When a business owner asks what AI means for their team, the question usually gets framed as a number: how many roles can this replace. That's the wrong starting point, and there's now real evidence it's an expensive one too. Evidence so far suggests that deeper workforce cuts do not, by themselves, predict better AI returns. The businesses seeing genuine value redesigned what their roles actually do, moved routine tasks onto AI, and rebuilt jobs around new combinations of skills that a task-by-task automation approach couldn't replicate on its own. That's a more interesting story than a layoff number, and it's already playing out at real companies on both sides of the Atlantic.
Why the headcount-first approach keeps disappointing
It's worth starting with the evidence that cutting first and figuring out the rest later doesn't reliably work, because it reframes everything that follows. A Gartner survey of roughly 350 executives at large enterprises, businesses with at least $1 billion in revenue using AI agents, intelligent automation, and similar autonomous technologies, found that approximately 80% of organizations piloting or deploying these autonomous business capabilities reported workforce reductions, but the level of reduction was about the same regardless of whether the deployment delivered high or low return on investment. Separately, Gartner has predicted that half of companies that cut customer service staff specifically because of AI will end up rehiring for similar functions by 2027, and Forrester expects roughly half of AI-attributed layoffs industry-wide to be quietly reversed, though the replacement roles may come back offshored or at lower pay. A survey of 600 HR leaders conducted by Careerminds, covering layoffs made in the year to February 2026, found only 8.4% said the restructuring delivered what was promised and that they'd do it again.
None of this means AI doesn't reduce the need for certain work. It means that treating "how many people can we remove" as the primary question tends to produce restructuring that doesn't hold up, because it skips the harder and more valuable question: which specific tasks, not which whole roles, can actually move to AI, and what does that free the remaining people up to do instead.
The IKEA example: reskilling instead of reducing
The clearest, most concrete version of doing this well comes from Ingka Group, the largest IKEA franchisee, based in Sweden with major operations in the Netherlands. When Ingka deployed an AI chatbot called Billie across its customer service channels, the bot was resolving close to half of all customer inquiries within two years. Rather than treating that reduction in routine inquiry volume primarily as a reason to cut customer-service headcount, Ingka reskilled roughly 8,500 call-center co-workers into remote interior design consultants; the resulting remote design business later reported more than €1 billion in annual sales. The AI didn't replace the team; it removed enough routine inquiry volume that the same people could be redirected into a different, sales-oriented skill combination that the company built specifically around the capacity the automation freed up.
Worth being precise about the fuller picture here, since this example gets cited a lot and deserves an honest update: in 2026, both Ingka Group and Inter IKEA separately announced layoffs, roughly 800 and 850 roles respectively, in office and Group Functions roles. The companies attributed these specifically to organizational simplification and faster decision-making rather than to the customer-service AI reskilling story, and they're worth noting as a separate development rather than a reversal of the Billie result. The reskilling case remains a genuine, well-documented example of redesigning work around AI rather than just cutting it; it just isn't the whole story of that company in 2026, and no single case study should be treated as a permanent guarantee.
What's happening across companies more broadly
A September 2026 analysis from the workforce intelligence firm TalentNeuron looked at seven large enterprises spanning both the US and Europe, including Salesforce, Klarna, Wells Fargo, Google, Microsoft, Citi, and BT Group, and found something that cuts against the simple job-elimination narrative: combined demand for HR, strategic workforce planning, and people-analytics roles across these companies rose 16% over two years, precisely the kind of role needed to manage a workforce being actively redesigned around AI rather than just shrunk.
The same research included a detailed case with a Fortune 100 manufacturer that had initially flagged a set of roles for full elimination. Analyzing those roles at the task level rather than the job level found that 34% of the flagged roles actually contained tasks requiring human judgment that were central to the company's own transformation strategy. Removing those roles wholesale, rather than reallocating the judgment-heavy tasks within them, would have eliminated capabilities the company needed for the very transformation it was trying to execute. TalentNeuron's own conclusion is worth stating in its actual, more careful form: no job and no human is 100% automatable, but plenty of individual tasks are, which is a meaningfully different and more precise claim than saying no employee can ever be replaced.
The employer data is more positive than the layoff headlines suggest
It's worth balancing the cautionary examples above with what employers are actually reporting about growth, since the loudest headlines tend to be layoffs rather than the reverse. ZipRecruiter's own 2026 economic research, based on a survey of more than 1,000 US employers, found that 35% expect AI to increase their total headcount going forward, and 24% say that's already happening, a genuine counterpoint to the assumption that AI is primarily a job destroyer. The same research found 74% of those US employers now view AI skills as a strong advantage or an outright requirement for at least some roles, which is really a statement about how job descriptions are being rewritten around new skill combinations rather than simply shrinking.
The skill-stack idea, and the live debate around it
There's a useful, if informal, framing for what's actually changing inside a restructured role, popularized in online business and career commentary rather than academic research, worth naming honestly as that: "skill stacking," most associated with creator and business writer Dan Koe, the idea that combining several skills, none individually rare, into one person's toolkit produces a combination that's harder to replicate than any single skill on its own. Applied to a business role, the logic is that AI commoditizes narrow, single-task skills quickly, drafting a first pass, formatting a report, transcribing a call, but the specific combination of skills a given role needs, industry knowledge plus client relationships plus the judgment to know which automated output to trust, remains genuinely harder to replace as a bundle.
