Zhao Yang Ng, principal at Baker McKenzie Wong & Leow, shares how HR leaders can ensure accountability and fairness when performance decisions are shaped by AI
As artificial intelligence systems become increasingly embedded in performance evaluations, appraisals, and workforce planning, employers in Singapore are facing a new kind of responsibility.
When employee terminations are influenced directly or indirectly by AI tools, the potential for missteps grows significantly.
Yet while the law continues to catch up with technology, the obligation for HR to uphold fairness, transparency, and sound judgment has never been clearer.
To help employers act decisively without exposing themselves to legal or reputational harm, HRD Asia spoke with Zhao Yang Ng, principal in the employment practice group at Baker McKenzie Wong & Leow, about what HR must have in place when using AI-powered tools in employee decision-making.
Applying tripartite standards to AI-linked dismissals
AI tools can enhance consistency in decision-making, but they do not replace the legal requirement to ensure fairness.
Ng emphasises that employment terminations must always be rooted in defensible, merit-based grounds.
“Employers must ensure that any dismissal decision, even if influenced by AI, is not based on unsubstantiated misconduct, unverified performance issues, or discriminatory grounds,” he says.
This principle remains consistent with Singapore’s Tripartite Guidelines on Fair Employment Practices and the upcoming Workplace Fairness Act 2025 (WFA), which will explicitly prohibit discrimination based on 11 protected characteristics.
While no AI-specific guidelines currently exist, Ng stresses that employers must still be transparent about how performance is assessed and how the tools operate.
“Employers should explain how the performance is assessed, what kind of data the AI model was trained on and what measures are in place to prevent biased outcomes.”
Equally important, he says, is preserving human agency.
"At least one employee should be involved in the final decision-making process in dismissals. This allows for at least one stakeholder within the firm to substantiate and justify the decision to dismiss the employee," Ng explains.
"Having this human oversight helps protect the company against wrongful dismissal and discrimination claims, especially when employees are also given the opportunity to present their side of the story before the dismissal decision is made," he adds.
If AI flags ‘poor performance,’ is that enough?
Terminating employment with notice in Singapore does not legally require a reason.
But Ng cautions that disclosing reliance on AI, especially if the employer cannot explain the system’s output, can significantly increase the risk of a wrongful dismissal claim.
"If an employer discloses that the reason for dismissal was due to an AI tool flagging the employee for poor performance, this can expose the company to a wrongful dismissal claim," he says.
In such a situation, the burden falls on the employer to show that the termination was grounded in job-relevant criteria.
"The employee may argue that the dismissal was based on unsubstantiated poor performance, and the burden would then fall on the employer to justify the grounds for termination."
Ng emphasises that HR must not treat the system as a black box.
"To mitigate this risk, employers using AI in dismissal decisions must be prepared to explain how the AI had influenced the dismissal decision, including how the tool collects and processes data to assess the employees' performance."
He also reminds employers that an AI tool’s logic must be understandable and explainable.
"Although there is no statutory definition or standard for poor performance metrics under Singapore law, companies must ensure that the indicators are relevant to the employee’s role. It is also essential that HR teams understand the inputs and logic behind the AI’s assessment, so they can provide a clear and fair explanation if challenged."
Accountability for AI vendor tools
The growing availability of AI-based HR products has led many companies to outsource appraisal and selection tools. But Ng is firm.
"Employers are legally responsible for HR decisions made using third-party AI tools.”
This includes understanding how the AI operates and ensuring the outcomes are not discriminatory.
“Algorithms may unintentionally discriminate based on facial features, accents, or other personal characteristics, leading to biased outcomes and potential legal claims under the new WFA.”
Ng urges employers to demand full visibility into the vendor’s system.
"Employers must ensure they have full visibility and control over how AI tools are used in hiring, appraisals, and terminations. This includes understanding how decisions are made, what data is being processed, and what safeguards are in place to prevent unfair or biased outcomes."
AI monitoring: obligation to inform staff
From a legal standpoint, Ng explains that Singapore’s Personal Data Protection Act (PDPA) includes an evaluative purpose exception.
This allows employers to use personal data to assess staff without consent, particularly for performance monitoring.
But that doesn’t eliminate the obligation to inform.
“Employers are still obligated to notify employees regarding the use of AI systems in monitoring performance,” he says.
In practice, Ng recommends disclosing the tool’s purpose and scope, particularly when it may influence employment decisions.
“Under the PDPA Advisory Guidelines Chapter 6, the evaluative purpose exception allows employers to collect, use and disclose personal data without needing the consent of the employee. However, employers are still obligated to notify employees regarding the use of AI systems in monitoring performance.”
He also advises updating internal policies to reflect how data is collected, what it’s used for, and what tools may be deployed in performance appraisals.
“Employers should proactively implement an internal AI policy and a HITL decision-making process to stay compliant with the current Tripartite Guidelines and reduce [the] risk of claims by leaving employees.”
Human-In-The-Loop (HITL) processes
Superficial review of AI outputs will not hold up in sensitive decisions. Ng says meaningful Human-In-The-Loop (HITL) processes must be the norm.
“While there is no hard and fast rule when it comes to determining the appropriate level of intervention, the key principle is that the reviewer must be able to understand how the AI arrived at its decision, and the individual must have the authority to override it if necessary," he says.
"The review process should not be a mere formality or rubber-stamping exercise; it must serve as a meaningful check to ensure fairness and accountability.”
He recommends a structured approach. Define who reviews the decision, at what stage, and what powers they have. Employers should also establish internal AI policies that specify:
- Which HR decisions require human oversight
- How outputs are validated
- Where issues can be reported internally
Especially in cases of termination or redundancy, human validation is more than a compliance step. It is a line of defense.
Appeal mechanisms in AI-backed HR systems
As the WFA moves closer to enforcement, Ng underscores the importance of building internal appeal channels, especially for decisions where AI played a role.
“With the WFA coming into force in 2026/2027, it is especially important for companies to have clear procedures that allow employees to challenge or appeal decisions influenced by AI, whether those decisions happen before hiring or after employment ends.”
Ng recommends formal procedures for appeals, ideally routed through a separate HR contact or review panel. What matters most, he stresses, is that employees are heard and that their concerns are addressed before they escalate externally.
What should HR implement now?
The WFA will come into force by 2026 or 2027, but Ng stresses that HR should act now, not later.
This means implementing an internal AI policy, a robust HITL review process, and clear appeal mechanisms that apply across the employment lifecycle, from hiring to exit.
Most importantly, HR teams must understand the tools they rely on.
"For these procedures to work effectively, companies must also understand how the AI tool or vendor uses data to reach its conclusion. This includes knowing what data is being collected, how it is processed, and how it contributes to the final output."
AI may offer faster assessments and cleaner data, but without accountability, it adds more risk than reward.
"Without this understanding, it is difficult to justify decisions or defend against potential discrimination claims. Transparency and control over AI systems are key to ensuring fair outcomes and protecting the company from liability."
When performance scores shape life-changing decisions, HR leaders must take every step to ensure those processes are fair, transparent, and defensible.
