AI can already replace nearly 12% of jobs - MIT

For HR leaders across Asia Pacific – artificial intelligence is no longer just a case study from Silicon Valley or Beijing. It is already reshaping how white‑collar work gets done, and researchers have put a very specific number on what today’s AI systems can handle.

MIT’s Project Iceberg, a large‑scale simulation of the U.S. labour market, finds that current AI tools are already capable of performing tasks worth 11.7% of total wage value, or about US$1.2 trillion a year. The model represents roughly 151 million workers across 923 occupations and more than 32,000 skills. 

For a regional hub like Singapore – built on finance, trade, professional services and high‑end manufacturing – these findings are hard to ignore. The same cloud‑based AI tools that underpin the American numbers are being deployed in banks on Shenton Way, in regional HQs at Marina Bay and in shared‑services centres from Changi to Jurong.

What Project Iceberg actually measures

Project Iceberg does not just speculate about the future. It systematically compares what people do with what existing AI tools can already do.

Each worker in the simulation is treated as an “agent” with a bundle of skills and tasks, located in a particular occupation and region, and matched against thousands of deployed AI systems. 

The central metric, the Iceberg Index, measures technical exposure: the share of an occupation’s wage bill tied to skills where AI has already demonstrated usable performance at least once. The authors stress that the Index “captures technical exposure, where AI can perform occupational tasks, not displacement outcomes or adoption timelines.” 

For Singapore HR leaders, the key point is this: 11.7% is not a prediction of headcount reduction. It is a measure of how much work could be shifted to AI today, if organisations choose to adopt and integrate the tools.

The visible tech impact is just the “surface”

Most of the public conversation has centred on AI coding assistants, tech layoffs and data‑science jobs. Iceberg shows that this is only a small and unusually visible part of the story.

If you look solely at current AI adoption in computing and technology roles, the researchers find exposure of about 2.2% of wage value – roughly US$211 billion. They label this the “Surface Index” and note that it is “only the tip of the iceberg.” 

Below that surface lies the real mass: technical capability “extends far below the surface through cognitive automation spanning administrative, financial, and professional services,” accounting for 11.7% of wage value, or around US$1.2 trillion. 

In other words, AI’s biggest immediate impact is not on software engineers – it is on the routine, rules‑based office work that keeps banks, insurers, logistics firms, manufacturers and government agencies running.

One summary of the study points out that high Iceberg scores in several U.S. regions are driven by “cognitive work—financial analysis, administrative coordination, and professional services.” For Singapore, where a significant share of employment sits in banking, wealth management, regional headquarters and business‑process hubs, that should sound uncomfortably familiar.

Which skills are easiest for AI to take over?

The report does not give a neat list of doomed job titles, but it is clear about the kinds of tasks AI already performs well.

The highest technical exposure is in structured, repeatable, digital tasks, including:

• Document processing and data extraction (contracts, forms, KYC files, invoices)

• Routine administrative and clerical work in offices and healthcare

• Standard financial analysis, reconciliations and regulatory reporting

• Workflow coordination, triage and scheduling across large teams and contact centres

The researchers note that financial institutions are already using AI for “document processing and analytical support,” and healthcare organisations are using it to automate “administrative tasks,” freeing clinicians for patient care. 

In a Singapore context, the exposure is likely to be highest in:

• Operations and middle‑office roles in banks and insurers

• Shared‑services and business‑process outsourcing operations handling AP/AR, payroll, HR admin, procurement and compliance

• Early‑career analyst and associate roles that focus on standardised reporting or documentation

• Administrative support roles in hospitals, universities and government agencies

Crucially, Iceberg is skills‑based. Many jobs will be partially automated: AI takes over the high‑volume, well‑defined work; humans focus on exceptions, relationship management, complex judgement and regional nuance.

Why this matters for HR in Singapore and the region

Project Iceberg was designed so governments can “identify exposure hotspots, prioritize training and infrastructure investments, and test interventions before committing billions to implementation.” That logic applies directly to Singapore’s national skills strategies and to corporate workforce planning across Asia Pacific.

