Two recent reports each capture a different picture of how AI is reshaping views on staffing and skills in Sweden. One measures expectation – what business leaders believe will happen to headcount. The other measures what's already happening – what employers are actually hiring for. Read together, they reveal a tension worth pausing on.
From 16 to 91 percent in a year
A survey Novus conducted on behalf of Svea Bank, among 500 Swedish companies with at least ten employees, finds that 91 percent now expect AI to reduce the number of employees at their own company within five years. The year before, the same figure stood at 16 percent. That's a 75-percentage-point shift in twelve months – from a minority believing it to something close to consensus.
The survey was conducted by phone from 3–24 October 2025 and published by Svea Bank in May 2026, feeding into the report "Drivkraft: AI".
An expectation running ahead of the evidence
That shift is worth placing next to another figure we've written about before: BCG's survey of more than 300 Nordic business leaders found that only 4 percent of Nordic companies see meaningful return on their AI investments so far. Two figures, the same period, the same market – pointing in different directions. Expectation that AI will cut headcount has raced to near-total agreement, while evidence that AI is actually delivering measurable value remains rare.
Expecting an outcome isn't the same as having worked out how to get there. Planning headcount reductions on AI's promise, before the workflow meant to carry that reduction has actually been rebuilt and tested, draws the conclusion before the evidence.
What the labour market already shows
A report from TechSverige, based on more than 16,000 job postings from Arbetsförmedlingen, Sweden's public employment service, offers a different angle. Demand for AI skills in job postings has grown 328 percent since 2016, and the number of professions where AI is mentioned in postings has nearly doubled, from 48 to 92. IT roles still account for roughly 60 percent of all AI-related demand, but the spread into analysis, engineering, business development and administration is clear.
This isn't a measure of expectation. It's a measure of what employers are actually spending money and hiring time on right now. And it doesn't point toward a narrower labour market, but a broader one: more roles, not fewer, are expected to work with AI as part of everyday work.
Two sides of the same shift
Placed side by side, the two reports tell a more nuanced story than "AI takes jobs." Business leaders expect AI to shrink headcount, while the labour market is simultaneously demanding broader AI skills across more professions, not a narrower specialist group. That suggests the real change isn't primarily about fewer employees – it's about what's expected of the people who remain.
The skill in demand is rarely a certificate. It's the ability to actually work with AI inside a workflow you own and understand – not to have heard of it. That's the same conclusion we drew from Solita's survey on the Nordic AI gap: having access to AI isn't the same as having rebuilt the work around it.
Where Nodal comes in
Our view is simple: don't plan headcount reductions on AI's promise. Plan them, if at all, on a workflow that has actually been rebuilt, measured, and shown to hold up. That order is what separates a company getting real value from AI from one that's merely expecting it.
Wondering what AI can realistically do – and not do – for your staffing and workflows? Get in touch with us at Nodal, and we'll have a straight conversation about where you actually stand.
