When AI recommends, who stays accountable?

Ask most people what governance means, and they will point to boards, risk and compliance. Rita Tsui, Founder of AsiaHRM and a member of the Hong Kong Chartered Governance Institute, starts with something less obvious: the everyday decisions that determine who gets hired, who gets promoted, whose performance concerns are addressed and whose are quietly overlooked.

Those decisions, she tells HRM Asia, shape organisational culture, trust and ultimately performance – and they are governance decisions, whether or not they are labelled as such. “People governance,” as she defines it, concerns how organisations exercise judgement, accountability and oversight in decisions that affect people. It looks beyond whether a policy exists and asks whether decisions are actually made fairly, consistently and in line with an organisation’s stated values.

To draw that distinction clearly, Tsui separates three related layers. HR administration is largely operational – the proper running of employment processes, records, payroll and people systems. HR governance sits above that, focused on the frameworks, controls, and oversight that govern the HR function and its associated risks. People governance takes the conversation further still, into how decisions are actually made across the organisation – many of them not by HR at all, but by line managers and senior leaders who exercise a significant degree of judgement and discretion.

That, she says, is where the gap tends to open.

When the policy is the same but the experience isn’t

An organisation may have excellent policies, sophisticated systems and a clearly articulated culture, and still deliver markedly different experiences to employees depending on who they report to. One manager applies performance standards consistently and explains difficult decisions openly; another leans on personal relationships or informal judgement. The policy is identical. The employee experience is not.

For Tsui, this is the heart of the matter. Governance, she suggests, is not embedded only in structures and policies but is expressed through behaviour – and the real test comes when a situation is not straightforward. When pressure mounts to deliver results, when a senior leader seeks an exception, or when a high performer behaves badly, informal norms can quickly outweigh formal policy.

Employees read these moments clearly. They do not judge an organisation’s values by what appears on a website or in a handbook, she notes, but by what they observe: whose behaviour is tolerated, how consistently rules are applied, and whether leaders themselves are willing to be held accountable. Culture, in her account, shapes what people come to believe is acceptable. A culture that rewards results at all costs invites workarounds; leaders who routinely make unexplained exceptions teach employees that process does not apply equally to everyone. Where leaders explain their decisions, welcome challenges and apply principles consistently – even when the outcome is unpopular – trust tends to hold.

What AI changes, and what it doesn’t

If the policy-behaviour gap is an old problem, Tsui argues that AI raises the stakes. Technology can now influence people decisions at far greater scale and speed – supporting recruitment, surfacing patterns in performance data, recommending candidates for promotion, and flagging employees deemed at risk of leaving. These tools can be genuinely valuable. What they do not do, she cautions, is remove the need for human judgement or accountability.

One risk is that organisations begin to treat a system’s recommendation as objective simply because it is data-driven. But an algorithm is not independent of the organisation that built or trained it. The data it draws on may reflect historical patterns, existing biases or past management decisions; where those were flawed, AI can reproduce or even scale them.

The second issue is accountability. When AI shapes a people decision, Tsui argues, organisations need to be clear about who remains accountable, because accountability cannot simply be handed to the system. Leaders should still be able to explain AI’s role, the information used, whether the recommendation was challenged, who held final authority, and how the decision could be reviewed. The stakes are highest in hiring, performance and promotion, where the outcomes bear directly on people’s careers.

The governance risk she believes organisations most underestimate is what she calls “accountability drift” – the tendency, as systems grow more sophisticated, to assume a recommendation is neutral and therefore leave it unchallenged. In her view, AI should sharpen decision-making, not replace responsible judgement. That points organisations toward governance mechanisms designed for using AI in people decisions: clear decision rights, appropriate oversight, transparency, mechanisms for challenge, and leaders equipped to understand both what the technology can do and where it falls short.

The path is not more policy

For leaders who recognise the gap, Tsui is clear that the starting point is not to write more policies. Most organisations already have plenty. The more useful question is whether those arrangements are working together to support good decisions and build trust.

She frames people governance maturity across several connected dimensions. Governance structure concerns clarity about decision rights, oversight and escalation. Accountability depends on whether leaders understand that they answer not only for business results but also for how people decisions are made. Culture and leadership behaviour determine whether the organisation encourages consistency, transparency, constructive challenge and ethical judgement. People practices cover whether recruitment, performance management and promotion are designed and applied fairly. Capability asks whether managers can handle discretion responsibly, recognise bias and, increasingly, understand the implications of AI. And wellbeing, she adds, belongs in the picture too, since inconsistent treatment and unclear expectations eventually register in engagement and trust.

READ MORE: Compliance asks if it was done. Governance asks if you can prove it

Within this thinking, Tsui has developed a self-assessment she calls the People Governance Maturity Diagnostic – a discussion tool intended to help organisations reflect on how governance is practised across these areas, rather than on whether the right policies happen to exist. The point, she says, is not to generate a score or add another checklist, but to help boards, chief executives and HR leaders identify their strengths and gaps and open a more useful conversation about where stronger governance capability would matter most.

She is reluctant to identify a single universal weak point, because organisations differ. But in her experience, the gap is rarely the absence of policy. It lies in translating governance principles into leadership behaviour and day-to-day decisions. A maturity approach, she suggests, is useful precisely because it shifts the question from “Do we have a policy?” to how consistently those principles are embedded – where inconsistencies occur, where discretion goes unchallenged, whether employees understand how significant decisions are made, and whether leaders can justify them.

The aim, she stresses, is not to pile on controls for their own sake. It is to improve decision quality, strengthen accountability and build trust. In her formulation, strong people governance shows up when governance becomes part of how leaders lead, rather than something leaders reach for only when a compliance issue arises.

That, ultimately, is her reframing of the AI moment. The technology is forcing organisations to rethink how decisions are made and who is accountable for them, but the underlying issue is not new. What AI changes is the speed and sophistication of people decisions; what it does not change is the need to make them responsibly, fairly and with judgement. The environment she describes is one in which structures, processes, technology and leadership behaviour work together – so that people do not simply see governance as a set of rules, but experience it through the way decisions are made.

Rita Tsui hosts AsiaHRM’s LinkedIn Live Governance Series, which airs every Wednesday and is supported by HRM Asia, exploring governance, leadership and people practices.