Displaced by a Forecast, Not a Fact

Companies are firing workers based on what AI might do, not what it has done. CFOs at 750 US firms expect 502,000 AI-linked workforce reductions in 2026, a ninefold increase from 2025, but research from the Atlanta Federal Reserve found that the productivity gains driving these decisions are perceived, not measured. These are not abstract forecasts. These are headcount decisions already in motion. Workers are being displaced by a forecast, not a fact.

What does the new data actually say about AI and jobs?

The numbers are specific and worth sitting with. The Duke CFO Survey, published through the National Bureau of Economic Research (2026), covered 750 firms across every major sector. 44% of those firms expect to reduce headcount because of AI. That translates to 502,000 projected workforce reductions in the US this year, up from 55,000 attributed to AI in 2025.

Related research from the Atlanta Federal Reserve (2026) makes it worse. The productivity gains executives believe AI delivers exceed the productivity gains their own organisations can measure. The gap between perception and measurement is not small. It is the gap driving restructuring decisions at companies that are otherwise profitable and growing.

Why are companies cutting faster than AI can deliver?

Because the market rewards it. When Block announced 4,000 job cuts tied to AI, its stock rose nearly 18% in the days following, according to Fortune (2026). That creates a feedback loop. Boards see AI-linked restructuring rewarded by investors. They accelerate timelines. Cognizant's workforce research projected these changes for 2032; they are happening now, six years ahead of schedule.

The incentive structure is clear. Companies are not cutting because AI proved it could do the work. They are cutting because investors expect them to act as if it can.

What is the AI productivity paradox?

The perceived productivity gain from AI exceeds the measured productivity gain. That is the paradox, and it matters because real people lose real jobs based on the perception, not the measurement.

The 502,000 workers facing cuts this year are not losing their jobs to a machine that replaced their output. They are losing them to a spreadsheet that predicted a machine would.

This is a specific case of a broader failure. GDP measures economic output. It does not measure whether people are safe, employed, or able to feed their families. A company can post record profit, cut thousands of workers, and GDP goes up. The OECD Well-being Framework exists precisely because economic output and human outcomes have been diverging for decades. AI-era restructuring accelerates that divergence. The gains accrue to shareholders. The costs land on workers, families, and a support system that was never designed for this speed.

What does this look like in Australia?

Commonwealth Bank posted a $5.45 billion half-year profit and announced a $90 million AI workforce program, which CBA describes as reskilling and internal mobility. CBA also confirmed role reductions in the same period. The invest-and-restructure pattern is arriving in Australia.

The timing compounds the pressure. The RBA raised rates to 4.10%. The National Debt Helpline recorded 15,857 contacts in February alone, the highest for the month since 2020. The Australian Financial Security Authority projects 13,750 personal insolvencies in 2026-27. Anyone facing displacement should calculate their runway immediately. The person displaced by AI-era restructuring in March 2026 is also facing a rate hike, stretched helpline capacity, and a safety net that takes weeks to activate.

What does this mean for workers displaced by AI in 2026?

It means capable workers are being cut from profitable companies with no bridge support waiting for them. These are people who were performing well, in roles that were profitable, at companies that are growing. They did not fail. A projection said they could be replaced. The support infrastructure was designed for gradual transition, not speculative restructuring at scale.

I see this pattern at TEKVA. People walk in having been cut from companies posting record results. The role was not automated. It was eliminated in anticipation of automation. The freefall gap, the window between the shock and the system responding, is where these people land. Financial counsellor waitlists run six to eight weeks. Centrelink's assets test excludes anyone who was earning a professional salary last month. Tools like TEKVA's Hardship Helper exist because people cannot wait that long. The bridge between panic and a plan does not exist in the current system. That is what we are trying to build.

What should funders and policymakers take from this?

Displaced workers need bridge support now, not retraining programs that arrive in three years. The Job Loss Action Plan is an example of what immediate, practical intervention looks like. And if companies are cutting on projections rather than proof, they should be funding transition for the people they displace. The $1.5 billion in new philanthropic capital for AI impact, including the OpenAI Foundation's $1 billion commitment and the $500 million Humanity AI coalition, should flow to frontline response, not just research.

The investor class that rewards AI-linked restructuring also manages the retirement savings of the workers being cut. Australian superannuation funds hold $3.9 trillion in assets. Their members are the same people losing jobs to forecasts. If institutional investors understood the downstream cost of displacement, the incentive structure would shift. Boards currently operate under fiduciary duties to the company. Scotland's Community Wealth Building legislation (2025) offers a model: embedding community impact into corporate governance, not as charity, but as obligation.

There are sustainability models worth watching. The Wellbeing Economy Alliance advocates for economic systems measured by human outcomes, not just output. The OECD's Beyond GDP initiative is building the measurement frameworks. These are not utopian proposals. They are operating alternatives already adopted by governments in Scotland, Iceland, New Zealand, and Wales. But the displaced worker in Melbourne this week cannot wait for a measurement framework. They need bridge support now.

If you fund workforce transition, partner with organisations managing displacement, or want to understand what is happening on the ground, get in touch.

Frequently asked questions