By Alvin Aloysius Goh & Eugene Chang

Singapore doesn't have an engagement problem. It has a judgement problem.

Building on the Singapore Workplace Report 2026: why the leaders being asked to fix engagement are the same ones deploying AI into decisions no one was formed to make.

Executive Summary

The Singapore Workplace Report 2026 establishes the problem beyond dispute: engagement is low, persistent and costly, and the manager sits at the centre of it. This paper is about what lies beneath that finding - and what it will take to change it.

Our argument is that Singapore's engagement problem is, at root, a judgment problem. For decades we have developed leaders at the level of skills: coaching, feedback, delegation, the techniques that can be taught in a workshop. But the work that actually determines whether someone leads well sits one level deeper, in judgment and identity: who a manager becomes, and the calls they make when no procedure can settle it. Skills can be trained. Judgment must be formed. A system that keeps investing in the first while assuming it produces the second will keep getting the engagement numbers it has.

Artificial intelligence makes this urgent, but not in the way most fear. The risk is not that machines will decide instead of leaders. It is that AI faithfully scales the judgment an organisation already has, with speed, confidence and the appearance of neutrality. Where judgment is sound, AI is a multiplier.

Where it is unexamined, AI automates the flaw.

And it does so just as it moves people off the execution work that skills were built for, and toward the design and discernment that only judgment, taste and human sense can supply. The competencies AI makes scarce are precisely the ones our development systems never prioritised.

Closing this gap will take more than better individuals. It requires forming judgment deliberately, and redesigning the organisations around managers so that good judgment becomes possible to exercise. The leaders worth having, in an AI-accelerated workplace, are not those who automate the most decisions. They are those who have examined the judgment AI is about to multiply.

There is a second layer beneath even that, and it is one Singapore has already lived through once. For a decade the country has funded the attempt to build better leaders with subsidised executive programmes, accredited modules, a national leadership-development initiative that has since concluded, without ever instrumenting whether any of it changed how a manager actually leads.

The signal Singapore cannot ignore

Singapore's engagement challenge is not new. But the SID-Gallup Singapore Workplace Report 2026 makes it harder to look away from. The report presents a genuine paradox: Singapore is one of the world's most prosperous and competitive economies, with per-capita GDP above USD107,000 at purchasing power parity. Yet only 14% of its workforce is engaged, below the Southeast Asian average of 25% and the global average of 20%. Engagement has sat at or near this level since 2019. Among workers under 35 it falls to 10%, against 16% for their older colleagues, with 53% reporting daily stress, compared with 37% of employees over 35.

The comparison sharpens further when set beside the region. Four ASEAN economies with materially lower per-capita GDP than Singapore all report higher engagement: the Philippines at 39%, Thailand at 34%, Indonesia at 27% and Malaysia at 25%. Singapore has not simply undershot a global benchmark; it has been outperformed by neighbours investing a fraction of what it spends on developing the people who run its workplaces. That is not evidence that Singapore invests too little. It is evidence that Singapore has never established what its investment actually buys, a question this paper returns to before it closes.

This should concern boards, chief executives and HR leaders alike. Engagement is not a soft metric. Gallup's global research links it directly to productivity and performance, and the Singapore findings were widely reported as an estimated productivity cost of some US$73.6 billion a year, around S$95 billion a year around S$95 billion a year.

When GDP growth is forecast to slow, a disengaged workforce stops being a cultural concern and becomes a strategic liability. Even Mr Dinesh Vasu Dash, Minister of State at the Ministry of Culture, Community and Youth and the Ministry of Manpower, has commented on the issue.

The report identifies four priorities: build manager capability, align culture with employee experience, reposition HR as strategic, and maximise talent density. These are sound. They give Singapore organisations a real place to start. But they are best read as the opening of a conversation, not its conclusion, because if the challenge has persisted through years of surveys, courses and culture programmes, the honest question is whether we have been solving at the right level at all. And it raises a harder one still: if 14% is where a decade of investment has left us, is that the floor — or simply where we have stopped looking?

