The most impressive AI customer service case and the lesson about what happens when cost trumps quality
24 de fevereiro de 2026
8 min read
Source: klarna.com
US$40M
Projected Savings
2/3
Chats via AI
700
Agent Equivalents
In February 2024, Klarna made the year's most impactful AI customer service announcement: its OpenAI-based assistant had handled two-thirds of all support chats in its first month — equivalent to the work of 700 full-time agents, across 23 markets and 35+ languages.
The projected savings were US$40 million in additional profit for 2024. Average resolution time dropped from 11 to under 2 minutes. Repeat queries fell 25%. Customer satisfaction matched human agents. It seemed like the perfect case — until it wasn't.
US$40M
Projected Savings
2/3
Chats via AI
700
Agent Equivalents
Klarna, the Swedish buy-now-pay-later fintech present in 23 markets, processed millions of support interactions monthly. Refunds, returns, billing queries, and dispute resolution were constant demands in 35+ languages.
The company needed to scale service without proportionally scaling costs. The OpenAI partnership seemed like the definitive solution — and initial numbers spectacularly confirmed this thesis.
Klarna implemented an OpenAI-based AI assistant to process all customer service chats globally. The system handled refunds, returns, billing queries, and dispute resolution, operating 24/7 in 35+ languages.
Initial results were extraordinary: 2.3 million conversations in the first month, 2/3 of all chats handled by AI, resolution time under 2 minutes (vs 11 before), -25% in repeat queries.
However, in 2025, Klarna partially reversed its strategy and reinvested in human agents. The company publicly acknowledged that "cost appears to have been too dominant an evaluation factor" and that service quality suffered from aggressive automation.
Klarna's case offers the most valuable lesson about AI in customer service: efficiency metrics cannot be the sole success criterion.
The numbers were impressive: US$40M in projected savings, 700 agent equivalents, resolution in <2 minutes. But customer-perceived quality declined to the point where the company reversed its strategy.
This doesn't invalidate AI use — quite the opposite. It demonstrates that the correct approach is hybrid: AI for volume, efficiency, and availability, with humans for empathy, complexity, and retention. Klarna tried to go 100% AI and discovered that perfect is the enemy of good.
The lesson for the market: implement AI to scale, but keep humans in the loop for quality. The balance between automation and empathy is the real competitive differentiator.
Catalisa was designed with a hybrid AI philosophy — exactly what Klarna learned it needed. Intelligent handoff between AI and humans is a native platform feature, with sentiment and complexity detection for automatic escalation.
Catalisa's Workflows allow precisely defining when AI should resolve alone and when it should involve a human, avoiding the trap of automating everything at the expense of quality.
KLARNA. “Klarna AI Assistant handles two-thirds of customer service chats in its first month”. klarna.com. Disponível em: https://www.klarna.com/international/press/klarna-ai-assistant-handles-two-thirds-of-customer-service-chats-in-its-first-month/. Acesso em: 27 fev. 2026.
OPENAI. “Klarna Case Study”. openai.com. Disponível em: https://openai.com/index/klarna/. Acesso em: 27 fev. 2026.
Schedule a demo and see how Catalisa can transform your business
Quantitative analysis of the real return on AI investment in customer experience, with consolidated data from Forrester, Boticário, Lotte, and Synthesia.
27 de dez. de 2025
·
10 min read
Analysis of how companies like Carrefour, Boticário, and Synthesia found the ideal balance between automated efficiency and human empathy in customer experience.
29 de nov. de 2025
·
9 min read
How Synthesia, Lotte, and Carrefour absorbed exponential growth in service demands without proportional hires, using AI as a force multiplier.
04 de fev. de 2026
·
9 min read
© 2026 Catalisa. Todos os direitos reservados.
Conforme LGPD
Meta WhatsApp Business
Dados no Brasil
We use cookies for analytics and load external resources (images) to improve your experience. By clicking "Accept", you consent to the use of cookies and third-party resources in accordance with our Privacy Policy.