Ask any leadership team about AI skills right now, and you'll hear some version of the same line: we need people who have them, and we're prepared to pay for that.
That made me curious. Is skills-based pay actually happening? The answer is complicated.
Even though AI is rewriting what a role requires faster than most job descriptions can keep up, pay is the slowest system in most organizations to adjust.
The Evidence
The Conference Board's newest salary budget survey found that AI skills are now the single biggest driver of base pay adjustments, ahead of leadership skills and data science. Not a nice-to-have. The top driver.
At the same time, PwC's latest Global AI Jobs Barometer found that workers with AI skills command a wage premium of over 60% in the open market. That's not a rounding error. That's the market pricing a real, measurable gap in capability.
And yet Payscale's 2026 Compensation Best Practices Report found that even though 6 in 10 companies have already rewritten job descriptions to call for AI skills, barely half of them are actually paying more for it, in base pay, bonus, or equity.
So the market is pricing the skill. Companies are naming the skill. And the evidence is mixed about whether employees are being rewarded for it.
The Answer You Are Most Likely to Hear
The most reasonable explanation is that companies are actively paying for critical skills; they are just doing it ad hoc: a manager pushes through a one-off exception, a counteroffer gets approved to stop someone from walking, a favored employee gets a "market adjustment" that never gets explained to anyone else. It’s the path of least resistance, and it’s fraught with risk.
Ad hoc isn't a fix. It's the absence of one, dressed up as flexibility. Every exception runs through whoever happens to be in the room when the decision gets made, with no consistent criteria for who gets it and who doesn't. That's not a stopgap while the real system gets built. It's how a pay equity problem gets built, one exception at a time.
The question worth asking isn't why managers keep reaching for exceptions. It's why exceptions are the only tool most of them have.
The Reality
Most pay structures are still built around jobs, not skills. A person's pay is tied to their title and level. Level is tied to a job description that was written months or years ago. If a skill doesn't show up in that job description, there's no lever anyone can pull to pay for it, no matter how much the market says it's worth.
So when a leader decides to reward AI capability, they have two bad options. Bend the merit budget to cover it, which waters down everyone else's raise. Or reach for the same ad hoc exceptions covered above, which is how the risk gets baked in rather than avoided.
This isn't a budget problem. It's a design problem. The structure has no place to put the money.
A Company Already Rebuilding This
IBM is a great example of a company that shows that a fix for this is buildable, not theoretical.
IBM built its reputation over the last decade on treating skills, not degrees, as the credential that counts. Its "New Collar" initiative removed degree requirements from roughly half of its U.S. job postings, down from 95% of roles requiring a four-year degree before the shift, and its digital badge system lets employees validate specific capabilities (cloud, security, AI, and more) that factor into advancement and pay conversations, independent of job title.
But every company doesn't need to be IBM. One company's program doesn't prove the model works everywhere. What IBM's approach does prove is narrower and more useful: a large, complex organization can build a real mechanism for recognizing skills that a job description never anticipated. The barrier other companies face isn't that this is impossible. It's that most haven't built the structure yet.
What Actually Fixes It
A few things separate the companies who are closing this gap from the ones still stuck naming skills they can't pay for:
- Define the handful of skills that actually move business results, and define them as capabilities a person can demonstrate, not as a line buried in one job description.
- Build a separate, small adjustment pool for skill-based pay, distinct from the annual merit budget, so rewarding a critical skill doesn't come at the expense of everyone else's raise.
- Tie the skill definition to what someone can actually do, not to keywords lifted from a job posting. A title with "AI" in it is not the same thing as AI capability.
None of this requires a bigger budget. It requires a structure that allows you to spend within budget on the things you say you value.
The Governance That Keeps This Honest
A separate pool solves the mechanism problem. It doesn't automatically solve for fairness, and left unmanaged, it can quietly become a new source of inequity instead of a fix for the old one. A few things keep it honest:
- Use objective, verifiable criteria (certifications, assessments, demonstrated project outcomes), not manager nomination. If "having the skill" is subjective, the pool inherits every bias already built into performance reviews.
- Run decisions through a calibration group, not individual manager sign-off, the same way you'd calibrate performance ratings.
- Set a fixed amount for each verified skill instead of negotiating case by case. Negotiated premiums produce unequal ones.
- Audit the pool itself on a regular cadence, by gender, race, tenure, and function, separate from your annual company-wide pay equity review.
- Track who gets access to the stretch projects and training that lead to certification. If that pipeline is uneven, the pay outcome will be too, no matter how clean the criteria look on paper.
Qspired's Take on Valuing Skills
The gap between naming a skill and paying for it isn't a talent problem. It's an architecture problem, and architecture problems are solvable.
This is not a failure to value skills. It's a failure to build pay systems that can act on that value.
If you handed every leader in your organization a dollar today to reward a critical skill that isn't written into anyone's job description, what would they reward? Without a shared answer to that question, you don't have a skills strategy. You have as many strategies as you have leaders.
The window to get ahead of this is now, not later. Every quarter AI reshapes more roles, the gap widens between what companies pay for and what they actually need. The organizations that close it first will not just pay more fairly. They will be the ones equipped to move talent and dollars toward the skills that matter, while competitors are still pricing job titles that no longer describe the work.
This is the work we do at Qspired: helping organizations build compensation strategy, job architecture, and skills-based talent systems that hold up as AI reshapes how work actually gets done.