Three implications stand out for HR teams.

1. Traditional HR metrics won’t give sufficient warning

The researchers show that standard macro indicators such as GDP, income and unemployment explain “less than 5% of this skills-based variation” in exposure. If HR only watches vacancy rates, attrition and engagement scores, it will miss where AI is quietly re‑shaping job content until much later.

For a labour market as tightly managed as Singapore’s – with foreign‑worker quotas, sectoral transformation maps and national skills frameworks – that lag could translate into skills mismatches and wage pressure in surprisingly exposed areas.

2. The productivity race will be intense in APAC hubs

The report stresses that AI’s impact depends on how quickly organisations package capabilities into deployable tools and integrate them into workflows. Regional HQs in Singapore are already under pressure to deliver more with leaner teams, and many multinationals will see a 10‑plus% automation opportunity as a straightforward way to improve margins and responsiveness.

For HR, that means AI won’t just arrive as isolated pilots. It will appear in global operating‑model redesigns, offshoring decisions and “efficiency programmes” that land on local teams.

3. HR will be judged on how exposed staff are treated

Iceberg shows where leaders have options; it does not tell them how to use them. That places HR squarely in the centre of decisions about redeployment, retraining and communication.

Will automation savings be partly reinvested in upskilling Singapore‑based staff? Will there be clear internal pathways for employees whose work is highly exposed? Or will workers discover they were at risk only when a regional restructuring memo from HQ lands in their inbox?

A practical playbook for Singapore and APAC HR

So what should HR teams in Singapore and across Asia Pacific do with this information?

1. Build a task‑level exposure map

Follow Iceberg’s skills‑centred approach. Go beyond job titles and:

• Break key roles into actual tasks (processing trade documents, KYC checks, report drafting, claims validation, payroll changes, basic analysis).

• Identify which tasks look like the “document processing and analytical support” and “administrative work” the report singles out as already automated. 

Even a rough, 70/30 style view of human vs AI‑suitable tasks will be far better than guessing.

2. Set an AI transition policy before the tools scale

Work with senior leadership to agree on principles in advance:

• Conditions under which headcount will be reduced versus redeployed

• Minimum retraining investment per exposed full‑time equivalent

• How changes will be communicated to staff and, where relevant, unions or worker representatives

• How Singapore’s regulatory framework (for example, fair employment guidelines, PDPA, sector rules) will be respected in AI deployment

Remember: Project Iceberg exists so policymakers can test scenarios “before committing billions”; HR should insist on the same discipline for AI‑driven workforce changes. 

3. Shift from jobs to skills and internal mobility

The Iceberg Index is deliberately “a skills-centered metric that measures the wage value of skills AI systems can perform.” HR can mirror this by:

• Building a skills inventory for critical roles across APAC

• Aligning with national initiatives like SkillsFuture to co‑fund upskilling

• Creating internal marketplaces and career paths that help employees move from high‑exposure tasks into roles that rely more on relationship‑building, complex judgement and regional expertise

4. Update what goes to the executive committee and board

Alongside vacancy and turnover data, boards should start seeing:

• An estimate of the share of wage spend tied to highly exposed tasks

• Progress metrics on retraining and redeployment for those tasks

• Scenario plans showing how different levels of AI adoption would affect headcount, skills demand and cost structure over three to five years

In a region where cost and productivity are under constant scrutiny, those numbers will speak loudly.

MIT’s researchers make the stakes clear: “The window to treat AI as a distant future issue is closing.” For Singapore and other Asia Pacific hubs, that window may be even shorter, given how quickly global firms can roll out new tools across their regional networks.

AI is already capable of doing a meaningful slice of the work people in Singapore are paid to perform. Whether that becomes a story of managed transition – with thoughtful reskilling and role redesign – or one of abrupt restructuring will depend largely on the choices HR leaders make now, before the iceberg fully surfaces in Asia Pacific.