Cutting Costs or Hobbling Growth? A Business-Side View on AI and Headcount
Singapore's leaders are urging companies not to use AI to cut headcount. The advice is thoughtful and partly right. But business leaders also deserve a clear account of the risks and constraints they carry, so they can make their own informed decisions.
The question on the table
On 28 September 2026, Senior Minister of State for Digital Development and Information Tan Kiat How told business leaders that companies using AI mainly to reduce manpower are "hobbling" their future growth (CNA). His argument, in brief:
As AI becomes cheaper and more widespread, intelligence itself stops being scarce.
Economic value sits in solving "bottlenecks" — customer, domain and industry pain points.
Experienced staff hold the context, judgement and relationships needed to find those bottlenecks.
Saving "one or two heads here and there" is a short-term gain. The longer-term game is to use AI to make people more capable, and to "bring your people along".
His remarks echo a wider official position. In Parliament's May 2026 debate on an AI transition with "no jobless growth", the government said it expects productivity gains to be shared with workers through fairer wages (MOS Jasmin Lau, 6 May 2026).
The backdrop explains why the message matters now:
Growth is up. GDP grew 5.9% year-on-year in Q2 2026, and MTI raised its full-year forecast to 4.5–5.5%, citing the surge in global AI-related investment (MTI, 11 Aug 2026).
Retrenchments are up too. 4,620 workers were retrenched in Q2 2026, the highest quarterly figure since 2020. Around 3,950 of them were PMETs (MOM data, via The Online Citizen, 21 Sep 2026).
Re-employment is harder. Only 54.9% of retrenched residents found work within six months, down from 60.7% (same source).
The economy grew in every quarter shown, yet retrenchments rose in each of the last four quarters. That gap is what the debate is about.
The concern for workers is legitimate, and much of the advice makes sense. The open question is whether it fits every business. This article looks at the same facts from the employer's side of the table.
Where the advice is right — and where it gets harder
Much of the argument is sound business sense. Customer relationships built over years rarely move to "a click and a prompt and a screen". Domain knowledge is hard to rebuild once it has walked out of the door. At FYT, we put it this way: AI speeds up tasks; people interpret and decide.
The tension lies in applying the advice to every firm:
It assumes growth. Redeploying people to "jump higher" works when demand is growing. In a fixed-volume function such as payroll or regional finance, faster work does not create more work to do.
Not every role holds deep context. The Minister's case is strongest for experienced, customer-facing staff. It is weaker for digitised, repeatable roles — the very roles AI is absorbing first.
Short-term gains can mean survival. For a firm facing higher energy prices and thin margins, saving costs this year may be what funds next year's growth.
He agrees it is early. The Minister himself described AI as being at a "nascent stage", with Singapore in the "adoption phase". That supports caution about fixed commitments while the picture is still forming.
The real disagreement is not about whether people matter. It is about who decides the right mix of people and technology for a given business — and who bears the cost if the decision is wrong.
1. Productivity was the point — and it was forecast
None of this should come as a surprise. For years, policymakers, economists and industry have said that generative AI can perform many of the digital, repeatable tasks people once did. Singapore, one of the most digitised economies in the world, was always going to feel this early.
Productivity means more output per worker. A business can use that gain in two ways:
Grow output with the same team — possible when demand is expanding.
Hold output steady with fewer people — likely when demand is fixed, as in many back-office and shared-service functions.
Where the volume of work is fixed, some reduction in headcount for the same size of business is an expected outcome, not a sign of bad faith.
Workers were also warned early. On 2 May 2024, Manpower Minister Dr Tan See Leng told an international audience at the St Gallen Symposium that people who have embraced AI and technology will displace those who have not. He linked this to Singapore's $1 billion, five-year investment in its national AI strategy to build the industry and equip people with the skills needed (The Straits Times, 4 May 2024). What we are seeing is, in part, that warning playing out.
His 2026 Committee of Supply speech set out the policy aim of steering AI adoption to "enhance our workers' potential, not displace or replace it" (MOM, 3 Mar 2026). Both statements can hold at once: displacement is a risk, and avoiding it is a goal, not a guarantee.