Reading the number more carefully

Before asking what the 14% demands of us, it is worth asking whether the number tells the whole story. Leadership Choices has argued elsewhere that in Asian contexts, employee surveys can under-capture the truth. This is not a question of whether Gallup's methodology is sound - it is globally established and carefully benchmarked. It is a deeper, cultural point: in relationship-leaning, high-context workplaces, a survey is often not the medium through which people choose to tell the whole truth.

In many Asian organisations, truth is relational before it is transactional. Employees may not disclose what they really think simply because a survey link has arrived in their inbox, even when anonymity is promised. Concerns surface after trust has been built, in a one-to-one conversation, through a respected intermediary, in small groups over a meal, or in the more relaxed candour that sometimes emerges after work. The point is not the setting but what it represents: a shift from formal reporting to relational safety. People speak honestly when the medium feels human, trusted and socially appropriate.

This is why survey data can look cleaner than reality. Responses in Asia are shaped by acquiescence, politeness, midpoint and hierarchy effects: people say yes to avoid conflict, soften criticism to preserve harmony, choose the middle to stay safe, or hesitate to criticise a process seen as coming from the top. Benchmarks adjust for calibration between countries; they cannot fully resolve whether the response captured real sentiment in the first place. This does not make the survey invalid. It makes it incomplete.

The 14% figure may not capture the full depth or texture of the problem, which is all the more reason to read it as a starting point for deeper inquiry rather than the end of one.

The report's own paradox

The report contains two findings that are striking on their own. Placed side by side, they are more striking still.

The first is about human judgment.

Singapore organisations routinely promote their strongest individual contributors into management, then reward those managers for personal output rather than for developing their people. In the report's leadership survey, the senior leaders taking part rated their own organisations' manager effectiveness at just 3.32 out of five, and the strength of their leadership pipeline lower still, at 3.05. Gallup's broader research attributes some 70% of the variance in team engagement to the direct manager. Read plainly, this is a candid admission from the leaders themselves: the systems Singapore uses to choose and grow the people who make its most consequential human decisions are working only modestly well.

The second finding is about machines. Asked, in that same survey, how confident they were that their organisation could realise value from AI while managing its risks, those leaders returned a mean of 4.11 out of five, the highest-rated item of all, with 85% agreeing or strongly agreeing and not one rating it below the midpoint.¹

Hold those two findings together, remembering they come from the same people. On the human people-decisions they have made for decades, these leaders rate their organisations only modestly effective. On the machine-assisted decisions they are just beginning to make, they express near-uniform confidence. The question this raises is not an accusation, and it is one worth sitting with honestly: on what basis is our confidence in governing AI's risks higher than our confidence in the human judgments we already make ourselves? And if the leaders most engaged and most willing to be studied are this assured, the gap is unlikely to be smaller among those who did not volunteer to be studied.

This is what sits around the corner. Singapore is unusually well-placed to adopt AI fast, well- capitalised, digitally mature, culturally disposed to treat technology adoption as progress. Those are real strengths. But they also mean AI will enter people-decisions here faster, at greater scale, and against less resistance than almost anywhere else. When a capability that powerful is deployed that quickly into a judgment system that is, by leaders' own rating, only modestly sound, it does not repair the judgment underneath. It multiplies it, sound or unsound alike, at scale, wearing the reassuring appearance of neutrality.

What used to merely influence a manager's behaviour may soon quietly govern it.

Skills can be trained. Judgment must be formed.

By judgment, we mean the capacity to make consequential decisions well under uncertainty, where the numbers are incomplete, the rulebook runs out, and the right call depends on weighing evidence, context, consequence and fairness.

It shows up in strategic choices, in reading whether the data in front of you is telling the truth, and in decisions about people. It is precisely what cannot be reduced to a procedure or a metric.

The report's first recommendation, build manager capability, is correct. The risk is that organisations read it too narrowly and reach, once again, for skills. Singapore has not lacked leadership training. Many organisations have invested in agile leadership, situational leadership, coaching, cross-cultural competence and personalised development. Yet the engagement needle has barely moved. That should make us ask whether the issue is not the absence of training, but the limits of training when it is not connected to identity, judgment and system design. In our own work with clients, we are seeing leaders arrive at this conclusion themselves: they have run the courses and certified the competencies, and still find their managers unable to grow committed, capable people. The appetite for something deeper is real.

