22MAR

'We went too far': Klarna rehires staff

4 min read

12:34UTC

The company that became the technology industry's proof of concept for replacing workers with AI is hiring humans again — and its CEO is publicly admitting the experiment failed.

← Back to 45,000 tech layoffs, half may be reversed Jump to analysis ↓

EconomicAssessed

Key takeaway

Klarna's reversal follows a documented pattern: wholesale automation of complex human interaction reliably fails.

Klarna CEO Sebastian Siemiatkowski replaced 700 customer service agents with AI. Customers reported "robotic responses" and "Kafkaesque loops" when attempting to resolve issues⁠¹. Satisfaction dropped. Siemiatkowski conceded publicly: "We went too far"⁠². The company is now rehiring human agents.

Siemiatkowski had staked his public credibility on AI as a direct labour substitute. In early 2024, he announced Klarna's AI assistant was handling the equivalent of 700 agents' workload within weeks of deployment, and presented the savings as proof the model worked. Klarna became the technology industry's go-to example of successful AI replacement — cited in earnings calls, investor decks, and boardroom presentations across sectors. That the reversal comes from this company — the most aggressive adopter, not a cautious incumbent — strips away the defence that failures elsewhere result from poor implementation. Customer service is text-based, repetitive, and high-volume: if AI cannot hold this ground reliably, the case for substitution in more complex service environments is weaker than the market has priced.

The failure fits a pattern now backed by accumulating data. An Orgvue survey of 300 HR managers found 55% of business leaders admitted wrong decisions on AI-driven layoffs; a third had already rehired 25–50% of the roles they cut, and one in three spent more on restaffing than they saved⁠³. Forrester independently placed the regret rate at 55%, predicting half the cuts would be quietly reversed — often offshore or at lower pay⁠⁴. Harvard Business Review research by Thomas H. Davenport and Laks Srinivasan found only approximately 2% of organisations reported layoffs tied to actual AI implementation⁠⁵. The other 98% cut based on projected capability.

The equity market has rewarded these cuts without pricing the reversal risk. Block's 40% headcount reduction sent shares up 22–25% in after-hours trading . Meta's planned 20% cut lifted shares approximately 3%. If Klarna's trajectory generalises — and the Orgvue, Forrester, and Gartner data suggest it will — those share-price gains rest on savings that partially evaporate once rehiring, retraining, and institutional-knowledge recovery costs arrive. The market has priced the cut but not the boomerang.

Deep Analysis

In plain English

Klarna, a major buy-now-pay-later company, replaced its entire customer service TEAM with AI chatbots to cut costs. Customers hated it — they got stuck in loops, received robotic answers, and couldn't resolve real problems. The CEO publicly admitted the experiment failed, and the company is now rehiring human agents. This is not simply an embarrassing story about one company. Research firms across three separate methodologies predict that half of all businesses making similar cuts will quietly reverse them by 2027. The savings promised by AI vendors and projected in board presentations did not materialise as expected.

Deep Analysis

Synthesis

Klarna's case reveals an 'automation residual' problem absent from vendor capability assessments. As routine queries are absorbed, the remaining human-or-AI workload skews toward complexity and customer distress. This creates a compounding dynamic: each percentage point of automation increases the average difficulty of the remaining queue, accelerating satisfaction decline beyond what headcount ratios would predict.

Root Causes

Customer service automation failures follow a structural adverse-selection pattern that standard productivity models miss. AI handles high-volume, low-complexity queries effectively, leaving the residual queue disproportionately composed of the most difficult, emotionally charged, and non-standard cases. The AI then faces precisely the interactions it is least equipped to resolve. Satisfaction metrics therefore fall faster than raw automation percentages imply — a non-linear relationship that cost-benefit models built on task-substitution rates systematically underestimate.

Escalation

The reversal dynamic is entering a second phase. As Klarna's admission becomes a publicised case study, other firms face reputational risk from persisting with visibly failing AI deployments. This reputational pressure could accelerate reversals beyond Gartner's 50% forecast and compress the typical 2–3 year failure recognition cycle.

What could happen next?

Precedent
Klarna's public admission establishes a reputational permission structure for other CEOs to reverse AI staffing decisions without career penalty — reducing the psychological cost of reversal across the industry.
Short term · Assessed
Risk
Companies that have not yet reversed failing AI customer service deployments face compounding customer lifetime value erosion the longer they delay acknowledgement.
Medium term · Suggested
Consequence
AI customer service vendors face heightened contractual scrutiny of satisfaction guarantees as Klarna becomes a standard due-diligence reference point in procurement decisions.
Short term · Assessed
Opportunity
Hybrid human-AI customer service models — AI triage with mandatory human escalation paths — are positioned to gain market share as pure-AI deployments visibly and publicly fail.
Medium term · Assessed

Source Landscape

This story draws on centre-right-leaning sources from United States

United States

Primary parallel: The late-1990s mass deployment of Interactive Voice Response (IVR) systems eliminated tens of thousands of telephone operators at firms including AT&T and American Airlines, producing an immediate satisfaction collapse and the cultural meme of pressing '0 for operator.' Most companies quietly restaffed human escalation paths within three to five years.

