Artificial intelligence is influencing decisions about employee promotions, performance ratings, and pay, with research suggesting that HR departments are using automated analysis to challenge management assessments and identify previously overlooked talent.
A study published by HiBob found that 57% of surveyed HR professionals using AI to review workforce decisions had reconsidered an employee initially passed over for promotion.
Among respondents using AI to inform relevant decisions, 43% reported more favourable outcomes for pay increases, while 40% reported more favourable promotion decisions.
Almost three-quarters, 73%, said they had used AI-generated information to encourage a manager to reconsider a proposed decision. Within that group, 42% said they had identified occasions when an employee’s or candidate’s skills or potential appeared to have been overlooked.
The findings reflect the experiences of surveyed HR professionals rather than an independent assessment of whether AI produces fairer employment decisions. Nevertheless, they illustrate how the technology is becoming involved in matters that directly affect employees’ earnings, career progression, and job security.
Technology influences important decisions —
HiBob found that AI was informing a range of HR activities, including performance ratings, cited by 64% of respondents, promotions at 51%, and pay rises at 50%. Forty-two per cent reported using the technology in decisions concerning redundancies.
Other applications included workforce analysis, candidate screening, recruitment recommendations, and reviews of employee performance.
Nearly one-third, 31%, said they had used AI information to support changes to job responsibilities, organisational structures, or required skills.
These applications extend beyond administrative activities such as scheduling interviews or preparing routine correspondence. They also introduce questions about which information is included in an assessment and whether automated recommendations give appropriate weight to individual circumstances.
Ronni Zehavi, chief executive and co-founder of HiBob, said: “It gives HR a second opinion to test assumptions, challenge potential bias and spot overlooked talent.”
That potential depends on the accuracy and completeness of the information being analysed, alongside the ability of HR professionals and managers to understand and challenge recommendations.
Human oversight remains essential —
The research found that organisations were applying different checks to AI-assisted decisions. Fifty-nine per cent of respondents said AI information was checked for accuracy and relevance, while 45% reported reviewing recommendations for unfair or inconsistent treatment.
A further 41% said they considered individual circumstances that automated analysis might have missed.
These safeguards are particularly important when historical workforce data reflects earlier management decisions that may themselves have involved inconsistency or bias.
An automated system using such information may reproduce existing patterns rather than identify their weaknesses.
The Information Commissioner’s Office has emphasised the importance of meaningful human involvement in automated decisions that significantly affect individuals, alongside appropriate data protection safeguards and opportunities to challenge outcomes.
Human oversight must involve a genuine assessment of the evidence and an opportunity to reach a different conclusion, rather than routinely approving a system-generated recommendation.
Employers also need to consider access to sensitive employee information, the retention of workforce records, and whether individuals understand how their personal data contributes to decisions affecting them.
The HiBob findings indicate that automated analysis is already influencing employment decisions across the organisations represented in its research. The reported benefits need to be assessed alongside the quality of workforce information, the reasons recorded for individual decisions, and the accountability of those responsible for the final outcome.
