This is not just inevitable. This is what should happen. But understanding where to apply AI should always be the first step. And having the conversational data to make this decision definitively is not only in your best interest, but also a benefit to your customers.