Klarna Went Too Far: What a 22% CSAT Crash Teaches HR Leaders About AI Workforce Reduction
When Klarna's CEO Sebastian Siemiatkowski publicly admitted "we went too far" with AI-driven workforce cuts, the fintech giant had already watched customer satisfaction plummet 22% — a metric that no amount of cost savings could paper over. Klarna's story is now the canonical cautionary tale for every HR leader weighing AI workforce reduction.
The Timeline: From Bold Bet to Public Reversal
Klarna began aggressively replacing customer service staff with AI between 2022 and 2024, eliminating roughly 700 customer service roles and shrinking its total workforce by approximately 40% — from around 5,000 employees to about 3,000.
In February 2024, the company announced its AI chatbot was handling 75% of all customer service chats, managing approximately 2.3 million conversations per month across 35-plus languages. The narrative was triumphant: AI was doing the work of hundreds of agents, faster and cheaper.
By December 2024, a hiring freeze had pushed headcount down another 22%, to roughly 3,500. But behind the efficiency headlines, cracks were widening. Customer satisfaction scores had dropped 22%, and repeat contact rates climbed on complex issues — customers were calling back because the AI hadn't actually resolved their problems.
By May 2025, Siemiatkowski made his public reversal, launching Uber-style rehiring of human agents at 400 SEK per hour. The course-correction was confirmed in September 2025, with a hybrid human-AI model becoming fully operational by early 2026.
As Siemiatkowski put it: "We probably over-indexed a little bit on that, and then in the last six months we have been trying to course-correct."
Where AI Worked — and Where It Didn't
Klarna's experience offers a clean separation between AI-appropriate and AI-inappropriate service domains.
AI succeeded on routine queries: order status checks, return processing, FAQ responses — high-volume, low-complexity interactions where speed matters more than nuance.
AI failed on complex interactions: billing disputes, fraud investigations, account closures, and any scenario requiring empathy or contextual judgment. These are precisely the interactions that drive customer loyalty — and where Klarna's bot left customers frustrated and unresolved.
Tech analyst Gergely Orosz described the chatbot as "underwhelming," characterizing it as primarily a routing filter rather than a genuine problem-solver. That assessment captures Klarna's core miscalculation: they treated AI as a full replacement when it was functioning as a triage tool.
Klarna Is Not an Outlier
Klarna's experience reflects a broader pattern. An IBM survey of 2,000 CEOs found that only 25% of AI projects deliver their promised ROI. Industry analysts have called Klarna "almost the poster child for bad AI deployment" — not because using AI was wrong, but because the all-or-nothing approach ignored the technology's current limitations.
The lesson is structural, not technological. AI performs well on bounded, repetitive tasks. It underperforms on tasks requiring judgment, context, and relationship management. Organizations that conflate these categories — as Klarna did — end up measuring efficiency gains in one column while customer trust erodes in another.
Three Takeaways for HR Leaders
1. Scope AI by task domain, not by headcount targets. Klarna's mistake was workforce-level: cut 700 roles, let AI absorb the work. A domain-scoped approach would have identified which interaction types AI could handle and preserved human capacity for everything else. Start with task analysis, not FTE reduction goals.
2. Phase rollouts with CX circuit breakers. A 22% CSAT drop doesn't happen overnight. Klarna lacked the feedback loops to catch degradation early and reverse course before damage accumulated. Build measurable quality gates — satisfaction scores, resolution rates, repeat contact metrics — into every phase of AI deployment. Define thresholds that trigger automatic human re-engagement.
3. Treat AI as augmentation infrastructure, not a replacement strategy. Siemiatkowski's own conclusion points here: "Really, investing in the quality of human support is the way of the future for us." The highest-performing model is not AI or humans — it's AI handling volume while humans handle value. HR leaders who frame AI deployment as workforce augmentation rather than workforce reduction will avoid the trap Klarna fell into.
Looking Forward
Klarna's hybrid model — now operational in early 2026 — may ultimately prove that the company learned the right lessons. Siemiatkowski has committed to a clear principle: "From a brand perspective, a company perspective, I just think it is so critical that you are clear to your customer that there will always be a human if you want."
For HR leaders evaluating AI-driven workforce changes, Klarna's journey compresses years of hard lessons into a single case study. The technology works — within its domain. The workforce strategy works — when it respects those boundaries. The failures happen in the gap between what AI can do and what organizations assume it can do.
The question is no longer whether to deploy AI in workforce operations. It's whether your deployment plan has the guardrails to catch a 22% CSAT drop before it happens — not after.
Sources: Entrepreneur, Tech.co, Fast Company, Fortune, HCA Mag, Digital Applied, Fintech Weekly, Customer Experience Dive.