A fair caveat. Official data suggests AI is not yet the main driver of job cuts. Of firms that adopted AI, only about 6% reported reducing headcount because of it (Minister Tan See Leng in Parliament, 7 May 2026, via People Matters). Most Q2 retrenchments were attributed to "business reorganisation or restructuring" (72%), a category that can include AI-enabled redesign but is not limited to it. Business leaders should hold both facts at once: the headcount effect is real and expected, and it is still early.
[embed: node/62e718d4-2aa4]
Restructuring can include AI-enabled redesign, but MOM's categories do not identify AI separately. The data cannot yet say how much of the rise is due to AI.
2. Who owns the gains? Risk, capital and the employment contract
A business is a bundle of risks that someone chose to take on. Founders and investors put up capital, organise resources, pursue opportunities, and absorb losses when things go wrong. Adopting AI is one more such bet, with its own costs:
licences, integration and data work;
legal and governance exposure while the rules are still forming;
reskilling existing staff;
the chance that the investment simply does not pay off.
Employment exists because a business needs people to perform a defined set of tasks, usually over defined hours. The contract sets out what each side owes. Rewarding people who drive business outcomes — through raises, bonuses and promotion — is already in a company's own interest. For everyone else, the assurance is that the company pays what was agreed.
Seen this way, official calls to keep headcount and "share the gains" raise an interesting tension. It treats productivity gains as a shared pool. Many owners see them as a return on risks they alone underwrote. It is difficult to argue that AI-driven gains come entirely from employees' efforts, even if employees clearly contribute.
There is also a governance point. Many firms in Singapore answer to overseas headquarters, boards and shareholders — some of them government-linked. How gains are allocated is set by those governance structures, not by public appeal alone.
3. We are still storming, not norming
In team-development terms, most organisations are still in the "storming" stage with AI: building, testing and rebuilding workflows. Stable norms for how work, pay and roles should look are some way off. Even official observers note that AI "hasn't gone deep enough" into business processes yet (Acting Manpower Minister Jasmin Lau, 22 Aug 2026, via The Independent). Committing to a permanent split of gains that have not yet settled is a hard ask.
Singapore's economic model adds a particular exposure. The country has deliberately built itself as a regional hub for corporate functions — HR, finance, shared services and regional headquarters. These are exactly the digitised, process-based roles where generative AI is already proving capable. The success of the hub strategy also makes this segment of the workforce more exposed. Early signs are visible:
Information & communications and financial services were among the sectors with the largest rise in Q2 retrenchments.
490 workers were retrenched due to overseas relocation, and 420 due to high costs (MOM data, via The Online Citizen).
PMET job vacancies fell, with total vacancies down to 68,600 from 73,300 in March (same source).
Meanwhile, the operating environment remains volatile:
Energy shock. The Middle East conflict disrupted the Strait of Hormuz. Brent crude roughly doubled to about US$141 a barrel, and electricity tariffs rose to 27 cents/kWh for Q2 2026, with further rises expected (DPM Gan Kim Yong, Ministerial Statement, 7 Apr 2026).
Inflation. Official inflation forecasts were revised upward (same source).
Concentration risk. MTI itself warns that a slowdown in global AI capital spending could weigh on the very sectors driving today's growth (MTI, 9 Sep 2026).
In other words, headline GDP growth is concentrated in AI-linked sectors. It does not mean every business is enjoying windfall profits. Many are still absorbing higher costs while funding a transformation whose payoff is uncertain.
4. The asymmetry of flexibility
From a CFO's desk, AI and people are very different kinds of commitment. One can be adjusted in weeks; the other is, rightly, protected by process and scrutiny.
AI capability | Human capital | |
How cost is set | Subscriptions, seats or token budgets | Salaries, benefits, CPF contributions |
Scaling down | Reduce seats or budget, often within a billing cycle | Retrenchment process, notice and benefits |
Oversight | Contract with the vendor | MOM notification within 5 working days for firms with 10+ staff (MOM); tripartite guidelines; union engagement where relevant |
Once raised, can it fall? | Yes | Rarely — pay rises are hard to reverse |
Reputational risk | Low | High — disputes can play out in person and online |
New obligations | Emerging AI governance rules | Workplace Fairness Act, expected in force by end-2027 (Hawksford) |
These protections exist for good reasons, and many employers support them. But they shape behaviour. When a gain is uncertain and the obligation is permanent, the rational choice is to be cautious about locking gains into fixed pay. A one-off bonus, profit-sharing tied to results, or funded training are far easier to commit to than a permanent wage uplift.
