min. readMost conversations about AI and jobs start from the same fear. Automation is a trade: machines for people, and someone always loses. One well-documented case from a global home furnishings retailer complicates that story.
A few years ago, the company introduced an AI assistant to handle customer service. In its first two years, the assistant could resolve 47 percent of customer questions on its own. Today that figure is 74 percent, according to public reporting on the program. A support function got faster and cheaper almost overnight. That part of the story is familiar. What the company did next is the part worth building an AI reskilling strategy around, not just admiring.
The choice most companies skip
When a chatbot starts absorbing routine questions, the obvious move is to shrink the team that used to answer them. This company took a different path. It looked at the roughly 8,500 call center employees whose workload had shrunk. Then it asked a harder question: what were customers still calling about, and could a person do that better than a bot?
The answer was design. Customers who got past the basic questions were usually trying to plan a kitchen, a closet, or a whole room. Those problems need judgment, taste, and a back-and-forth a bot cannot yet offer. So the company retrained those 8,500 workers as remote design consultants instead of letting them go. The retraining took about two years for the original group. It now takes five to six weeks for new hires.
That is a real cost. Training at that scale is not free, and it does not pay off in a single quarter. Leaders under pressure to show fast AI returns rarely choose it. This company did. That choice was a judgment about the business, not a guaranteed formula for every company.
What the numbers earned
The results are specific enough to check. Reskilled workers now staff a remote-sales channel. That channel has been the company’s fastest-growing sales line for three straight years. It grew 15 to 20 percent annually and brought in the equivalent of over a billion euros last fiscal year, up from the year before. The company also reports its customer happiness score has climbed to 89 percent, up from 60 percent before the AI assistant existed.
None of that erases the layoffs elsewhere in the broader business. Public reporting notes the company’s corporate structure cut hundreds of office-based roles the same year, unrelated to this program. The company did not attribute those cuts to AI. They also left the remote-sales team untouched. The honest version of this story is not “this company never cuts jobs.” It is narrower and more useful. The specific team affected by this AI rollout kept their jobs. They moved into higher-value work and helped build a channel worth over a billion euros a year.
Why this matters beyond one company
We see a version of the same fear in almost every conversation about adding AI to an engineering or support function. Leadership wants the efficiency, but the team hears “fewer of us.” That fear is not irrational. It is also not the only outcome available.
This is our position, not just an observation about one retailer. When a tool makes a piece of work faster or cheaper, demand for that work has historically grown rather than shrunk. Economists have called this the Jevons paradox since the 1800s. We think the pattern holds for engineering now. Narrow tasks disappear. Judgment work tends to grow instead: deciding what an AI system should own, reviewing what it produces, staying accountable for what ships. We laid out that argument in more depth in AI and jobs: what the Jevons paradox reveals about the future of work.
What the company got right
This case works because someone did the harder analysis before making the harder promise. They looked at what the AI could not do well. They found the higher-value work hiding behind that gap, and built a training path into it. That is a design decision, not a lucky break. It is the kind of decision that benefits from technical partners who have built AI-augmented systems before, not just deployed a chatbot and hoped.
Good AI implementation and good nearshore engineering practice overlap more than people expect. Both depend on scoping the work honestly. What should the machine own? What should stay human? What does the team need to learn to do the second part well? We wrote about that overlap in nearshore in the AI era: the part that stays human. We covered the judgment AI still cannot replace in the rise of agentic AI and why human-on-the-loop is the new standard.
A question worth sitting with before the next AI rollout
This story is not a template anyone can copy exactly. Few companies have the balance sheet or the adjacent revenue opportunity that made a two-year retraining program worth the wait. But the underlying question travels well. When a piece of work gets automated, does anyone in the room ask what the freed-up people could do instead? Or does the conversation stop at headcount?
So maybe the useful question is not whether AI will change a given role. It probably will. It is whether anyone has looked hard enough at what’s left over to find where the people go next.
Most conversations about AI and jobs start from the same fear: that automation is a trade, machines for people, and someone always loses. One well-documented case from a global home furnishings retailer complicates that story.
A few years ago, the company introduced an AI assistant to handle customer service. In its first two years, the assistant could resolve 47 percent of customer questions on its own. Today that figure is 74 percent, according to public reporting on the program. A support function got faster and cheaper almost overnight. That part of the story is familiar. What the company did next is the part worth building an AI reskilling strategy around, not just admiring.
