AI Can Reduce Bias in Pay Decisions — If It Doesn’t Embed Them

Total rewards practitioners are increasingly exploring artificial intelligence (AI) tools as a remedy for bias in compensation decisions. But, what if those tools amplify bias instead? 

Experts advise a two-prong response strategy:

• Make sure your house is in order before you bring AI in to scale pay determinations.
• Keep a ready eye on these tools once you implement them.

“AI doesn’t absolve us of judgment — it demands more of it,” said Marta Turba, the vice president of content strategy at WorldatWork. “Compensation professionals need to stay close to how these tools are built, what data they use and how outputs are applied. The goal is to make compensation more fair, not just more efficient — and that takes oversight, transparency and care.”

This article delves deeper into this subject.

Access bonus Workspan Daily Plus+ articles on this subject:

• Audit Checklist to Uncover and Mitigate AI Bias in Compensation
• AI Tools May Pave the Path to Greater Objectivity in Skills-Based Pay

How AI Tools Can Reduce Bias

According to the experts interviewed for this article, the uses for AI tools in compensation decisions, particularly in helping to address and reduce bias, are myriad and growing. These tools can:

• Automate routine or repetitive HR tasks.
• Gather and interpret large data sets, such as benchmarking metrics from multiple sources — especially useful for large or global organizations.
• Provide real-time compensation analysis and reviews of salary structures.
• Offer a bird’s-eye view of an overall compensation system, bringing patterns to the surface that might not have been noticed in isolation.
• Incorporate corporate-wide salary information and market data to better equip managers to make equitable, consistent compensation decisions.
• Urge managers and total rewards (TR) professionals to think through and justify their salary decisions or performance ratings — capturing that information, identifying patterns and codifying them to create more robust data and systems moving forward.
• Generate personalized talking points to help managers explain pay decisions to employees.
• Provide data to back up compensation decisions.
• Apply compensation rules the same way every time, helping avoid bias.
• Hunt down inconsistencies and outliers, compare roles across demographic groups and regularly conduct pay equity audits.

“AI can reduce bias by creating a single, consistent, inspectable and fair methodology for determining pay,” said Sara Hillenmeyer, the senior director of data science at Payscale, a compensation software and data provider. “This gives flawed and biased humans less wiggle room to make biased pay decisions. It’s much harder to fix the implicit bias in millions of decision-makers than it is to improve a single AI model or process that helps create guardrails and ensure consistent pay.”

(Hillenmeyer and co-panelists Ben Eubanks of Lighthouse Research & Advisory and Paulo Fava of ITX Corp. discussed this subject in a session at WorldatWork’s Total Rewards ’25 conference titled “Brace for Total Transformation: How AI is Reshaping Pay Decisions”.)

How AI Tools Can Exacerbate Bias

But these experts say the same tools that businesses are calling upon to eliminate pay bias can, if not utilized thoughtfully and strategically, actually magnify biases instead.

Turba advised employers to consider the following factors, which — if fed into AI models — can embed and augment prior or existing biases into future compensation decisions on a much larger scale:

• Incorrect or inconsistent historical pay decisions within your own organization (e.g., employees hired into the wrong role, promoted too late, paid outside of stated salary guidelines or demoted without a pay adjustment).
• Workers not being hired, promoted or paid equitably due to individual instances of intentional or unconscious manager bias.
• Long-perpetuated patterns of paying women or minority employees, for instance, less than their peers for similar roles .

“But it goes deeper than [internal compensation history],” Turba said. “Even if your current processes are solid, your employees come with pay histories shaped by previous employers and a long history of systemic inequities baked into society. Those legacy pay factors affect current salaries, expectations and how roles are benchmarked. When AI models pull in this data, they can unintentionally perpetuate those inequities — unless you intervene. AI doesn’t magically eliminate bias; it reflects the system it’s trained on.”

AI tools can help paint a better picture of issues such as pay compression, but they also can inflate them, said Nancy Romanyshyn, the senior director of total rewards strategy and solutions at HR technology company Syndio.

For instance, consider an employer that offers an incoming employee a salary deemed competitive with the market — but would not pay an internal employee a similar rate when promoting them to the same role because that salary leap falls outside the organization’s promotional increase caps.

“With that kind of disconnect, if you don’t correct the design, it’s only going to be exacerbated by new AI tools that are doing it faster and at a larger scale,” Romanyshyn said. “You still need to look at the outcomes. You still need to be critical of how those decisions are happening.”

There’s still significant work to be done in this area: a recent IBM study on global AI adoption found that 74% of surveyed organizations had not taken steps to reduce bias in AI tools they’d implemented, and 61% weren’t necessarily able to explain AI-driven decisions.

Tips for Getting it Right

To implement AI tools in a way that reduces rather than amplifies pay bias, Hillenmeyer said TR should partner with:

• IT and data teams to ensure integrity;
• Legal and compliance teams to help the organization remain compliant with pay equity and AI regulations; and,
• The HR team and teams focused on promoting an equitable and diverse workplace to help keep equity and representation at the forefront.

Build in checks and balances to monitor the outcomes of AI-driven pay recommendations — use the tools as a decision-supporter, not a decision-maker, Romanyshyn said. For instance, after implementing AI tools directed toward compensation practices, she said you’ll likely want to run pay equity analyses more frequently than before.

“If you scale compensation processes, you need to scale your monitoring at the same time,” Romanyshyn said.

It’s also vital to stay up-to-date on AI regulations, as well as pay transparency laws — so these tools remain an asset, not a liability, said Gordon Frost, a global rewards solution leader at consulting firm Mercer.

“For those who invest in upskilling their HR teams to leverage the new AI tools that are available, while also developing the appropriate governance and oversight practices, I think there is potential,” he said. “It will allow these teams to develop deeper insights into their existing pay practices, to understand where they may have instances of unfair pay decisions that need to be addressed.”

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:

• Equitable and Diverse Workplaces
• Employee Compensation
• Organizational Performance
• People Analytics