5. The unintended outcome: fewer jobs, not fairer ones
The government and employers want the same end result: good jobs for Singaporeans. The risk is that well-meant pressure produces the opposite.
Singapore still leads the region as a base for business. In AmCham's 2026 survey, 58.8% of respondents preferred Singapore for an Asia-Pacific regional headquarters, far ahead of Hong Kong at 17.6% (Asia Business Daily, 15 Apr 2026). The same survey is also a warning: South Korea's ranking fell, and respondents' top concern was "labor policy and labor market flexibility" (71%).
The lesson is not that Singapore is losing its edge. It is that labour flexibility is one of the things investors weigh most closely. The line between reasonable protection and lost competitiveness can be crossed gradually.
If employers come to feel that:
productivity gains will attract expectations to share them,
headcount decisions will draw public scrutiny, and
each new rule adds cost and reduces flexibility,
then the likely response is not higher wages. It is fewer roles placed in Singapore to begin with. Hiring in the next market, or not hiring at all, carries no retrenchment exercise and no headline. That quieter outcome could affect overall employment more than any single wave of layoffs.
This also raises a fair question of accountability. If guidance to businesses proves costly — in jobs relocated or firms that struggle — who bears that cost? Businesses and their employees usually do. That is a reason for owners to evaluate public guidance on its merits for their own situation, rather than treat it as a directive.
6. The other side of the ledger
An objective assessment has to state the opposing case at its strongest. Here is what supporters of gain-sharing would say:
Social licence. Singapore's pro-business model rests on broad public support. If growth feels "jobless" to many households, that support can erode — with consequences for every employer. Parliament passed a motion in May 2026 on "An AI Transition with No Jobless Growth" (NTUC).
Public investment. Government has committed over S$10 billion to workforce initiatives over five years, much of it helping firms adopt AI and retrain staff (People Matters). Officials have framed this as a partnership in which a "fair deal for workers" is expected in return (Mothership, May 2026).
Business self-interest. Firms that keep and reskill people may build stronger capability. The government cites PSA, which reached record cargo volumes while redeploying more than 2,000 workers into higher-skilled roles (MOS Jasmin Lau, 6 May 2026).
Evidence of upside. Among AI-adopting firms, about 70% saw higher worker productivity, and adopters showed higher revenue and total employment (MTI, 9 Sep 2026).
It is also worth noting that official positions are not uniform. Minister Tan See Leng has argued that the best use of any AI surplus is to fund accessible upskilling, rather than framing workers as passive recipients. That is closer to an investment view than a redistribution view. The real debate may be less about whether to share gains and more about how — wages, bonuses, training, or redeployment.
7. What business leaders should weigh
The decision on how to use AI gains belongs with each business, its board and its shareholders. Public guidance is one input, not the answer. Before committing, leaders may want to work through these questions:
Where do our gains actually come from? Separate AI-driven savings from demand growth, pricing and cost pressures. Be honest about what is durable.
Are we growing output or trimming inputs? If demand is growing, redeploying people may earn more than removing them. Take up the Senior Minister of State's question: where are your customers' and your industry's bottlenecks, and who in your team understands them best?
Which form of sharing fits our risk? Variable pay, profit-sharing and funded training carry less long-term risk than permanent wage rises.
Who are our critical people? Reward and promote those who drive outcomes. That is sharing gains where it matters most to the business.
If we must reduce headcount, can we do it well? Follow notification rules and tripartite guidelines, and support outplacement. How a company exits people shapes its reputation with those who stay.
What is our location strategy? Test whether Singapore still gives the best mix of talent, cost and flexibility for each function — and revisit it as rules evolve.
How do we tell the story? Explain AI decisions clearly to staff, unions and the public before others explain them for you.
At FYT, our view is simple: AI accelerates tasks; humans interpret and decide. The same holds here. The data can inform the choice, but leaders must make it — and own it.
Comments