AI has promise to reshape payroll, but it still needs human oversight

"Payroll is a mission-critical, organization-wide effort that requires hours of repetitive, time-consuming tasks and navigating changing rules and regulations around local, state, and federal payroll and tax compliance," says Mariah Hantis, senior director of global people operations, payroll, and benefits at Deel in Philadelphia. "AI is already helping organizations like ours automate some of the manual, data-driven processes. By helping with administrative aspects, AI helps HR teams free up time for strategic decision-making and employee engagement."

That thinking was echoed in Gartner’s Hype Cycle for HR Technology from July 2024. Mary Baker, senior director of public relations at Gartner in Arlington, Va., says the report highlighted how “hyperautomation” in HR — an emerging technology fueled by AI — offers opportunities to improve efficiency and reliability across transactions and workflows prone to manual data entry errors and delays, such as payroll.

The report found that AI can optimize payroll by detecting input errors and catching anomalies in payroll calculations, overtime processing, and payment patterns.

However, while AI in payroll is considered highly feasible, the Gartner research also noted it offers low business value, making it low-risk but yielding marginal gains. This highlights the importance of balancing the adoption of AI with realistic expectations about its impact.

The human element in payroll processing

Hantis issues a cautionary note about the use of AI in both the payroll and HR realms.

"HR decisions are meant to be informed by AI, not made by it," she says. "Getting people paid requires nuanced judgment, empathy, and complex problem-solving. AI alone is not enough to handle such sensitive tasks without a 'human in the loop.'"

This sentiment underscores a key limitation of AI: its inability to replace human judgment in areas that require emotional intelligence and complex decision-making. While AI can handle data-driven tasks efficiently, final decisions—especially those impacting employees directly—still need human oversight. After all, a paycheque is not something that can be mishandled.

Baker points to another report from Gartner, 5 Total Rewards Predictions for 2024 and Beyond, which found organizations are experimenting with AI in transactional total rewards tasks, such as creating job descriptions, pay communications, and pay philosophy statements. As AI continues to advance, organizations will have the opportunity to expand their AI use cases within total rewards.

Enhancing compensation planning and pay equity

The potential uses of AI also extend to compensation planning and pay equity comparisons.

"I imagine AI can be used to speed up all jobs across payroll and pay management," Hantis says. "Deel AI, our AI tool, can be used by professionals to query not only the company’s global knowledge set but your own HR and people data."

Organizations could harness AI for immediate tasks, such as helping managers make compensation decisions, evaluating managers' compensation decisions for potential biases, or forecasting the impact of certain pay decisions on pay equity. Gartner anticipates that as a first step, organizations will likely explore vendors offering these capabilities and begin to identify the risks and ethics around using AI for sensitive total rewards components.

Privacy concerns and data security

With the integration of AI comes significant privacy risks.

"Privacy is a huge concern, especially in an area like HR," Hantis says. "You need to understand how and where any data particular to an organization can be shared by an AI tool. Lean on your IT and compliance teams for guidance."

Copying and pasting sensitive payroll data into AI tools like ChatGPT can pose risks if not managed properly. Organizations must ensure that data privacy regulations are adhered to and that any AI tools used comply with these standards. Understanding how AI platforms handle and store data is crucial to maintaining confidentiality and compliance.

Embracing AI amid uncertainty

For professionals who are fearful or unsure of AI, Hantis offers some practical advice: "It’s already here and likely being used by your HR technology stack or by individuals within your organization today. The best thing you can do for your organization is to get ahead of AI use and its implications on the business. A good idea is to come up with an internal policy on the use of AI."

Understanding AI's capabilities and limitations is crucial for workplace leaders. By proactively engaging with AI technologies, organizations can better control their implementation and mitigate potential risks. Developing internal policies can provide guidelines for responsible AI use and help address concerns related to ethics and compliance.

Starting points for AI integration

For those unsure where to begin with AI in the workplace, Hantis recommends exploring available tools and resources. "If you’re unsure of where to go to learn more about how to use AI at work. We’re biased, but definitely check out Deel AI as a potential place to start introducing AI into your HR workflows," she suggests.

It's important for organizations to evaluate AI tools carefully, considering factors such as functionality, security, compliance, and how well they integrate with existing systems. Collaboration with IT and compliance teams is essential to ensure a smooth and secure implementation.

"As we start to implement AI in day-to-day operations at a research or task-based level, it’ll be interesting to watch the use cases evolve," Hantis says. "For example, how can we use AI to forecast future payroll costs or avoid unnecessary mistakes?"

Baker points to Gartner research suggesting that as AI continues to advance, organizations will have the opportunity to expand their AI use cases within total rewards. However, human "checkers" who review AI outputs will be essential stakeholders as these experiments begin. This human oversight is critical to ensure that AI recommendations are accurate, fair, and free from biases.

Enter some text...