It is worth being precise about what "has not lacked leadership training" means, because the claim is stronger than it first appears.

Singapore has built one of the densest leadership-development ecosystems yet, without a published account of whether the pipelines it funded actually strengthened. A decade of leadership-training expenditure has been managed as an input, tracked by enrolment and completion, and never once connected to an outcome. This is not incidental to the argument about judgment; it is what makes the argument urgent rather than merely interesting. Singapore cannot yet say with confidence that the skills layer worked, let alone assume it produced the judgment underneath it.

There is a harder version of that admission still to make. Public funding for this ecosystem is not conditioned on outcome; it is conditioned on delivery. A programme provider is paid when a seat is filled and a course is completed, whether or not the manager who sat through it leads any differently a year later.

But it means Singapore has spent a decade unable to distinguish a subsidy that changed a manager's behaviour from one that changed nothing beyond an attendance sheet. It is a gap SHRI's own outcome-indexing work now sets out to close.

The reason is that the report's most uncomfortable finding is not about skill at all. It is about formation. We promote strong individual contributors and assume that performance in one role proves readiness for a very different one.

But management is not the next rung of the same ladder. It is a different craft. For the individual contributor, value is created through personal expertise, speed and output. For the manager, value is created through others: setting context, building trust, developing people, making trade-offs, and turning strategy into human commitment. That transition is easy to describe and hard to live. Many new managers look at the role from the outside and see only more meetings, more politics, more accountability for outcomes they cannot fully control. It is a tax on the work they love, not a path worth choosing.

Skills training does not resolve that. A manager can master feedback models, coaching questions and delegation techniques and still never make the deeper shift the role demands. They still define their worth by personal output, still avoid the hard conversation, still treat management as status or control. Capability without formation produces a more polished version of the same accidental manager. What has to change is not only what a manager can do, but who they understand themselves to be.

Judgment is also contextual. A people- decision that appears fair in one culture, team or organisational history may land very differently in another. A direct feedback model, a performance rubric, a promotion criterion or an AI-generated recommendation can look neutral while ignoring hierarchy, face, trust, market maturity or the relational norms through which work actually gets done. This is why leadership judgment cannot be reduced to a generic best-practice checklist, and why it is exactly the capacity most at risk when we begin to automate.

The layer the report does not reach

There is a dimension the report does not address, not through oversight, but because it lies outside the frame of a workplace survey. The report tells us that managers matter, that they are often poorly chosen, and that better systems must support them. It does not ask what managers must be equipped to judge.

This matters because most consequential people-decisions are not, at their core, technical. Whether a struggling employee needs support or exit; whether a complaint reflects a difficult person or a failing system; whether to trust a performance rating that reads as fair but was never checked against context, these cannot be trained as modules. They are acts of judgment: weighing fairness, consequence and human circumstance under uncertainty, with incomplete information and real stakes for a real person.

Judgment of this kind has quietly slipped out of the leadership-development conversation, not because it stopped mattering, but because it became inconvenient to teach. It resists the competency checklist, the dashboard and the two-day certificate. In a market that rewards what is legible, the slower work of forming judgment gets crowded out by what can be measured and sold.

Singapore adds a particular twist: we have tended to treat fairness and conduct as matters for the regulator to enforce rather than capacities leaders must personally hold. A strong tripartite framework and a robust compliance culture are genuine national strengths, but they can also let the commercial sphere quietly assume that judgment is handled elsewhere, that if a decision is lawful, it is sound. It is not the same thing.

AI removes that comfort. Used well, AI can strip away low-value administrative load and give managers more timely insight and more personalised support than they have ever had before. Used poorly, it will scale weak judgment with the confidence and consistency of a system. When a tool recommends a promotion, drafts a dismissal rationale, flags an underperformer or scores an engagement survey, it produces an output that looks authoritative and context-free. A leader trained in skills but not formed in judgment has no reliable way to tell when that output is right, when it is subtly wrong, and when it is confidently wrong in a way that will surface only later, at scale, in the very disengagement Singapore is trying to reverse.

