AI in Singapore: Grow GDP Without Losing Jobs

Singapore’s AI push has exposed a truth most companies prefer to ignore: you can increase output while reducing headcount—and the national GDP numbers will still look fine.

That’s not a theoretical risk. It’s already embedded in how many firms are adopting generative AI and automation: remove labour from workflows, ship the same work faster, and book the productivity gains. Economists commenting on Singapore’s Economic Strategy Review update (Jan 2026) made the point bluntly: growth is necessary for jobs, but it doesn’t guarantee jobs. If AI-driven productivity gains accrue mainly to owners of capital, technology, or data, GDP can rise while wages stagnate and job opportunities thin out.

For this edition of the AI Business Tools Singapore series, I want to turn that macro conversation into a practical guide for business leaders: How do you adopt AI to stay competitive while protecting (and upgrading) the “good jobs” your company needs to scale?

GDP can rise while jobs fall—and business leaders should plan for it

Answer first: Yes, GDP can grow while jobs disappear, because GDP tracks output, not employment.

When companies automate tasks, they often produce more with fewer people. That shows up as higher productivity and profitability—both of which contribute to GDP. Professor Nick Powdthavee (NTU) described exactly this mechanism: firms can replace workers with AI, increase output, lower costs, and raise GDP, without increasing labour demand.

This matters to companies because it changes the operating environment:

• Your competitors can scale faster with smaller teams, putting pricing pressure on everyone else.
• Hiring may get harder in the middle: not because jobs vanish entirely, but because roles get redesigned and the “old” job descriptions stop matching the work.
• Your best people will move to firms that use AI to make their work more valuable, not more replaceable.

The implication isn’t “pause AI.” It’s the opposite: adopt AI deliberately—and treat workforce design as part of the AI project, not an HR clean-up later.

The hidden driver: “jobless productivity” inside workflows

In many Singapore companies, AI adoption starts in the most automatable areas:

• customer support replies
• marketing content drafts
• finance reconciliations
• HR screening
• sales outreach
• reporting and analysis

These are full of routine, repeatable tasks. AI tools can compress those tasks dramatically. The danger is that leadership then measures success only as:

“How much labour can we remove from this process?”

Economists in the CNA piece argue for a better framing:

“Which tasks still benefit from human judgment, accountability, and interaction?”

That switch in mindset is where “good jobs” come from.

“Good jobs” in an AI-rich economy aren’t accidental—they’re designed

Answer first: A “good job” is a role where humans do the work that AI can’t reliably own—judgment, context, relationships, accountability—and AI handles the rest.

OCBC’s Selena Ling raised a point many firms gloss over: when policymakers say “quality jobs,” what do they mean—wages, progression, stability? Business leaders should be equally specific.

Here’s a practical definition I use with clients:

A good AI-era job has three traits:

1. Clear accountability: a human owns outcomes (not just activities).
2. Compounding skills: the role builds expertise that transfers to future workflows.
3. AI augmentation built in: the job is faster and more impactful because AI is part of the standard operating procedure.

If your AI rollout results in roles that are narrower, more monitored, and easier to replace—expect churn.

The stance I’ll take: cost-cutting-only AI creates fragile companies

AI used purely as a cost-cutting tool can look great in quarterly numbers, but it tends to create:

• brittle processes (too much trust in models that still make mistakes)
• weak customer experiences (fast responses, low empathy)
• over-centralised “AI gatekeepers” (one team becomes the bottleneck)
• demoralised teams (people feel replaced, not upgraded)

Professor Powdthavee’s warning about bias and mistakes is the operational reality: AI will be wrong in confident ways. If you remove too much human review, you don’t just cut costs—you also increase risk.

Human–AI collaboration: the simplest way to protect jobs and raise output

Answer first: The safest path to both growth and jobs is human–AI augmentation, where AI removes the repetitive load and humans own decisions and exceptions.

This isn’t motivational poster stuff. It’s workflow engineering.

A practical augmentation model looks like this:

• AI drafts (responses, analyses, summaries, first-pass code)
• Humans decide (approve, reject, escalate, rewrite)
• AI monitors (flags anomalies, missing info, policy issues)
• Humans handle edge cases (high-stakes, ambiguous, emotionally charged)

The result: productivity increases because people can do more, not because there are fewer people.

Examples: what “augmentation roles” look like in Singapore SMEs

You don’t need a huge budget to design around augmentation. Some common redesigns:

• Customer support → Customer resolution specialist

  • AI proposes replies and pulls policy and order history
  • human handles escalation, goodwill decisions, and sensitive issues

• Marketing executive → Demand gen operator

  • AI produces variants, landing page drafts, and audience ideas
  • human sets positioning, ensures compliance, and runs experiments

• Finance ops → Controls & exceptions analyst

  • AI categorises invoices and reconciles transactions
  • human investigates anomalies, fraud signals, and supplier disputes

• Sales development → Account research & meeting quality lead

  • AI generates account briefs and call prep
  • human focuses on discovery, relevance, and relationship-building

These roles are more defensible and typically more satisfying.

