Three in four employers now expect workers to hit a baseline level of AI proficiency within about two years, according to a new HiBob survey of 1,200 AI decision-makers, but the same research finds companies are already tying that proficiency to promotions and pay well before they have agreed on how to define, source, or teach it.
The skills bar just moved, and it is already gatekeeping careers
HiBob’s AI Skills 2026 report, based on a Censuswide-fielded survey run between February 3 and March 12, found that 75% of decision-makers expect moderate AI proficiency to become the standard for most non-technical roles by 2028. That expectation is not staying theoretical. Sixty-seven percent of the organizations surveyed already link AI skills to promotion criteria, and 50% tie them directly into performance ratings.
The bar is moving fastest in the Nordics, where 79% of decision-makers expect AI proficiency to become standard, followed by the UK, DACH, and Benelux at 77%. The United States sits close to the global average at 74%, while Australia and New Zealand trail at 68%. The pattern holds everywhere: AI proficiency is becoming a baseline job requirement well before most employers have built the systems to hire, train, or fairly evaluate it.
Intent is outrunning infrastructure
HiBob’s researchers describe a consistent gap between what companies say they are doing and what they can actually show. On sourcing, 68% of decision-makers say their organization has a defined strategy to find AI-skilled candidates, but only 24% use concrete levers like applicant-tracking-system tagging or dedicated talent communities to do it. Just 23% say they specifically design internal mobility or reskilling pipelines to surface AI-skilled employees already on staff, meaning most companies are chasing outside hires for a capability they may already have inside their own workforce.
Learning and development shows the identical pattern. Seventy-three percent of organizations say they invest in AI upskilling, yet no single program dominates: funded learning reaches 30% of employers, prompt and procedure libraries 29%, and protected practice time 29%. Training content is just as scattered, with no single topic used by more than 27% of organizations. The topics that do lead, workflow redesign (27%), security and acceptable use (26%), documentation (25%), and data verification (24%), show employers are training for safe, reliable execution rather than novelty. Fittingly, HiBob found that proactively reviewing AI output quality and documenting workflow decisions, each cited by 52% of respondents, rank as the most valued everyday AI behaviors, ahead of raw technical skill.
Managers are being handed a job they are not ready for
The report’s sharpest finding may be about who is supposed to make all this work. Direct managers are the group organizations most often expect to build AI capability across their teams, cited by 24% of respondents as the primary owner. But among the companies assigning that responsibility to managers, only 36% consider those managers highly prepared to deliver it. Ownership varies by region in ways that suggest no consensus model exists yet: in the US, technology providers (28%) are expected to drive AI capability slightly more than managers (24%); in Benelux, employers push it to the corporate level (25%) rather than individual managers (13%); in the Nordics, organizations lean on outside vocational education and training pathways (25%) instead.
The financial stakes of getting this right are already visible. Employers told HiBob they are willing to pay at least a 10% salary premium for scarce AI expertise, most often in automation (34%) and in AI safety, ethics, and governance (34%), with output evaluation close behind (33%). Skills that companies cannot define consistently are already being priced into offers.
HiBob’s answer: a shared framework instead of ad hoc tagging
HiBob used the findings to launch an AI Skills Framework and a companion AI Skills Assessment Guide on July 23, alongside Bob’s Skills and Jobs Catalogues, a centralized repository meant to standardize how role-critical skills, AI capability included, get defined and tracked across hiring, performance, and workforce planning.
“Too many organizations are still trying to manage a changing workforce with disconnected systems and fragmented skills data,” said Ronni Zehavi, HiBob’s CEO and co-founder. “The winners in the AI era will be the organisations that can define a shared language of skills, identify where capability truly sits, and pivot talent to where it creates the most value. That is the system we are enabling today.” HiBob, which counts more than 5,400 organizations including eToro, Fred Perry, and SmartRecruiters as customers, is effectively selling the standardization layer its own survey shows most HR teams have not built.
What this means for the HR leader
The risk here is not that companies are moving too slowly on AI skills. It is that recruiting, learning, and performance management are each defining “AI-skilled” differently inside the same organization, a recruiter tagging capability in the ATS, a learning team stocking a prompt library, a manager scoring it in a review, with no shared standard connecting the three. That is exactly the kind of inconsistency that surfaces later as a fairness or documentation problem once a passed-over promotion or a compensation decision gets challenged.
It also feeds directly into a broader capacity question HR is only beginning to own: not just who is AI-skilled, but how much of the organization’s work should run through AI agents versus people, a question we explored in HR’s newest job of planning capacity across humans and AI agents. A shared skills taxonomy is the prerequisite for making that capacity call with any rigor, rather than by department-level guesswork.
Before adopting a vendor framework or building one internally, HR leaders should audit where AI-skill judgments are already happening informally: in interview scorecards, ATS filters, and review templates. Write down the criteria each function is actually using today. HiBob’s own data suggests those criteria will not match across recruiting, learning, and performance management. The employers most exposed here will not be the ones with the least advanced AI skills strategy. They will be the ones that cannot explain, in writing, why one employee’s AI proficiency earned a raise and another’s did not.
Source: HiBob