There is a structural shift beneath this. For most of their careers, managers have created value in the middle of the work, taking direction, executing against it, and being judged on whether the output passed. That is where hard skills live, and it is exactly the layer AI is now absorbing. What it leaves behind are the two ends: deciding what should be done in the first place, and discerning whether what the machine produced is actually sound. Both ends run on judgment, discernment and human sense rather than technical execution. And both are precisely what a career spent rewarding execution never had to develop. AI does not only change the manager's tools. It moves the manager to the parts of the work where judgment was always going to matter most, and finds many unprepared for the shift.

Consider what this looks like in practice. A client-service team has posted its best numbers in years since adopting an AI tool, throughput up, response times down, the dashboard green, and leadership wants to hold the team up as the model for the division. The manager could take the win. But something nags: the experienced operator's discomfort with a result that looks better than the work beneath it feels.

He suspects the tool is scoring well by optimising what is easy to measure, and quietly thinning the slower work that actually holds clients. He cannot prove it yet. Doing the job properly means interrogating his own celebrated numbers before he banks them, digging into what the system is really rewarding, and insisting the measure serve the client and the business rather than the other way round. Nothing forces him to. No one is asking. The easier, entirely defensible path is to accept the applause and move on.

What separates the two paths is not conscience. It is whether his experience has taught him to distrust a number that looks too clean, and to walk into his director's office and say, against his own interest, that the results everyone is celebrating may be measuring the wrong thing. That is judgment. It is not a skill the AI tool can supply, and not one a certificate confers. It is formed, or it is absent, and in an AI-accelerated workplace, the difference between the two is the difference between a system that catches its own blind spots and one that scales them.

The future manager will not be valuable for outsourcing more decisions to AI. They will be valuable for knowing which decisions must never be fully outsourced, and why.

Two conditions: the will to lead, and a system that lets you

Before asking more of managers, it is worth seeing them clearly. Most managers are not failing for lack of care or intelligence. Many are carrying roles that have expanded faster than their preparation, authority and support, expected to coach, decide, motivate, transform and now govern AI-enabled work, often while still being measured by the output logic of the job they were promoted out of. The problem is rarely the person alone. It is the gap between what the role now demands and what the organisation has done to make it possible.

That is why formation cannot be only an individual project. Ending accidental management requires attending to both sides of the managerial transition at once: the intrinsic and the extrinsic. The intrinsic question is whether a person can come to see leadership as a meaningful path of contribution rather than a burden or a prize. The intrinsic question is whether they choose, internally, to be formed. The extrinsic question is whether the organisation creates the time, authority, incentives and cultural permission for good leadership to be practised at all.

Both must hold. A person may want to lead and be crushed by a system that rewards the wrong things. A system may offer every support while the person never makes the inner shift. This is where the report's own instinct, that structure matters more than exhortation, becomes decisive, and where its evidence is strongest. Its cases pair capability- building with structural change: reducing team sizes, widening the manager's remit for development, treating wellbeing as a matter of design rather than individual resilience. That is the extrinsic half done well, and the report deserves credit for naming it.

The warning is that delayering, pursued for agility and cost, often removes management roles without removing management work. Coordination, coaching, sensemaking and conflict resolution do not disappear; they land on fewer, more tired shoulders. Gallup's global data shows manager engagement falling nine points since 2022, to 22% in 2025, and concludes that this decline accounts for most of the recent global slide. The middle layer that once held team morale and quietly mentored the next generation is itself being thinned out, precisely when AI is being handed to it as relief. But a depleted manager with more dashboards is still a depleted manager. AI creates capacity for the human work of leadership only when it is deployed by someone with the judgment to use it well. Absent that, it is simply one more thing to manage.

The instrument we still need

There is a further reason this cannot be left to individual conviction alone. Even where formation succeeds, even where a manager genuinely makes the inner shift from output to people, from control to trust, Singapore currently has no reliable way of knowing it happened. The instruments in place measure attendance, satisfaction and completion. None of them measure whether a manager, twelve months on, actually leads differently. That is not a hypothetical gap. It is the precise reason a national leadership-development initiative could conclude without anyone being able to say whether it changed how Singapore's managers manage.