It's worth noting there's a real counter-argument in the same online discourse, not a settled consensus: some commentators argue that broad, shallow generalism is now the risky path specifically because AI gives everyone access to basic competence in any single skill cheaply, meaning the safer bet is deep specialization in the areas AI struggles with rather than a wide, moderate skill stack. Both views are worth holding loosely rather than treating as settled; the practical, task-level evidence from TalentNeuron's research above is a sturdier foundation for actual restructuring decisions than either side of that debate on its own.
What this actually means for redesigning a role
The practical version of all this, for a business actually restructuring a role rather than reading about one, is to work at the task level the way the manufacturer in the TalentNeuron case did, not the job level. Take a role, list what it actually does across a representative week, and sort those tasks into what's genuinely repeatable and automatable versus what depends on judgment, relationship, or context that doesn't reduce to a repeatable process. We cover that specific split, and how to calculate the financial side of it, in AI Automation vs Hiring: The Real Financial Comparison and in Can AI Replace an Employee?.
What's specific to this article is the next step, the one IKEA's case and the TalentNeuron research both point to: once the automatable tasks move to AI, the interesting question is what new combination of remaining skills becomes valuable with that freed-up time, not just how much smaller the team can get. A support role that loses its routine-inquiry volume to AI doesn't automatically become a smaller support role; it can become a different role entirely, one that combines the relationship and product knowledge it always had with a new consultative or sales-adjacent skill the business couldn't previously afford to build into that position.
A framework for redesigning, not just reducing
This is a Kubera AI planning heuristic, not a universal benchmark, meant to structure a redesign decision rather than default to a headcount target.
The Kubera Role Redesign Model works in three steps:
- Task-level audit, not role-level. List what the role actually does across a representative week, and classify each task as automatable execution work or judgment work that doesn't reduce to a repeatable process, the same split covered in more depth in our Execution vs Judgment framework. Skipping straight to "how many of these roles do we still need" is exactly the mistake the Fortune 100 manufacturer case above illustrates.
- Identify what the freed capacity is actually worth doing. Once routine tasks move to AI, ask what new skill combination the business could use more of, the way Ingka's freed capacity became design consultation rather than just fewer support agents. This step is where most restructuring stops short, treating freed time as a cost saving to bank rather than a capacity to redeploy.
- Redesign the role deliberately before deciding headcount. Only after steps one and two does a genuine headcount question become answerable, and it often produces a different answer than starting with a target number would have.
Where this plays out in practice
Illustrative scenario, not a specific Kubera client: a mid-size business with a small customer support team automates routine order-status and scheduling inquiries, the way covered in our WhatsApp automation guide and voice agent guide. Rather than reducing the team by the volume automated, the business identifies that its support staff already have strong product knowledge and client rapport, and redirects the freed time toward proactive account check-ins and upsell conversations the team never previously had capacity for. The team stays the same size; what the role actually consists of changes meaningfully, and the business gains a revenue-generating capability it didn't have before, not just a smaller cost line.
FAQ
Does this mean AI never leads to fewer jobs? No. Some roles genuinely shrink or disappear where the underlying work was almost entirely automatable to begin with. The point is that treating headcount reduction as the primary goal, rather than a possible outcome of a proper task-level redesign, is the approach with a poor track record so far.
Is the IKEA reskilling story still true given the 2026 layoffs? The reskilling result itself, roughly 8,500 workers moved into a remote design consulting business that later reported more than €1 billion in annual sales, is a separate, well-documented case from the 2026 layoffs, which the companies attributed to unrelated organizational simplification. Both are part of the same company's story; neither cancels the other out.
What's the biggest mistake businesses make when restructuring around AI? Analyzing roles at the whole-job level instead of the task level. The TalentNeuron research above found over a third of roles flagged for full elimination at one company actually contained judgment tasks the business needed, a mistake only visible once you look at tasks rather than job titles.
Is "skill stacking" a proven strategy or just a popular idea? It's a popularized framing from online business and career commentary, not an academic or peer-reviewed finding, and there's genuine disagreement even within that commentary about whether broad skill combinations or deep specialization is the safer bet. Treat it as a useful way to think about role design, not a settled formula.
Do new roles actually get created, or is this just cushioning layoffs? Evidence points to genuine new demand in some categories: workforce-planning, HR, and people-analytics roles specifically grew across the companies studied by TalentNeuron, and ZipRecruiter's research found a meaningful share of employers expect AI to increase total headcount, not just hold it steady.
How do I figure out which tasks in a role are safe to automate? Start with tasks that are repetitive, well-documented, and don't require relationship continuity or case-specific judgment. We cover this split in more detail, including a scoring approach, in Can AI Replace an Employee?.
Should a small business bother with this kind of redesign, or is it only relevant for large enterprises? The task-level principle applies regardless of company size, though a small business will naturally do this analysis on a handful of roles rather than at the scale of the seven-company TalentNeuron study. The underlying question, what should this specific role look like once routine tasks move to AI, is the same either way.
What happens if we get the restructuring wrong? The data above suggests this is common: roughly half of AI-attributed layoffs may eventually reverse per Forrester's estimate, and only a small share of surveyed HR leaders who'd already restructured said it delivered as promised. Getting the task-level analysis right before changing headcount is cheaper than reversing a cut later.
Does automating routine tasks always free people up for higher-value work, or does the business need to plan for that deliberately? It needs to be planned deliberately. Freed-up time that isn't redirected toward a specific new capability tends to just dissipate rather than turn into the kind of result Ingka achieved with its design-consultant reskilling.
If you're trying to work out which tasks in a specific role are genuinely ready to automate, and what the freed-up capacity could actually be redirected toward, that's exactly the kind of redesign worth mapping out before any headcount decision gets made.