Counter-parallel: Amazon's fulfilment centre automation from 2012 onward deployed over 500,000 robots while simultaneously expanding its human workforce — demonstrating that augmentation strategies, rather than wholesale replacement, can sustain service quality where interaction complexity varies widely.

Why it’s relevant now

The IVR parallel is directly actionable: it suggests Klarna's reversal is structurally predictable rather than aberrant, and the speed of reversal — under two years — tracks the IVR experience closely enough to be treated as a base rate.

Rehiring 700 trained customer service agents involves recruitment costs typically estimated at 50–150% of annual salary per role, plus a 3–6 month ramp-up period during which neither AI nor new hires operates at full efficiency. At median UK customer service salaries of approximately £25,000–£28,000, total restaffing costs may reach £17–£29 million before productivity normalises — a figure absent from Siemiatkowski's public accounting.

The longer-term economic risk lies in customer lifetime value erosion during the AI deployment period. Satisfaction declines in financial services correlate with elevated churn: a 10-point NPS decline typically corresponds to 1–3% annual revenue attrition in subscription and credit products. Klarna serves over 85 million active customers; even modest churn acceleration during the AI period could represent revenue losses exceeding the wage savings generated.

Consensus view: MIT economist David Autor's task-based labour framework predicts that non-routine interpersonal tasks resist substitution while routine cognitive tasks automate successfully — exactly the division that customer service AI failure exposes at scale.

Counter-view: Tsinghua University's Institute for AI International Governance (I-AIIG) has argued that AI customer service failures in Western markets partly reflect training data biases and cultural specificity. Chinese deployments of comparable systems within WeChat-integrated services report higher satisfaction, suggesting failure may be context-specific rather than universal to current AI capability.

Key tension: Whether Klarna's failure reflects an inherent ceiling of current large language models in relational contexts, or a correctable deployment error, remains analytically unresolved — and the distinction determines whether the rehiring reversal is temporary or structural.

Sources:Fast Company·Washington Times·Entrepreneur

Mentions:Thomas H. Davenport →Orgvue →Meta →Harvard Business Review →Forrester →Gartner →Klarna →Sebastian Siemiatkowski →

First Reported In

Update #2 · 45,000 tech layoffs, half may be reversed

Fast Company· 22 Mar 2026

Read original →

Different Perspectives

Salesforce, Synopsys and TD Bank Group

Salesforce, Synopsys and TD Bank Group each filed quarterly disclosures in late August booking restructuring charges, or none at all, without naming AI as a cause. Their silence matters because Challenger's tracker shows AI as a stated reason fell to fourth place in August even as the year-to-date AI-cut total still leads at 116,175.

Singapore, South Korea, Taiwan and Indonesia

Singapore launched its Skills and Workforce Development Agency on 16 September, giving citizens six months of free premium AI tools, while South Korea ring-fenced its AI tax windfall in a new Future Response Fund. Taiwan kept funding its AI build past NT$190bn and Indonesia rewired vocational training around AI literacy, betting state-built skills beat a market-led adjustment.

ver.di, CGT Fonction Publique and CCOO

Germany's ver.di banked a 3.3% pay rise on 1 September and opened talks on a Tarifvertrag Transformation covering dismissal bans and reskilling, while France's CGT rejected Paris's AI negotiating timetable the same week. Spain's CCOO went further on 21 September, proposing to tax companies by the jobs they generate rather than wait for the next bargaining round.

BIS General Manager and Federal Reserve governors

The BIS's General Manager said on 10 September that AI displacement remains limited, even as the BIS's own survey found nearly 80% of firms plan to automate roles. Two Federal Reserve governors made the same point in July, arguing the labour-market data does not yet show a mass-firing event.

Bank of Canada, ONS and ECB

The Bank of Canada found the job-finding gap between AI-exposed and unexposed occupations widened from 2.2 to 13.9 percentage points since 2015-19, while separations barely moved. That framing, a hiring freeze rather than a firing wave, is echoed by the ECB's finding that euro-area AI use hit 52% of workers in 2026, concentrated among the university-educated.

Office for National Statistics

Deferred its Transformed Labour Force Survey beyond November 2027 and disclosed a May 2026 telephone-collection failure. The ONS carries no AI-attribution layer at all, so Britain sits outside this month's cohort of measuring states by its own admission.