The absence is a choice, not a technical limitation. Internationally, the measurement Singapore lacks already exists. The World Management Survey, built over two decades by researchers at the London School of Economics, Stanford, Harvard and MIT, has assessed management practice quality across more than 35,000 organisations in 35-plus countries, established a causal link between management quality and productivity, and directly informed industrial policy in the United Kingdom, Australia, France and New Zealand. If a single academic research programme can build a credible, internationally comparable measure of how well people are actually managed, its absence from Singapore's own policy architecture is not because the problem is unmeasurable. It is because nobody has yet been required to measure it.

This is where SHRI's own institution-building work becomes directly relevant to the argument this paper makes. Through the Singapore Reskilling Outcome Index, SHRI has already begun shifting one corner of national workforce policy from activity metrics, namely completions, enrolments, credits used, to impact metrics: employment transition, wage trajectory, role quality, sustained employability. The judgment gap this paper describes is the same failure, surfacing in a second domain. A leadership-formation agenda that is funded, delivered and never verified will produce exactly what the last decade produced: another Gallup number, three years from now, confirming what nobody checked in real time, the invoice for a decade nobody audited.

Closing that gap does not require new legislation or a new bureaucracy. It requires boards to place manager effectiveness on the governance agenda as a tracked metric rather than an HR update buried in an engagement dashboard; organisations already using publicly subsidised leadership training to attach voluntary six- and twelve-month outcome tracking to that spend, ahead of any government requirement to do so; and a Singapore Management Capability baseline, the kind SHRI, working with academic partners such as NUS or SMU, is positioned to pilot, adapting the World Management Survey precedent to the specifics of tripartite bargaining, family-owned enterprise concentration and SME density that define the local context. Singapore's own Fractional Leadership Integration Programme (FLIP) and Senior PMET Transition Bridge point to a further move: the experienced leaders now exiting the workforce through the super-aged transition are a lower-cost, higher-fidelity source of coaching capacity for newly promoted managers than another subsidised course, converting a demographic pressure into exactly the capacity an under-prepared management layer needs.

None of this replaces formation. It disciplines it, making sure that when Singapore finally does the harder work of building judgment rather than skills, it does not repeat the one mistake that made the judgment gap invisible for a decade in the first place: investing in good faith without ever finding out whether the investment worked.

The work ahead

Singapore's engagement challenge will not be closed by another survey cycle or another catalogue of courses. The SID-Gallup findings are a signal, and a valuable one. They point past themselves to a deeper question about how organisations select, form, support and sustain the people closest to the employee experience.

There is something genuinely freeing in that. If the problem were simply that Singapore's people do not care, it would be nearly intractable. But that is not what the evidence shows. It shows capable, committed people working inside systems that were never designed to form the judgement good leadership requires, and that is a far more solvable problem. It means the years of flat engagement scores are not a verdict on Singapore's workforce. They are a sign that we have been working hard at the wrong level.

For too long, management has been treated as the reward for strong individual performance, a layer of administration, or an escalation point when work gets hard. Seen differently, it is a distinct discipline that can be learned. One that rests on the identity to see leadership as a path worth choosing, the judgment to make consequential human decisions well, and the organisational conditions under which both become possible to sustain. None of these is beyond reach. Each can be built deliberately, once we decide to.

This is where SHRI's leadership work is now focused: helping managers understand the human work of leadership, equipping them with practical craft, strengthening the judgment responsible AI use demands, and helping organisations redesign the systems that make good management viable. And, alongside it, building the means to verify that any of this changes how Singapore is led, rather than simply assuming it does.

Singapore does not need more accidental managers with better tools. It needs formed leaders who can build trust, grow capability, exercise judgment, and decide, in an AI-accelerated workplace, which human decisions must remain human. That is the work ahead, and it is work within our reach.

---

/

¹ The manager-effectiveness, leadership-pipeline and AI-confidence ratings all come from the same SID-Gallup leadership survey, in which a small number of senior leaders (n=14) rated a set of workplace statements. This is a directional signal from a self-selected group, not a population estimate — but because the ratings share one sample, the contrast between them is a like-for-like one: the same leaders who rate their people-systems only modestly express near-uniform confidence about AI. The 14% engagement figure and the manager's share of engagement variance are drawn separately from Gallup's large-sample instruments.