The $1.4 Billion Lesson: How to Turn AI Displacement Into a Dividend

Every major technology shift in history has produced the same panicked question: what do we do with the people it displaces? We asked these questions about the loom, the assembly line, the spreadsheet, etc. We are asking it again now, loudly, and with the same binary thinking that flattens a genuinely complicated problem into a false choice. Either AI takes the jobs or it does not. Either workers are safe or they are not. Either you cut headcount or you fall behind. The problem is not that the question is hard. It is that we keep refusing to treat it that way, reaching for simple answers when the situation demands something more rigorous, more systemic, and more honest about the tradeoffs involved.

What gets lost in that binary is the same thing that got lost two hundred years ago. Several years ago, I moderated a conversation with journalist Brian Merchant, author of Blood in the Machine. The book rehabilitates the Luddites from the 1800s from their modern caricature as technophobes who were simply scared and avoiding progress, and restores them to what they actually were: skilled craftspeople who understood automation deeply, had often helped improve the machines being used against them, and were not objecting to technology itself, but to the specific terms on which it was being introduced, without their input, without accommodation, and entirely for the benefit of factory owners. In fact, Merchant calls them "some of the earliest policy futurists" because they proposed ideas such as taxing factory owners to fund retraining for displaced workers. They were not afraid of the future; they wanted a seat at the table where it was being decided. This is a sophisticated position, and it got caricatured into a slur precisely because nuance is inconvenient when you are trying to move fast/break things.

Today, centuries later, we are still having the same argument. The factory owners are wearing different clothes, the machines are invisible, and the displaced workers have LinkedIn profiles, but the core question — who benefits from automation, and on what terms — has not changed at all. And it is still not being answered with the seriousness it deserves.

The story most people get wrong

In 2021, IKEA introduced an AI chatbot named Billie and within two years it was handling nearly half of all customer service inquiries the company received. According to Ingka Group's own reporting, Billie managed 47% of all customer queries directed to call centers: order tracking, product availability, store hours, the endless stream of transactional questions that kept their 8,500 customer service reps occupied but added nothing to the customer relationship beyond basic resolution.

The obvious move would have been to declare victory and reduce headcount accordingly, but IKEA did not do that. What they did instead was more difficult, slower, and more uncertain.

Ingka Group, IKEA's largest franchisee, looked at those 8,500 people and asked a different set of questions. What do these people actually know? Where else could that knowledge create value? These workers had spent years becoming deeply fluent in IKEA's product catalog, learning how to talk to customers, developing instincts for what people actually want when they walk into a store or pick up a phone. The company and customer know-how were already there, what was missing was a business unit that could put those assets to work at a higher level.

So IKEA built one. The company retrained those thousands of customer service workers as remote interior design consultants, supported and trained by AI.  This was not a painless process; reskilling at that scale takes time, investment, and organizational will. But the commercial outcome is documented: according to PYMNTS reporting on Ingka's financials, the remote interior design channel generated $1.4 billion in Ingka's 2022 fiscal year, accounting for 3.3% of total revenue. The company has set a target to grow that share to 10% by 2028.

What had been a pure cost center became a premium revenue line. The AI did not eliminate 8,500 jobs; it created the conditions for a new business. But it is worth being precise about what made that possible: it was not the chatbot. It was the organizational decision to treat employees as assets rather than liabilities, and to invest accordingly.

This is not a feel-good story. It is a hard one that happened to go well.

It is tempting to read the IKEA case as a parable, a company that chose people over profit and was rewarded for its virtue. That framing is both flattering and dangerous, because it implies the outcome was a natural consequence of good intentions rather than the product of deliberate, difficult choices made under genuine uncertainty.

IKEA did not know this would work when they started. They made a bet that their call center workers' existing expertise could be extended into a new service model. That bet required investment in training programs, a willingness to build a business unit from scratch, and patience with a transition that did not produce immediate returns. Most organizations facing the same situation choose the faster, more certain option: reduce headcount, capture the efficiency gain, move on. The reason IKEA's outcome looks exceptional is not that the strategy was exotic, but rather that following through on it is genuinely hard, and most organizations are afraid to go this route.

This is precisely what the Luddites understood about their own situation. The craftspeople being displaced in the 1800s by the power loom were not interchangeable labor. They carried accumulated knowledge, quality standards, and customer relationships that the automated machines could not replicate. The factory owners who ignored that were not just being cruel; they were being shortsighted. The cloth their machines produced was of inferior quality, and it damaged the reputation of the entire industry. Automation without regard for the expertise it displaces tends to produce exactly that: faster output, lower quality, diminishing trust. That is not a historical footnote, but rather a pattern that repeats, and it is repeating now in the form of AI-generated code with vulnerability rates and productivity effects that researchers are only beginning to quantify.

The real insight from the IKEA case is not that reskilling is good. It is that the workers whose roles are being automated carry something that is genuinely hard to replace, and that the organizations willing to do the harder work of identifying and redirecting that value are building advantages that the headcount-reducers may not be.

IKEA is not alone. But the honest picture is more complicated than the headlines suggest.

