What IKEA’s AI Reskilling Bet Can (and Can’t) Teach Smaller Businesses ## Executive Summary When IKEA’s parent company, Ingka Group, deployed an AI chatbot called Billie that handled 47% of customer calls, it chose not to lay off the affected workers. Instead, it reskilled roughly 8,500 employees and redeployed many into interior design and home-furnishing advisory roles. That new service line generated approximately 1.3 billion euros in 2022 revenue. The story has become a popular case study for the argument that companies should reskill workers rather than replace them with AI. But the narrative oversimplifies what happened, and for smaller businesses without IKEA’s resources, the lesson requires significant translation. This article examines what IKEA actually did, where the evidence is strong, where the causal claims are shaky, and what practical options exist for businesses that lack the budget to retrain thousands of employees. ## Why the “Reskill vs. Replace” Debate Matters Now Most businesses deploying AI tools face a version of the same question: when automation takes over routine work, what happens to the people who used to do it? The default answer for many organizations has been to reduce headcount. Klarna leaned into AI-driven customer service, then had to reinvest in human agents when quality dropped on complex interactions. Ford replaced experienced quality engineers with automated systems, then rehired more than 300 of them after defects increased. Commonwealth Bank of Australia replaced 45 customer service agents with AI and saw workloads spike for the employees who remained. These are not hypothetical scenarios. They are recent, documented outcomes at well-resourced companies. The pattern suggests that replacing workers with AI works well for high-volume, routine tasks, but creates problems when the remaining work requires judgment, empathy, or institutional knowledge. This tension is not new. When ATMs arrived in the 1970s and 1980s, banks initially expected to eliminate teller jobs. Instead, the reduced cost per branch led to more branches, more customer-facing roles, and a shift toward advisory services. When accounting software automated tax preparation in the 1990s, firms that reskilled preparers into advisory and compliance consulting roles grew, while those that simply cut staff shrank. The technology changed, but the strategic question remained the same. ## How IKEA Turned an AI Chatbot into a New Business Line IKEA’s approach followed a specific sequence. Ingka Group deployed Billie, which handled 47% of customer calls and saved an estimated 13 million euros in operational costs. Rather than using those savings to fund layoffs, IKEA analyzed customer demand data and identified an unmet need for personalized interior design help. It then retrained affected customer service employees for advisory roles in the new service. The interior design and home-furnishing advisory service generated approximately 1.3 billion euros in revenue in 2022, representing about 3.3% of IKEA’s total revenue that year. Those numbers are impressive, but the causal chain deserves scrutiny. The source material presents Billie’s deployment and the advisory service’s success as a connected story: AI freed up workers, workers were reskilled, reskilled workers generated new revenue. That narrative is plausible, but several links are unverified. The timeline is unclear. There is no public evidence establishing whether the advisory service launched because of Billie, before Billie, or independently. The 1.3 billion euros in revenue could reflect market demand, IKEA’s brand strength, or product innovation rather than (or in addition to) employee redeployment. The cost of reskilling 8,500 employees is also absent from the public record. Training programs at that scale typically run into the tens of millions of euros when accounting for curriculum development, lost productivity during training, and support infrastructure. Whether the 13 million euros in Billie savings offset the reskilling investment, or whether IKEA treated it as a separate strategic bet, is unknown. ## The Evidence for Reskilling Over Replacement The strongest argument for reskilling comes not from IKEA’s revenue numbers but from the failure cases on the other side. Klarna aggressively adopted AI for customer service, then publicly acknowledged that quality suffered on complex and sensitive interactions. The company reinvested in human-supported service after the problems became apparent. Ford let go of experienced quality engineers in favor of automated systems and AI tools. When quality issues mounted, it rehired more than 300 of those engineers. The rehired workers not only identified problems the AI had missed but also trained younger employees and improved the AI systems themselves. IBM replaced hundreds of HR professionals with AI agents that handled roughly 94% of routine requests. But the remaining 6%, cases requiring judgment, empathy, or nuanced decision-making, proved difficult for the AI to manage. IBM subsequently increased hiring in areas where human capabilities remained essential. Commonwealth Bank of Australia replaced 45 customer service agents with AI and reported spiked workloads for the employees who remained. Each case illustrates the same pattern: AI handles volume well but struggles at the edges where human judgment matters most. The counterargument is also present in the data. IBM’s 94% success rate on routine HR work is a strong outcome. For high-volume, low-complexity tasks, replacement can work. The failures tend to emerge in the remaining slice of work that requires context, discretion, or emotional intelligence. ## Where the Reskilling Argument Gets Complicated The reskilling narrative has real limitations