The choice most companies skip
When a chatbot starts absorbing routine questions, the obvious move is to shrink the team that used to answer them. This company looked at the roughly 8,500 call center employees whose workload had shrunk and asked a different question: what were customers still calling about, and could a person do that better than a bot?
The answer was design. Customers who got past the basic questions were usually trying to plan a kitchen, a closet, a whole room, problems that need judgment, taste, and a back-and-forth a bot cannot yet offer. So the company retrained those 8,500 workers as remote design consultants instead of letting them go. The retraining took about two years for the original group and now takes five to six weeks for new hires.
That is a real cost. Training at that scale is not free, and it does not pay off in a single quarter. Leaders under pressure to show fast AI returns rarely choose it. This company did, and the fact that it did is a judgment about the business, not a guaranteed formula for every company.
What the numbers earned
The results are specific enough to check. The remote-sales channel built around those reskilled workers has been the company’s fastest-growing sales channel for three straight years, growing 15 to 20 percent annually, and brought in the equivalent of over a billion euros in the last fiscal year, up from the year before. The company also reports its customer happiness score has climbed to 89 percent, up from 60 percent before the AI assistant existed.
None of that erases the fact that the broader business has had layoffs elsewhere. Public reporting notes the company’s corporate structure cut hundreds of office-based roles the same year, unrelated to this program. Those cuts were not tied to AI, and they did not touch the remote-sales team. The honest version of this story is not “this company never cuts jobs.” It is narrower and more useful: the specific team affected by this specific AI rollout kept their jobs, moved into higher-value work, and helped build a channel worth over a billion euros a year.
Why this matters beyond one company
At Abstra, we see a version of the same fear in almost every conversation about adding AI to an engineering or support function: leadership wants the efficiency, but the team hears “fewer of us.” That fear is not irrational. It is also not the only outcome available.
This is our position, not just an observation about one retailer. When a tool makes a piece of work faster or cheaper, demand for that work has historically grown rather than shrunk. Economists have called this the Jevons paradox since the 1800s, and we think the pattern holds for engineering now. The narrow tasks disappear. The roles built on judgment, deciding what an AI system should own, reviewing what it produces, staying accountable for what ships, tend to grow instead. We laid out that argument in more depth in AI and jobs: what the Jevons paradox reveals about the future of work.
This case works because someone did the harder analysis before making the harder promise. They looked at what the AI could not do well, found the adjacent, higher-value work hiding behind that gap, and built a training path into it. That is a design decision, not a lucky break, and it is the kind of decision that benefits from technical partners who have built AI-augmented systems before, not just deployed a chatbot and hoped.
This is where good AI implementation and good nearshore engineering practice overlap more than people expect. Both depend on scoping the work honestly: what should the machine own, what should stay human, and what does the team need to learn to do the second part well. We wrote about that overlap directly in nearshore in the AI era: the part that stays human, and about the judgment AI still cannot replace in the rise of agentic AI and why human-on-the-loop is the new standard.
A question worth sitting with before the next AI rollout
This story is not a template anyone can copy exactly. Few companies have the balance sheet, the timeline, or the adjacent revenue opportunity that made a two-year retraining program worth the wait. But the underlying question travels well: when a piece of work gets automated, does anyone in the room ask what the freed-up people could do instead, or does the conversation stop at headcount?
So maybe the useful question is not whether AI will change a given role. It probably will. It is whether anyone has looked hard enough at what’s left over to find where the people go next.
FAQ
- What is an AI reskilling strategy? An AI reskilling strategy is a deliberate plan to retrain employees whose work is partly automated into new roles that use the capacity AI freed up, rather than reducing headcount. One well-documented retailer used this approach when its AI assistant began handling routine customer service questions.
- Did that company really avoid layoffs when it introduced AI? For the specific team affected, yes. The company retrained around 8,500 call center employees into remote design consultant roles instead of cutting those positions, according to public reporting. The broader business has had separate corporate layoffs since, but those were not attributed to AI and did not affect the remote-sales team.
- How much revenue did this AI reskilling strategy generate? Public reporting puts the remote-sales channel, staffed largely by the reskilled customer service team, at the equivalent of over a billion euros in the most recent fiscal year, making it the company’s fastest-growing sales channel for three consecutive years.
- Can smaller companies apply the same AI reskilling strategy? The core idea scales down even if the exact program does not. Any company automating part of a role can ask what work is left over that needs human judgment, and whether existing employees could be trained into it, rather than defaulting to reducing the team.
- What should a company look for in an AI implementation partner? A partner who can help identify what should stay human before deciding what to automate, not one who treats every AI rollout as a cost-cutting exercise by default.