Klarna made headlines in 2024 when its AI agent took on the work of 700 human agents, then spent much of 2025 rehiring after customer satisfaction fell and CEO Sebastian Siemiatkowski admitted the company had pushed too far. The backtracking shows why AI customer support automation works as triage, not a wholesale replacement, and why the hardest part is not the technology but the implementation.
Where automation earns its place
AI handles repetitive, high-volume questions with clear answers. That includes order status, password resets, shipping timelines, return eligibility, and basic troubleshooting for problems with a defined set of failure modes. Intercom's Fin AI agent, one of the widely used tools in this category, reports a first-contact resolution rate above 50% for most businesses that deploy it.
The fine print on resolution rates
That 50% figure depends heavily on the quality of the company's help documentation. Companies that invest in writing specific, thorough articles see the bot resolve more issues on its own. Those that lean on the model to improvise from thin content get weaker results. A help article that ends with a vague suggestion rather than a precise answer pushes the customer toward a human agent - exactly the outcome the bot was meant to avoid.
Why Klarna reversed course
Siemiatkowski said in 2025 that the company had been too aggressive in replacing people with automation. Customer satisfaction dropped, and the experience showed that while the AI could answer many queries, it could not replace the judgment of a skilled agent for complex or emotional situations. The company began rehiring support staff after the satisfaction hit.
Why this matters for customer support
Customer support leaders who treat AI as a triage layer - letting bots handle the routine while routing what remains to humans - see better retention and higher satisfaction scores than those that aim for full replacement. The lesson from Klarna's experience is that the technology works, but only when paired with a clear content strategy and a willingness to invest in both human and machine resources. Leaders building these workflows can find guidance in programs like AI for Call Center Supervisors, which focus on balancing automation with human oversight.
Your membership also unlocks:
