Divergence, not uniform growth
The single most important pattern is that average pay growth figures have stopped describing most people's experience.
Roles that are scarce, technical or hard to automate have seen strong increases. Roles that are abundant, routine or easily distributed have seen stagnation, sometimes real-terms decline once inflation is counted. The average of those two experiences describes neither.
The practical implication is that the relevant question is not whether salaries are rising, but whether yours specifically is in a category where employers are competing. Those are entirely different questions with different answers.
What AI is actually doing to pay
The effect of AI on wages so far has been less about elimination and more about redistribution.
Pay has risen sharply for people who build these systems — machine learning engineers, data engineers, infrastructure specialists — and for people who can apply them credibly within an existing domain. A marketing analyst or a lawyer who works fluently with these tools is measurably more valuable than one who does not.
Downward pressure has appeared in roles built around routine text and data production, where output is now cheaper to generate. The pattern to watch is not whole professions disappearing but tasks within them being reallocated, which changes what a role is worth without changing its title.
Remote work and geographic pay
Remote work has partially decoupled salary from location, but far less completely than early predictions suggested.
Some employers now pay a single rate regardless of where an employee lives, which has raised pay substantially for people in lower-cost regions. Others adjust by location, which limits the arbitrage. Many have pulled back towards hybrid arrangements, restoring some of the geographic premium.
The durable effect is on competition rather than on rates. Employers hiring remotely compete with a wider pool, and workers in mid-sized cities have access to salaries that were previously restricted to expensive capitals — provided their role can genuinely be done at a distance.
Persistent shortages
Several sectors have shortages that no short-term hiring adjustment resolves, because the constraint is training capacity and demographics rather than the business cycle.
Healthcare is the clearest case: ageing populations increase demand while training pipelines for clinicians take a decade or more to respond. Skilled trades face a demographic cliff as experienced workers retire faster than apprentices replace them, which is pushing wages for electricians, plumbers and technicians up steadily. Engineering shortages persist in energy, infrastructure and semiconductors.
These are the areas where wage pressure is most likely to continue, precisely because supply cannot adjust quickly.
Inflation and real pay
A pay rise below inflation is a pay cut, and this distinction has done more to shape the last few years of worker behaviour than any other factor.
When inflation rose sharply, nominal increases that looked generous in isolation still represented falling real income, which contributed heavily to job-switching and industrial action across several countries. As inflation moderates, modest nominal rises can represent genuine real gains.
When you evaluate an offer or a review outcome, subtract the local inflation rate. That figure — not the headline percentage — is what changes your standard of living.
Pay transparency is changing negotiations
A growing number of jurisdictions now require employers to publish salary ranges in job adverts, and several restrict asking candidates for their salary history.
The effect is to reduce the information advantage employers traditionally held. Candidates can see the band before applying, which makes it harder to anchor an offer to someone's previous underpayment and tends to compress unexplained gaps between colleagues.
If you are job hunting, use it. Published ranges are the strongest evidence available when you negotiate, and they are far more reliable than aggregated averages.
What to do about it
A few conclusions follow reasonably directly from the above.
Build skills in areas where supply is genuinely constrained rather than where interest is high — those are not the same thing. Learn to use AI tools well within your existing field; the premium currently sits with people who combine domain expertise with fluency in them, not with either alone.
Check your pay against the current market at least annually, since internal raises reliably lag external offers. And when you evaluate any increase, subtract inflation before deciding whether it was actually a rise.