AI could already replace nearly 12% of jobs - MIT

According to new research from MIT’s Project Iceberg, today’s AI systems are already capable of taking over tasks equivalent to 11.7 per cent of the U.S. labour market, representing roughly US$1.2‑trillion in annual wages. With a total workforce of about 151 million workers modelled in the study, that share translates into work now technically and economically within reach of AI for nearly 18 million jobs’ worth of activity. 

For the human‑resources profession, the question is no longer whether AI will change employment. It is how quickly organizations will choose to act on this capability — and whether HR will shape that transition or be dragged along behind it.

What the Iceberg Index actually measures

Project Iceberg simulates the U.S. workforce as 151 million “agents” spread across 923 occupations, each tagged with skills, tasks and locations, and mapped against more than 32,000 discrete skills and thousands of real-world AI tools. 

The core metric, the Iceberg Index, doesn’t count layoffs. It measures technical exposure: where AI systems can already perform the tasks attached to specific skills at a competitive cost. The authors are explicit that the Index “captures technical exposure, where AI can perform occupational tasks, not displacement outcomes or adoption timelines.” 

In plain language: the study estimates how much human work could be shifted to AI today if organizations chose to adopt and integrate existing tools at scale.

That distinction matters. For HR, the 11.7‑per‑cent figure is not a forecast of immediate job cuts; it is a map of where the ground is already weak under employees’ feet.

The tip of the iceberg is visible. The rest is not

So far, public attention has fixated on high-profile layoffs in technology and IT roles. Yet the study suggests that visible AI adoption in computing and technology represents only “the tip of the iceberg,” amounting to about 2.2 per cent of total wage exposure, or roughly US$211‑billion. 

The bulk of immediate AI capability sits in less glamorous corners of the economy: financial operations, administrative support, customer service, logistics, and professional services. Fortune’s summary of the work 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 particularly high exposure in finance, health care and professional services. 

If your organization employs knowledge workers who process information, move numbers through systems, schedule or route work, or standardize documents, the Iceberg numbers say a meaningful fraction of that work is now technically automatable.

How many employees are really at risk?

Because Iceberg measures tasks, not jobs, it would be simplistic to equate 11.7 per cent of wage exposure with 11.7 per cent of employees losing their roles. Many jobs will be partially automated, with AI handling chunks of work rather than entire positions.

Still, some basic arithmetic is instructive for HR planning:

• Total workforce modelled: 151 million workers. 

• Share of wage value that AI can currently perform at competitive cost: 11.7 per cent. 

Applied directly, that implies AI is already capable of performing work equivalent to about 17–18 million full‑time jobs in the U.S. economy.

In practice, this exposure will manifest in several ways that HR will recognize:

• Headcount growth plans slowed or frozen in areas where AI can absorb incremental workload.

• Roles redesigned so that one employee, augmented by AI, does the work previously done by several.

• Natural attrition not fully backfilled, with AI systems taking over routine components.

• Targeted reorganizations where specific job families (for example, entry-level analysts or coordinators) are consolidated.

In all of these scenarios, the tasks Iceberg flags as automatable will quietly disappear from human job descriptions, even if official job titles remain.

Why HR cannot wait for the CIO

Project Iceberg was explicitly designed as a planning tool for governments making billion-dollar bets on training and infrastructure. It allows states to explore how different policy scenarios affect “workforce exposure patterns before committing billions to infrastructure and training programs.” 

HR leaders should treat it the same way: as a stress-test for their people strategy.

Three implications stand out.

1. Exposure is granular, not generic.
The simulation tracks exposure down to the level of skills and counties, not just industries. That means two employees with the same title may face very different futures depending on their task mix and location. HR analytics teams will need to go beyond job families and map actual work content: who spends their day reconciling invoices, drafting boilerplate contracts, or triaging service tickets?

2. Early movers will capture the productivity gains.
The study’s authors highlight that AI’s impact depends not just on technical capability but on how quickly organizations integrate tools into workflows and change management. Those who treat 11.7 per cent as a distant worry may find competitors using the same statistic as a target for margin improvement or service expansion.

3. Redeployment, not replacement, will define HR’s reputation.
Iceberg does not prescribe which jobs vanish; it shows where organizations have options. That puts HR squarely in the role of designing what happens next: redeploying exposed workers into growth areas, redesigning roles around uniquely human strengths, and negotiating new social contracts with employees who know some portion of their job could be done by a machine.

What this means for Canadian employers

The Iceberg model is U.S.-focused, built on American occupation and wage data. But Canada’s labour market is tightly coupled to the U.S. in both industry mix and technology adoption. Financial services, health care administration, government, and professional services — all heavily exposed in the U.S. estimates — are pillars of the Canadian economy as well.

HR leaders north of the border should resist the temptation to dismiss Iceberg as foreign policy research. Instead, they can treat the 11.7‑per‑cent benchmark as a starting point and ask:

• If similar AI capabilities were applied to our own processes, where would technical exposure cluster?

• Which Canadian roles mirror the U.S. occupations the Index finds most at risk — and which are buffered by regulation, union contracts or unique local skills?

• Do our current reskilling programs, tuition supports and career paths assume a world where jobs change slowly, or one where 10–15 per cent of work content can shift in a few planning cycles?

Given Canadian debates over productivity and competitiveness, the bigger risk may be under‑adoption rather than over‑automation.

A new mandate for HR

The Iceberg findings offer HR a rare opportunity: advance warning.

Instead of reacting to layoffs announced by finance or technology functions, HR leaders can use tools like Iceberg to:

• Build task‑level exposure maps for their own organizations, identifying where AI could already replace or augment work.

• Negotiate explicit AI transition plans with business leaders — including commitments on redeployment, retraining spend per exposed employee, and timelines for role redesign.

• Shift talent models from hiring for fixed jobs to cultivating adaptable skill portfolios that can move across roles as automation reshapes work.

• Update workforce metrics so executives see not only headcount and vacancy rates, but the share of labour in highly exposed tasks — a new KPI for the AI era.

The research coming out of MIT is blunt: the window to treat AI as a distant issue is closing. For HR, that means moving beyond ethics panels and pilot projects to something more uncomfortable but necessary — putting concrete numbers on how much of today’s work could, right now, be done by machines, and planning human futures accordingly.

Because whether 11.7 per cent becomes a story of mass displacement or managed transition will depend less on algorithms than on the decisions HR helps organizations make in the next few years.