AI could already replace nearly 12% of jobs, says new MIT study

New research from MIT’s Project Iceberg suggests today’s artificial intelligence systems are already capable of taking over tasks worth 11.7 per cent of the U.S. labour market, or about US$1.2‑trillion in annual wages. With roughly 151 million workers modelled in the study, that means AI could now technically and economically perform work equivalent to nearly 18 million full‑time jobs south of the border. 

For Canada — a smaller, services-heavy economy tightly integrated with U.S. supply chains — the question is not whether similar pressures will arrive, but how quickly, and whether HR will help manage the landing or merely clean up the aftermath.

What the Iceberg Index really tells us

Project Iceberg simulates the American workforce as 151 million “agents” spread across 923 occupations, each tagged with the specific skills and tasks they perform and mapped against more than 32,000 skills and thousands of real AI tools. 

Its core metric, the Iceberg Index, measures technical exposure: where current AI systems can already perform the tasks attached to particular skills at a competitive cost. The authors stress that the Index “captures technical exposure, where AI can perform occupational tasks, not displacement outcomes or adoption timelines.” 

In other words, this is not a layoff forecast. It is a map of where employers already have the option to shift human work to machines.

That nuance matters for Canadian HR teams. The 11.7‑per‑cent figure does not mean one in nine employees will vanish next year. It means a meaningful slice of the work Canadians do today could, in principle, be done by AI — and that strategic decisions in the next few planning cycles will determine how much of that potential is realized, and at whose expense.

The visible layoffs are only the tip

Public attention has largely fixated on high‑profile job cuts in Silicon Valley and corporate IT departments. Yet Iceberg’s authors argue that visible AI adoption in computing and tech amounts to just “the tip of the iceberg,” representing about 2.2 per cent of total wage exposure — roughly US$211‑billion in work. 

The real action lies in more mundane domains: financial and professional services, administrative support, logistics and back‑office operations. A summary of the research notes that current systems could already take over tasks tied to “roughly 11.7% of total wage value, or around $1.2 trillion in pay,” with especially high exposure across finance, health care and professional services. 

Canadian employers are heavily concentrated in the same sectors. Banks, insurers, pension funds, health‑care administrators, law firms, consulting shops and shared‑services centres all lean on the kinds of knowledge work Iceberg flags: processing information, moving numbers through systems, drafting and reviewing standard documents, scheduling and routing work, and triaging service requests.

When that work is broken down into discrete tasks, the Iceberg math implies a significant fraction is already technically automatable — even if head offices haven’t yet pushed the button.

What might 11.7 per cent look like in Canada?

Statistics Canada’s Labour Force Survey shows an employment rate hovering around 60 to 61 per cent in 2025, based on monthly samples of about 100,000 respondents. While the agency doesn’t yet quantify AI exposure, the structure of Canada’s labour market is similar enough to the U.S. that Iceberg’s numbers should be a wake‑up call.

Because the Index measures wage value, not headcount, it’s misleading to directly apply 11.7 per cent to Canada’s workforce and declare an equivalent share of jobs “replaceable.” Many roles will be partly automated: AI absorbs routine components, while humans handle exceptions, client relationships and complex judgment calls.

Even so, Canadian HR leaders can expect the exposure to show up in familiar ways:

• Expansion plans quietly scaled back in functions where AI can soak up incremental workload.

• Job redesigns where one AI‑augmented employee does what two or three employees once did.

• Attrition that isn’t fully replaced, with automation taking over repetitive tasks.

• Targeted reorganizations of specific job families — junior analysts, coordinators, claims processors, contact‑centre staff — whose day‑to‑day work maps closely to Iceberg’s high‑exposure skills.

On paper, employees keep their job titles; in practice, the tasks that once filled those titles start to evaporate.

Why HR, not just IT, has to lead

Iceberg was built for governments contemplating “billion‑dollar” investments in training and infrastructure and wanting to understand workforce exposure “before committing billions to infrastructure and training programs.” If Ottawa and the provinces see value in stress‑testing policy scenarios, Canadian corporations should be doing the same with their talent strategies.

Three lessons emerge for HR.

1. Exposure is about skills and tasks, not just jobs.
The model tracks 32,000 skills across counties and occupations. That means two employees with identical titles in Toronto and Calgary may face different risks depending on their actual task mix. HR analytics teams will need to move beyond org charts and map what people really do all day — from reconciling invoices to summarizing medical notes — and then assess which of those tasks are already demonstrably automatable.

2. The productivity gains will not wait for consensus.
Iceberg distinguishes between what AI can do and what organizations choose to implement. Canadian employers competing with U.S. firms will face pressure to match productivity improvements driven by AI, especially in cost‑sensitive functions. Hesitation may not prevent disruption; it may simply mean it arrives under duress, through budget cuts and outsourcing, rather than through planned redeployment.

3. HR’s reputation will hinge on what happens to exposed workers.
The Index doesn’t specify which jobs disappear; it shows where leaders have choice. That places HR at the centre of designing pathways for employees whose work is at risk — and of setting expectations early. Will automation savings be partially reinvested in reskilling? Will there be guaranteed internal mobility windows before redundancies? Or will workers discover their exposure only when severance packages are issued?

A Canadian playbook for an AI‑exposed workforce

If Iceberg gives Canadian employers advance warning, what should HR do with it?

First, build your own exposure map. Use job‑task analysis and process mapping to identify where employees spend time on activities similar to those Iceberg highlights as automatable — from document classification to standard report writing. Even a rough estimate of task‑level exposure will be more useful than generic lists of “AI‑impacted roles.”

Second, negotiate explicit AI transition plans with the C‑suite. Before a single pilot scales, HR should be at the table setting policies on redeployment, retraining and timelines. Project Iceberg was designed so policymakers can test scenarios before making irreversible investments; HR should insist on the same discipline for workforce decisions.

Third, shift from jobs to skills. Iceberg is, at its core, a skills‑centred KPI. Canadian employers can follow suit by inventorying critical skills, funding targeted upskilling and building internal marketplaces that let employees move away from highly exposed tasks into growth areas — for example, from routine claims processing into fraud analytics or customer advocacy.

Finally, update the metrics that reach the boardroom. In addition to vacancy rates and turnover, boards should be seeing a simple AI‑era indicator: the share of the company’s wage bill tied to highly exposed tasks. In a country where productivity is a perennial concern, that figure will be hard to ignore.

MIT’s researchers warn that “the window to treat AI as a distant future issue is closing.” For Canadian HR leaders, that window may be even smaller, squeezed between U.S. competitive pressures and domestic expectations around fairness and social responsibility.

The Iceberg Index shows that AI could already do a non‑trivial share of the work Canadians are paid for today. Whether that turns into a story of managed transition or avoidable dislocation will depend on choices employers — and their HR leaders — make now, before the numbers show up in the next round of restructuring announcements.