It would be convenient to present a clean roster of companies doing this well and declare a trend. The reality is messier. Most organizations, even those publicly committed to workforce investment, are still in early stages of figuring out what reskilling actually means in practice, and the gap between stated intention and operational follow-through is wide.

JPMorgan Chase has been among the most deliberate in making this argument publicly. Rather than simply cutting roles where AI has absorbed the workload, the bank has built what it describes as a comprehensive redeployment strategy for AI-displaced workers, actively redesigning career paths and creating mobility across functions. According to a Deloitte survey cited by HR Executive, only about a third of organizations are doing anything similar. Most are stopping at AI fluency training, which raises awareness but does not fundamentally redesign what people do or how their value is measured.

DBS Bank in Singapore offers one of the more operationally serious examples. Facing the need to integrate AI across a large organization without triggering widespread job losses, the bank embedded AI literacy and role-based reskilling directly into its long-term workforce strategy, using AI-powered coaching platforms to provide personalized, on-demand learning at scale. The result was thousands of employees transitioning into AI-enabled roles while day-to-day operations continued. 

PwC committed $3 billion to upskilling 275,000 of its own employees as a calculated competitive bet. Their reasoning, outlined in their 2025 Global AI Jobs Barometer is that you cannot buy your way out of an AI skills transition by hiring externally, because the skills change faster than the market can supply them. AT&T faced an earlier version of this when the telecommunications industry shifted from hardware to software, and committed to reskilling 100,000 employees over several years rather than replacing them. The parallel to the current moment is direct, but so is the caveat that both of those transitions took longer, cost more, and produced more uneven results than the headline numbers suggest.

Whether all of these programs translate into the kind of commercial outcome IKEA achieved is a question the data has not yet fully answered. We are early in this, and the honest position is that the evidence base for large-scale reskilling success is still being built.

What the data actually says, and what it does not

The dominant narrative around AI and employment is one of inevitable displacement and loss. A more careful reading of the evidence produces a more complicated picture, not a reassuring one, but a more accurate one.

The World Economic Forum's Future of Jobs Report 2025, drawing on responses from over 1,000 leading global employers representing 14 million workers across 55 economies, projects 170 million new roles created by 2030 against 92 million displaced, a net gain of 78 million positions globally. That number is frequently cited as evidence that everything will be fine. It is worth being more careful: net job creation at a macro level does not tell us whether the right people will be in the right places to fill those roles, or whether the transition costs will be distributed equitably. The aggregate can be positive while the experience for specific workers, industries, or regions is genuinely damaging. Two diverging things can be true.

What is clearer is the skills premium now attached to AI capability. According to PwC's 2025 Global AI Jobs Barometer, which analyzed close to a billion job postings across six continents, workers with AI skills commanded a 56% wage premium over peers without them in 2024, up from 25% the prior year. Job postings in roles most exposed to AI grew 38% between 2019 and 2024. Industries most exposed to AI saw three times higher revenue growth per employee than those least exposed. The market signal is consistent: it is not trying to eliminate people. It is trying to find people who know how to work with the technology. But knowing how to work with it requires investment, time, and access that are not evenly distributed, and that gap is the part of the story that tends to get glossed over in the optimistic framing.

The Luddites understood this dynamic intuitively, even without the data. Their objection was never to technology advancing, but rather to technology advancing in a way that captured all of the value at the top and left nothing for the people whose expertise made it possible. Two hundred years later, the data suggests that organizations willing to share that value are generating three times the revenue per employee of those that are not. The incentive is there; the execution remains the hard part.

The question most leadership teams are not asking, and why it is harder than it looks

Most AI workforce conversations are happening at the wrong level of complexity. The question being asked in most boardrooms is: where can we reduce headcount through automation? That is a tractable question with a measurable answer, which is part of why it dominates. The question that produces better outcomes — what new value could our people create if AI handled what they are doing now? — is harder to answer, harder to measure, and requires organizational imagination and visionary leadership.

In the case of IKEA, the job titles were a container for a set of skills that could be redirected, but recognizing that required someone to look past the job description to the underlying capability, design a reskilling program around it, build a new business unit to receive it, and sustain the investment long enough to see results. That is not a straightforward set of decisions. It involves real risk, real cost, and real organizational architecture. The reason it is worth talking about is not that it is easy. It is that the alternative, treating displaced workers as a balance sheet problem, often turns out to be the more expensive choice over time. In the end, the chatbot did not build IKEA's interior design business. The people did. The chatbot just made room for it. And making that room, turns out to be where the real competitive advantage lives.

The central argument in Blood in the Machine is that technology is never inevitable, that humans have agency over how we live with the machines, and that the terms on which automation is introduced are always a choice. That was true in 1811 and it is still true now. But exercising that agency well requires resisting the pull toward simple frameworks and sitting with the harder version of the question: not just whether to reskill, but whom, for what, at what cost, over what timeline, and with what support structures in place. These questions reward sustained, serious attention and the appetite to engage in them in a rapidly-changing, uncertain context.
‍

Natalia Martínez-Kalinina is an organizational psychologist and operator working at the intersection of people, culture, and business performance. Her work focuses on how organizations design team structures, leadership systems, and talent strategies to perform in moments of rapid technological and organizational change.