AI skills carry a nearly $18,000 premium, but pay bands aren’t budging

Many employers aren’t refusing the premium outright. They are routing it through variable pay, meaning bonuses and performance pools that can be scaled back.

In the US digital sector, base salary growth has normalized at 2.5 percent to 3.8 percent, but employers steer variable bonuses and performance pools toward specialists in AI automation, cloud architecture and data pipelines. In Canada, base adjustments hold at 2.6 percent to 3.9 percent, and Randstad reports that specialized AI base pay growth has matured, with employers shifting toward discretionary bonuses tied to measurable enterprise productivity.

Crofoot pointed to internal development when the market’s skills premium outpaces the budget. “When you can’t buy the talent on the market, employers need to think about developing that talent internally,” she said. “That’s when we see employers are investing in training programs, in coaching and mentoring, in developing some of those AI capabilities, and in identifying adjacent roles that could more easily move into those AI occupations or AI-adjacent occupations.”

Candidates expect that investment. According to Randstad research, 49 percent of tech candidates in the US and Canada reject job offers that lack funded certifications or continuous development.

Marie-Noëlle Morency, head of brand, content and communications at Randstad North America, said candidates now link AI expectations directly to pay and training. “[An employer] will ask for AI and [one] will expect to have some AI skills,” she said, describing candidates’ expectations. “But it will need to reflect in the comp plan, and it will need to reflect in the training offer and upskilling.”

Crofoot pointed to an underused retention lever: showing AI-skilled employees where the skill takes them next. Career pathing is a “huge tool for retention” in the AI era, she said, and like a bonus, it doesn’t lock in a permanent raise. “It’s definitely underused today across employers.”

Job architecture is the bottleneck

Crofoot said job architecture, the structure of roles, levels and pay grades that compensation teams often own, moves too slowly to price fast-changing skills.

“We are not seeing that job architecture is keeping up with the pace of skill change,” she said. “Job architectures, I think by design, are very static and it takes a long time to make any adjustments because you have so many parts of the organization that have a stake.”

Rather than wait for a redesign, organizations are defining each job as a basket of skills and flagging critical skills across a job family, Crofoot said. “It’s marrying the two and working within the constraints of slow-moving job architecture changes,” she said.

The data points to a narrower spend than a blanket premium: pay for specific, verified skills, use variable pay where base budgets can’t stretch, and fund the internal pipeline for skills too expensive to buy. That also speaks to the AI skills gap facing Canadian employers and the rising bar of AI fluency as a baseline hiring requirement.

“Look at pairing AI adoption with funded in-house training, to prevent top candidates from going elsewhere with other job offers,” said Jeff Johnson, head of marketing at Randstad North America.