What Role Should (or Shouldn’t) AI Play in Payroll?

Nearly half of American workers are uncomfortable with their employers using artificial intelligence (AI) in HR-driven payroll processes, according to a recent survey.

Twenty-two percent of the nearly 10,000 workers surveyed by PayrollOrg, a payroll education and training company, said they were “very uncomfortable” with that AI use case, while 24% said they were “somewhat uncomfortable.” In comparison, only 25% said they were “somewhat” or “very comfortable” with AI being used in payroll.

The findings echo PayrollOrg’s 2024 survey, when 39% of respondents said they were uncomfortable with AI calculating their pay and 53% were uncomfortable with AI tools answering payroll questions instead of a person.

Meanwhile, 25% of employers surveyed by PayrollOrg said they have moved meaningful AI experiments into production, but only 11% of their payroll teams are using AI features in their technology.

“The HR and Rewards space is warming to AI, but adoption is still in its early stages,” said Robert Sheen, the CEO and founder of Trusaic, a pay equity and regulatory compliance software company.

WorldatWork members can access a bonus article on this subject:

Laying the Foundation

Forthcoming research from Trusaic’s EU Pay Transparency Readiness survey found that 52% of organizations are evaluating AI for compensation and rewards, while 27% are already using it extensively or in limited areas.

“Compensation is still transitioning away from spreadsheets and rigid annual cycles, and AI represents a real shift here: It can take over the grunt work and surface real-time insights that help comp and rewards practitioners make better-informed calls,” Sheen said.

He stated that when done well, AI can move pay evaluation from a reactive, once-a-year exercise to a proactive, ongoing program.

“But that potential only holds if it’s built on a strong compensation philosophy grounded in pay equity,” Sheen said. “Without that foundation, every decision AI informs is flawed from the start. That’s the real risk.”

Current Uses and Opportunities

Today, AI is mostly improving the machinery behind payroll: automating repetitive work, identifying anomalies, supporting employee inquiries and helping teams manage exceptions more efficiently, said Tom McMullen, a senior client partner at consulting firm Korn Ferry.

“For employees, the best AI may actually be invisible,” he said. “They don’t want AI-powered payroll — they want payroll they never have to think about.”

Tauseef Rahman, the U.S. workforce rewards solutions leader at consulting firm Marsh, added that most AI implementation in payroll operations is currently upstream of the actual pay decisions.

Among organizations using AI-enabled compensation capabilities, Marsh research shows:

55% apply it to job matching or leveling;
38% to enable market-pricing recommendations;
13% to pay equity analysis;
12% toward manager decision-support tools;
11% to inform salary increase recommendations; and,
9% for budget optimization.

The experts interviewed for this article stated the bigger opportunity is moving from processing transactions to anticipating problems, surfacing unusual patterns, and giving employees and managers better information about pay. But according to McMullen, human judgment, explainability, governance and auditability become much more important when AI shifts from administering pay to influencing pay decisions.

Most currently used AI solutions involve large language models (LLM) that ingest and parse large amounts of external data, explained Jason Adwin, the national compensation and career strategies practice leader at consulting firm Segal.

LLMs can produce mathematical-based outputs, such as how to split salary increases across a team given specific budgets and performance distributions. In this way, technology can boost efficiency and productivity — but the decisions are left to humans, Adwin said.

“When you’re making decisions about pay, it’s inherently a very personal decision to employees,” he continued. “And when you’re having a conversation with an employee about how much you’re paying them, a manager needs to take accountability.”

What the AI Future Holds

Adwin can imagine a potential future where agentic AI accounts for an individual manager’s track record to make suggestions and recommendations around pay, but he predicted that use case is at least a decade off and still wouldn’t replace having human eyes and judgment on pay decisions.

“The weekend an employee came in and saved your client — some of those things just can’t be quantified in a language model,” he said. “A person is going to need to make that decision. I say that because I don’t think I’d want someone relying solely on AI to determine my pay.”

The potential isn’t simply faster salary recommendations, Marsh’s Rahman said. The opportunity is more coherent compensation management, where job matching, market data, pay equity, budget constraints and manager decisions can be considered together rather than in separate spreadsheets, systems and approval chains.

According to Rahman, the strongest near-term use case is decision support: helping compensation teams identify anomalies, model tradeoffs, test budget scenarios and surface equity risks before a manager makes a decision.

Marsh’s latest U.S. Compensation Planning Survey suggests the market sees this next step. Organizations are “seriously considering” AI for:

Market-pricing recommendations (26%);
Pay equity analysis (25%);
Salary increase recommendations (22%);
Job matching (22%);
Budget optimization (22%); and,
Manager decision support (22%).

“For rewards leaders, the question is not whether an algorithm can produce a number — it can,” Rahman said. “The question is whether we have built the job architecture, market-pricing logic and governance backbone that makes that number logical and communicable.”

Editor’s Note: Additional Content

For more information and resources related to this article, see the pages below, which offer quick access to all WorldatWork content on these topics: