From Sanskriti Khandelwal | Product & Market Analysis
Compensation Trends 2026: Flat Merit Budgets, a Steep AI Skills Premium
On this page
Three of the largest salary budget surveys in the US agree to within a tenth of a point: employers plan to raise pay by about 3.4% to 3.5% in 2027. In the same year, job ads asking for AI skills advertised pay between 28% and 62% higher than comparable ads without them. Both findings are sourced, and both can be true at once. The compensation trends story for 2026 is the mechanism that connects them.
Key takeaways
- The big three salary budget surveys landed within 0.1 points of each other. WTW reports 3.4% for 2027 against a 3.5% actual for 2026, while Mercer and Payscale both report 3.5% total budgets for 2027.
- The median raise no longer beats inflation. US consumer prices rose 3.7% in the 12 months to September 2026, above every 2027 total salary budget in this comparison.
- The AI skills premium and the merit budget measure different things. Lightcast and PwC compare advertised pay across job ads, while merit budgets measure the yearly change for people already on payroll.
- Funding the premium from a flat pool squeezes everyone else. In an illustrative 3.5% pool, giving 5% of payroll a 15% market adjustment leaves about 2.9% for the remaining 95%.
Compensation trends 2026: three salary budget surveys, one number
Direct answer. US employers plan salary increase budgets of about 3.4% to 3.5% for 2027, per WTW, Mercer and Payscale surveys run between March and July 2026. Mercer puts the merit portion at 3.2%. Job ads requiring AI skills advertise pay 28% to 62% above comparable ads, per Lightcast and PwC. Those premiums are paid mostly outside the merit budget.
The agreement between the three surveys is the first thing worth noticing. They use different samples, different fielding windows and different definitions. They still converge on the same narrow band.
WTW surveyed companies between March and May 2026 and received 34,024 responses from 156 countries, including 1,650 US organizations. It found average US budgets of 3.4% planned for 2027, slightly below the 3.5% actually delivered in 2026. Lori Wisper of WTW described the outlook as the "land of 3%" for the foreseeable future.
Mercer's July 2026 QuickPulse covered 1,001 US organizations across 15 industries. It found a 3.2% merit budget and a 3.5% total budget for 2027, the fourth consecutive year of similar numbers. Payscale's survey closed on 12 June 2026 with 1,266 usable submissions and reported 3.5% for 2027, up from 3.4% in 2026.
Merit budget 2026 and total budget are not the same line
Merit is the portion of the pool tied to individual performance ratings. The total budget adds promotions, cost-of-living adjustments and market adjustments. When you read "salary increase 2027" in a headline, it is usually the total figure.
The gap matters for anyone planning a raise cycle. Mercer's 0.3 point difference between merit and total is where promotions and special adjustments live. That small residual is also where most companies would have to find money for an AI skills premium, if they paid it through base salary at all.
Projections drift down, slightly and consistently
Planned budgets are not final budgets. Mercer's October 2025 survey expected 3.2% merit and 3.5% total for 2026. Its March 2026 survey of 756 employers found 3.1% merit and 3.4% total actually paid.
A tenth of a point is small. It is also directional, and it has run in one direction. I would budget for the 2027 actuals to land a tenth below the July numbers, not above them.
The merit budget against inflation
The Bureau of Labor Statistics reported on 14 October 2026 that consumer prices rose 3.7% in the 12 months ending September, up from 3.4% a month earlier. That is above every 2027 total budget in this comparison.
The timing is awkward for compensation teams. The surveys were fielded between March and July, before the latest acceleration. A budget set in a 3% inflation world is now being approved in a 3.7% one.
For your employees, the practical reading is simple. An average performer receiving a 3.2% merit increase is taking a small real pay cut if prices keep moving at September's pace. That is why the AI premium question carries an edge. It is a debate about who gets protected from inflation and who does not.
WTW asked companies what was holding budgets back. 32% cited cost management, 28% a tighter labour market and 27% inflation concerns. WTW also found 69% of employees staying with their current employer and only 22% of companies adding headcount.
Low quit rates remove the usual pressure on budgets. When people are not leaving, the cheapest response is to hold the pool flat and spend selectively. That is exactly what the rest of the data shows employers doing.