“Training is hard because the market moves”—so change how you upskill

Professor Lawrence Loh (NUS) described the real challenge: jobs are continuously redesigned, so training can feel outdated the moment it’s done.

A better approach is continuous, workflow-based upskilling:

1. Pick 2–3 workflows (not 20 tools) that matter to revenue or cost-to-serve.
2. Build AI playbooks: prompts, checklists, guardrails, examples.
3. Track performance weekly: cycle time, error rate, customer satisfaction.
4. Assign role-based proficiency (Bronze/Silver/Gold), not “course completed.”

This converts “learning” from an event into an operating habit.

Singapore’s niche opportunity: AI governance and assurance is a job creator

Answer first: Singapore can create high-value jobs by becoming strong at AI governance, standards, and assurance—and companies can benefit by building these capabilities now.

Economists in the article suggest Singapore is well-positioned to lead in:

• interoperability and standards
• responsible data-sharing
• auditing and governance
• model risk management
• safety testing
• privacy engineering

Assistant Professor Goh Jing Rong (SMU) pointed out the opportunity for “AI assurance” services—work that requires human judgment and professional accountability.

For businesses, this isn’t only a national strategy. It’s a competitive edge. As AI becomes more autonomous (Prof Loh’s point about agentic AI), customers, regulators, and partners will ask tougher questions:

• How do you prevent data leakage?
• How do you detect hallucinations in customer-facing content?
• Who is accountable for automated decisions?
• What’s your incident response plan when AI produces harmful output?

If you can answer those clearly, you win trust—and trust converts.

A simple “AI governance starter kit” for companies

You don’t need a massive compliance team to be responsible. Start with five items:

1. AI use register: where AI is used, by whom, for what purpose.
2. Data rules: what can and cannot be pasted into tools.
3. Human review policy: which outputs require approval (by role).
4. Model risk checks: bias, accuracy sampling, and failure logging.
5. Customer disclosure: when users should be told AI is involved.

This also creates internal roles that don’t disappear when the next model version ships.

A practical playbook: adopt AI tools without shrinking opportunity

Answer first: Build an AI roadmap around workflows, then redesign roles around the human tasks that remain valuable.

Here’s a concrete 30–60–90 day plan that fits many Singapore SMEs and mid-market firms.

Days 1–30: Pick the right AI use cases (not the flashiest)

Choose workflows with measurable pain:

• high ticket volume in support
• long proposal or tender cycles
• slow reporting and decision cadence
• repetitive compliance documentation

Define success metrics upfront:

• time saved per case
• reduction in rework
• customer satisfaction (CSAT)
• conversion rate (for sales/marketing workflows)

Days 31–60: Build “human-in-the-loop” operations

Implement with guardrails:

• require citations or source references for factual claims in drafts
• run A/B tests on AI-assisted outputs vs baseline
• create escalation paths for sensitive decisions
• set up a feedback loop: humans label bad outputs and why

The goal is reliability, not novelty.

Days 61–90: Redesign roles and progression paths

This is where job quality is either protected or quietly eroded.

Do three things:

1. Split tasks into three buckets
   • automate (routine)
   • augment (AI drafts, human decides)
   • human-only (high-stakes, empathy, accountability)
2. Update job descriptions to reflect the new reality.
3. Create progression tied to outcomes (resolution quality, pipeline impact, risk reduction), not manual effort.

If you skip this step, you’ll likely end up with “shadow AI” (people using tools informally), inconsistent quality, and a workforce that feels exposed.

What to watch in 2026: the metrics that matter more than GDP

Answer first: For businesses, the most important numbers aren’t macro GDP prints—they’re internal measures of value distribution and capability building.

Track these four indicators quarterly:

• Revenue per employee (productivity)
• Wage growth by role family (are gains shared?)
• Internal mobility rate (are people moving into redesigned roles?)
• Customer experience stability (CSAT, complaints, churn) during automation

If productivity rises while wages and mobility stall, you’re creating the company-level version of “GDP up, jobs down.”

Singapore’s national strategy is pointing in a direction: don’t assume growth automatically produces jobs. Companies should adopt the same discipline: don’t assume AI adoption automatically produces capability.

The next year will favour businesses that treat AI as a way to raise the value of human work, not delete it. That’s how you stay competitive without hollowing out your own bench strength.

If you’re mapping your AI roadmap for operations, marketing, or customer engagement, the question to keep on the whiteboard is simple:

Which parts of our customer promise require human judgment—and how are we designing roles so people can deliver that promise faster with AI?