A lot of us worry artificial intelligence will eventually take our jobs, but AI is already changing one of the most important aspects of our careers: how much money we make.

Compensation surveys suggest the average pay increase next year will be about 3.5%, in line with recent norms. The range of raises around that midpoint will be wider than usual, however.

Whether you land on the high or low end could come down to what a machine-learning model thinks you're worth.

More firms are using AI tools to comb job boards for competitors' salary ranges in an effort to pinpoint market rates. It's an exercise made easier by the expansion of pay-transparency laws. Third-party companies now offer AI salary benchmarking as a service, helping employers understand what it costs to hire and retain people in specific roles.

This could be good news if an AI model indicates you're underpaid. But if your pay is already on the high side, AI might dash your hope for a hefty raise even if you're an all-star.

'We tell you exactly how much to pay your employees'

Stello AI is among the companies analyzing vast quantities of salary data to help other businesses set their pay levels. It uses a mix of public compensation information in job listings and online forums like Glassdoor, and data purchased from payroll companies.

When Stello takes on a new client, one of its first tasks is to assess the pay of every employee and make recommendations.

"We tell you exactly how much to pay your employees," says Chief Executive Amee Parekh. A certain type of worker gets special attention: "Here are your high performers who are underpaid. If you have any budget for raises, these are the first ones you should fix."

Companies' increased focus on pay equity in recent years is a factor, Parekh adds, though the rationale behind these raises is often pragmatic. People who could make more money elsewhere are flight risks.

Stello's analyses also identify a different category of workers-those who are overpaid. These are often people who job-hopped at an opportune time and scored pay packages that are more generous than what today's tepid labor market would bear.

If you are one of them, the good news (aside from your fat paycheck) is that you are unlikely to be fired simply for making too much money. Parekh says she's never seen that happen, partly because companies understand the cost of finding and onboarding a cheaper replacement might negate any salary savings.

But you'd be wise to keep proving your worth and brace for a modest raise, if you get one at all.

The skills that pay

Good managers will explain small raises to well-paid workers, says Tauseef Rahman, U.S. workforce reward solutions leader at professional-services firm Marsh, formerly known as Mercer.

"They'll say, 'You're already at the top of the range, which means this year you will not be getting as much,'" Rahman says. "'We still value you and you know what? We should be talking about your promotion path instead.'"

This makes sense, though I wonder how people will respond to a bot's determination that they make plenty already.

A Marsh survey of more than 1,000 U.S. employers found the average planned merit increase for 2027 is 3.2%, the lion's share of a total raise budget of 3.5%. The findings of a survey by The Conference Board were nearly identical.

It's an indication that companies will pick and choose who gets a pay boost instead of evenly spreading cost-of-living adjustments across their teams.

"Many organizations are saying, 'I have to make some really tough choices,'" says Diana Scott, The Conference Board's U.S. Human Capital Center leader. "People are being more differentiated, and employers are being more intentional in terms of how they use those budgets. There's not a lot of peanut buttering, as there might have been in the past."

Companies plan to reward employees who demonstrate AI savvy, according to The Conference Board's survey. Businesses are also willing to give more sizable pay bumps to people whose leadership, judgment and interpersonal skills would be hard to replace.

New compensation model needed

It's reassuring that you don't necessarily need to be an AI whiz to up your earnings. Still, AI is messing with some of the metrics that traditionally factor into compensation decisions.

For example, many professional-services firms have long measured productivity in billable hours. Should employees' paychecks shrink if AI reduces the time required to complete certain tasks?

Conversely, do sales representatives who are judged on lead generation deserve to make more if AI enables them to identify and pursue more prospects?

Lesley Uren, chief executive of Korn Ferry Consulting, says client companies are wrestling with these questions and thinking about how compensation models should change.

"You have to really rethink what good looks like, what success looks like," she says. "How are we going to reward you now?"

Uren adds that one vision for compensation is to tie pay more closely to impact and outcome, and de-emphasize activity measures, like hours worked, that are frequently used as proxies for productivity.

That's a complicated endeavor because impact can be intangible. This underscores the conundrum of AI's effect on pay.

AI could make your value as an employee more subjective, even as it attempts to objectively determine the going rate for your services.