How each salary budget survey is built
You cannot judge a compensation number without its method. The table below puts all five sources in this post on the same page, with what each one actually measures.
| Source | Headline figure | What it measures | Sample and window |
|---|---|---|---|
| WTW Salary Budget Planning | 3.4% for 2027, 3.5% actual 2026 | Average US salary increase budget | 34,024 responses, 156 countries, 1,650 US; March to May 2026 |
| Mercer QuickPulse, July edition | 3.2% merit, 3.5% total for 2027 | Average US budgets, zeros included | 1,001 US organizations, 15 industries; mid-July 2026 |
| Payscale Salary Budget Survey | 3.5% for 2027, 3.4% in 2026 | Average total base-pay increase | 1,266 submissions, US, Canada and 19 other countries; closed 12 June 2026 |
| Lightcast, Beyond the Buzz | 28%, about $18,000 a year | Advertised salary gap, ads with AI skills vs without | More than 1.3 billion job postings; published July 2025 |
| PwC Global AI Jobs Barometer | 62%, up from 57% | Advertised wage premium for AI skills | More than 1 billion job ads, 27 countries and territories; June 2026 |
Budget figures are employer plans and actuals reported by HR and compensation teams. Premium figures are derived from advertised pay in job postings, not payroll. The two groups answer different questions, which is the argument of the next two sections.
All three budget surveys are self-reported by employers, usually by the compensation function. They capture planned and delivered pools, averaged across the whole workforce. They say nothing about how the pool is split between people.
That last point is the one most coverage skips. An average of 3.5% is consistent with everyone getting 3.5%. It is equally consistent with a tenth of the workforce getting 10% and the rest getting less than 3%.
Lightcast and PwC both work from posted job advertisements. They compare the salary advertised in postings that list AI skills against postings that do not. Neither study observes what any individual is paid after hiring.
What the AI skills pay premium actually measures
Lightcast's July 2025 report analysed more than 1.3 billion postings and found that postings including AI skills offered 28% higher salaries, nearly $18,000 a year. It also found 51% of postings requiring AI skills sat outside IT and computer science occupations as of 2024.
PwC's 2026 Barometer found a much larger figure. It measured an average 62% premium for AI skills, up from 57% in the prior edition. The range across industries ran from 16% in government and public sector work to 118% in consumer markets.
Why the two premium estimates differ by more than double
The press materials do not give enough method to reconcile 28% and 62%. The likely sources of the gap are the comparison group, the country mix and the time window. Lightcast's figure is a year older and its release does not specify controls for occupation or seniority.
I would not average them. Treat 28% as a lower anchor from an older, larger posting set, and 62% as the current upper anchor. Either one is an order of magnitude larger than any gap inside the merit budget data.
Why the AI skills premium and the merit budget cannot be subtracted
The tempting headline is "AI talent gets 28% while everyone else gets 3.2%". That sentence compares a level with a rate of change. It is the most common error in this topic.
A premium is a snapshot. It says that, on a given day, an ad for a role requiring AI skills offers more than a comparable ad without them. A merit budget is a flow. It says how much the pay of people already employed rises over a year.
Someone hired at a 28% premium in 2026 may then receive the same 3.2% merit increase as everyone else in 2027. The premium is paid once, at the point of hire or promotion, and the merit budget never sees it.
The industry spread inside the budget surveys is tiny
Mercer's merit budgets by industry run from 2.9% in Consumer Goods to 3.8% in High Tech. That full spread is 0.9 points. PwC's premium spread across industries is 102 points.
This is the core answer to whether AI talent is pulling pay away from everyone else. In the budget data, it cannot be seen. In the posting data, it is enormous. The divergence lives in the gap between the two methods, and you will only see it in your own payroll.
How an AI skills premium pulls pay apart inside a fixed pool
If the total pool holds at 3.5% and part of it is spent on market adjustments for AI-skilled staff, the rest of the workforce gets less. The arithmetic is unforgiving and simple to run.