that the IKEA story tends to obscure. Scale assumptions. IKEA is a global company with mature HR infrastructure, deep financial reserves, and an existing training apparatus. Reskilling 8,500 people requires institutional capacity that most organizations simply do not have. A 50-person company facing the same question operates under fundamentally different constraints. Survivorship bias. The companies cited, both successes and failures, were selected because their stories are dramatic. We do not know how many companies quietly replaced workers with AI and achieved satisfactory results, or how many attempted reskilling and found it unworkable. Implementation quality matters. Klarna’s and Commonwealth Bank’s problems may reflect poor AI implementation rather than a fundamental flaw in the replacement strategy. Better AI systems, clearer escalation paths, or more gradual rollouts might have produced different results. Employee outcomes are invisible. None of the cited cases include data on employee satisfaction during transition, retention rates after reskilling, or whether reskilled workers performed well in their new roles over time. Reskilling may reduce displacement anxiety but introduce performance stress, especially when employees move from routine tasks to advisory roles requiring different skills. The missing counterfactual. There is no analysis of what would have happened if IKEA had pursued replacement instead. It is possible that both strategies would have succeeded, or that a hybrid approach would have outperformed either. ## What This Means for Small and Mid-Sized Businesses The core principle, ask “what higher-value work can people do?” before defaulting to layoffs, is sound regardless of company size. But the execution looks very different at smaller scale. An SMB with 20 customer service employees cannot afford a multi-year reskilling program with dedicated curriculum designers and transition support. What it can do is think carefully about which tasks AI handles well (routine inquiries, scheduling, data entry) and which tasks still require human judgment (complex complaints, relationship management, sales conversations that require context). The practical question for most smaller businesses is not “should we reskill or replace?” but “how do we redeploy the time AI frees up?” If a customer service representative spends 60% of their time on routine inquiries and AI handles those, the question becomes what the representative does with the recovered hours. That might mean deeper customer relationships, proactive outreach, or cross-training into adjacent roles. It does not require a formal reskilling program. Historical precedent supports this incremental approach. When self-checkout systems arrived in retail, some chains cut cashier headcount while others redeployed cashiers to customer assistance, stock management, and specialty departments. The companies that redeployed tended to report better customer satisfaction scores, though the effect varied widely by implementation. ## How to Apply the Reskilling Principle Without Enterprise Resources Start with a task audit, not a role audit. Instead of asking which jobs AI can eliminate, map the tasks within each role. Identify which tasks are routine and automatable, and which require judgment, relationships, or creative problem-solving. Most roles contain both. Redeploy hours before redeploying people. When AI takes over routine tasks, first look for ways to fill recovered time with higher-value work in the same role. A bookkeeper freed from data entry can spend more time on cash flow analysis. A support agent freed from password resets can handle escalations and retention calls. Measure what matters. Track customer satisfaction, employee retention, and revenue per employee before and after AI deployment. If you reduce headcount and customer satisfaction drops, the savings may be illusory. If you redeploy time and revenue per employee rises, the approach is working. Plan for the 6% problem. IBM’s AI handled 94% of routine HR requests successfully. The lesson is not that AI fails but that someone needs to handle the remaining cases, and those cases are often the most consequential. Ensure you have clear escalation paths before removing human capacity. Be honest about what you can afford. Formal reskilling programs with new curricula and extended training periods may be out of reach. But informal approaches, cross-training, mentorship, gradual role expansion, are available to almost any organization. The goal is intentionality, not scale. Watch for workload redistribution. Commonwealth Bank’s experience highlights a common failure mode: removing some workers and expecting the rest to absorb the overflow. If AI is handling routine work but remaining employees are overwhelmed by complex cases, you have not solved the problem. You have concentrated it. ## Conclusion IKEA’s story is a useful illustration of what becomes possible when a company treats AI as a tool for workforce evolution rather than workforce reduction. The 1.3 billion euro revenue figure makes the case memorable, even if the direct causal link between Billie and that revenue is less certain than the narrative implies. The counterexamples from Klarna, Ford, IBM, and Commonwealth Bank reinforce that replacement-only strategies carry real risks, particularly for work that requires human judgment. For smaller businesses, the takeaway is not to replicate IKEA’s program but to adopt its question: when AI handles the routine work, what becomes possible with the people and time you free up? The answer will vary by industry, company size, and financial position. But asking the question before defaulting to headcount reduction is the single most important step, and it costs nothing.