The scenarios below are illustrative, not measured. They assume a $10 million payroll, a 3.5% total pool and that every adjustment comes from inside that pool.
| Share of payroll adjusted | Adjustment size | Cost | Average left for everyone else |
|---|---|---|---|
| None | None | $0 | 3.50% |
| 5% ($500,000) | 10% | $50,000 | 3.16% |
| 5% ($500,000) | 15% | $75,000 | 2.89% |
| 10% ($1 million) | 15% | $150,000 | 2.22% |
| 10% ($1 million) | 20% | $200,000 | 1.67% |
Calculation: remaining pool divided by the unadjusted share of payroll. For the third row, ($350,000 minus $75,000) divided by $9.5 million equals 2.89%. These are not survey results.
The third row is the one I would expect a mid-sized software company to recognise. Closing roughly half of Lightcast's 28% gap for a small group drops everyone else below 3%. With CPI at 3.7%, that is a visible real pay cut for the majority.
Promotions shrink the residual further
Mercer reports employers expect to promote about 8.4% of their workforce in 2027, down from 8.6% in 2026 and 9.9% in 2025. Fewer promotions free up a little money. They also close one of the main routes by which a skills premium reaches existing staff.
Mercer also reports average salary structure adjustments of 2.6%. Ranges are moving more slowly than merit, which means AI-skilled hires priced at market will sit higher in their range from day one.
Where the AI skills premium actually gets paid
If the premium does not show up in merit budgets, it has to go somewhere. The WTW data gives the clearest clue about where.
WTW found that 36% of US companies are hiring at higher salary ranges, 34% are increasing retention bonuses or spot awards, and 32% are raising starting salary ranges. Each of those actions sits outside the annual merit pool.
Brittany Innes of WTW described employers as becoming more deliberate, targeting "the roles, skills and talent segments that matter most." That is a polite way of describing a two-speed pay system.
Variable pay is the preferred channel
HCA Magazine reported in October 2026 that employers are routing AI pay premiums through bonuses and performance pools rather than base pay. The same piece cited a Randstad survey in which more than 63% of employees said their pay remains tied strictly to traditional job titles and bands. That Randstad figure reached me through secondary coverage, so treat it as directional.
Variable pay is the rational choice for an employer unsure whether a skill premium will last. A bonus can be cut next year. A base increase compounds in every future merit cycle, pension contribution and severance calculation.
Payscale found only 32% of US companies plan across-the-board increases in 2027, down from 36% in 2026. Mercer's March data had just 4% of its participants giving equal, across-the-board increases. Pay is becoming more differentiated even where budgets are flat.
My position: this is the real compensation trend of 2026. Budgets are not growing, and distribution inside them is becoming steeper. AI skills are the most visible reason for that, not the only one.
The same pattern shows up in other places this blog has tracked. The spread between senior and entry-level engineering pay is covered in the analysis of AI engineer salary spread. The question of which human capabilities keep their value is the subject of the piece on the tacit knowledge premium.
Where this argument is weakest
The case above rests on comparing two kinds of evidence that were never designed to be compared. Here is where it can fail.
Posted pay is not paid pay
Both premium figures come from job ads. Ads that mention AI skills may be disproportionately for more senior or specialised roles, which pay more for reasons unrelated to AI. The press materials I could open do not state how either study controls for that. Payroll evidence on the same question is mixed, as set out in the review of what payroll data shows about AI and jobs.
The budget surveys may already contain the premium
Mercer's High Tech merit budget leads at 3.8%. Part of that lead may be AI-related market pressure, captured exactly where the budget surveys can see it. If so, the premium is not hidden, just diluted by averaging across thousands of non-AI roles.
The diffusion counter-case is real
Lightcast found 51% of AI-skill postings sat outside IT in 2024. If AI fluency becomes a baseline expectation across most roles, the premium compresses towards zero. In that world, the fixed-pool squeeze I describe is a transitional cost rather than a lasting divide. I think that is the most likely long-run outcome, and I would not bet on its timing.
Finally, the illustrative scenarios assume the adjustment comes from inside the pool. Many companies fund market adjustments from a separate line. If yours does, the squeeze lands on margin instead of on colleagues.
Frequently asked questions
What is the average salary increase for 2027?
US employers plan average salary increase budgets of about 3.4% to 3.5% for 2027. WTW reports 3.4% from 1,650 US organizations, Mercer reports 3.5% total and 3.2% merit from 1,001 organizations, and Payscale reports 3.5% from 1,266 submissions. These are planned budgets surveyed between March and July 2026, and Mercer's past actuals have landed slightly below its planned figures.
What is the merit budget for 2026?
Mercer's March 2026 survey of 756 US employers found an average merit increase of 3.1% actually paid in 2026, against 3.2% planned the previous October. The total increase, including promotions and other adjustments, was 3.4%. WTW reports a 2026 actual total of 3.5%, and Payscale reports 3.4%. All three averages include employees who received no increase.
How much more do AI skills pay?
Lightcast's July 2025 analysis of more than 1.3 billion job postings found ads listing AI skills offered 28% higher salaries, about $18,000 a year. PwC's June 2026 Barometer, covering more than 1 billion ads in 27 countries, found a 62% average premium. Both figures measure advertised pay in job postings, not what employees are paid after hiring.
Are salary increases keeping up with inflation in 2026?
Not for the average employee. US consumer prices rose 3.7% in the 12 months to September 2026, according to the Bureau of Labor Statistics. Planned 2027 salary budgets of 3.4% to 3.5% sit below that rate, and the 3.2% merit portion sits further below it. Whether that becomes a real pay cut depends on where inflation settles during 2027.
Which salary budget survey is most reliable?
None is clearly superior, and their agreement is the useful signal. WTW has the largest global sample, Mercer separates merit from total budgets and publishes actuals each spring, and Payscale includes smaller organizations. Use Mercer for the merit and total split, WTW for sample breadth, and compare all three against your own industry rather than relying on one average.
How do companies pay the AI skills premium if budgets are flat?
Mostly outside the merit budget. WTW found 36% of US companies hiring at higher salary ranges, 34% increasing retention bonuses or spot awards, and 32% raising starting ranges. Trade press reporting in October 2026 described employers routing AI premiums through bonuses and performance pools. Those channels can be reduced later, which base salary increases cannot.
Where to start before your 2027 cycle closes
Most 2027 budgets are not final. Mercer found 87% of organizations still had preliminary numbers in July, so there is time to change how yours is built.
Split the pool on paper before anyone allocates it. Write down the merit line, the promotion line and a separate market adjustment line, with the roles that would draw on the third. If the third line comes out of the first, run the table above with your own payroll and show the leadership team what the average performer will receive against 3.7% inflation. That one number tends to end the argument about whether to fund adjustments separately.
Then decide, role by role, whether the premium belongs in base or in variable pay. The answer depends on whether you believe the skill will still be scarce in three years. For the hiring side of that judgement, the guide to interview questions that test AI fluency is the practical companion, and the piece on performance review criteria for AI-assisted work covers how to rate people once they are in.
Related on this site
For how hiring signals are shifting at the bottom of the ladder, read why entry-level hiring data conflicts.
References
- WTW, Employers shift from bigger pay budgets to smarter pay strategies, WTW finds, 15 July 2026. Used for US budgets, sample, caution factors and off-budget actions.
- Mercer, US employers stick with moderate pay increases for 2027, 31 August 2026. Used for 2027 merit and total budgets, industry merit spread, promotions and budget status.
- Mercer, Annual increases paid out as expected, 27 April 2026. Used for 2026 actual merit and total increases and across-the-board share.
- Payscale, 2026-2027 Salary Budget Survey, 4 August 2026. Used for 2026 and 2027 budgets, sample and across-the-board share.
- Lightcast, AI skills command 28% salary premium as demand shifts beyond tech industry, 23 July 2025. Used for the 28% premium and the share of postings outside IT.
- PwC, 2026 Global AI Jobs Barometer press release, 15 June 2026, read via its PR Newswire distribution because the PwC page refused automated access. Used for the 62% premium and industry range.
- US Bureau of Labor Statistics, Consumer Price Index, September 2026, 14 October 2026. Used for 12-month CPI-U inflation.
- HCA Magazine, AI skills carry a nearly $18,000 premium, but pay bands aren't budging, 19 October 2026. Used for the variable pay channel and the Randstad survey figure.
Weakest point in the source base: both AI premium figures were read from press releases, which do not state their controls for seniority or occupation. The PwC citation should be upgraded to the full report methodology. All budget figures are employer plans current as of October 2026 and will be superseded by spring 2027 actuals.
Related reading