AI Certainty Comes from Distance

In June 2025, speaking at the Aspen Ideas Festival, Jim Farley, Ford's chief executive, said artificial intelligence would replace "literally half of all white-collar workers" in the United States.

A year later, in June 2026, Ford executives described to reporters what the technology had actually done inside their own company.

They explained that automated quality inspection had fallen short of what Ford expected, and many of the engineers whose knowledge the tools needed had already left. The repair work had already been running for three years and had brought in 350 experienced engineers, some of them former employees with others recruited from suppliers.

Charles Poon, Ford's vice president of vehicle hardware engineering, said Ford had misjudged what would come from feeding its design requirements into artificial intelligence.

That three-year repair effort put the beginning of the rebuilding at around 2023, roughly two years before Farley’s forecast. So, Ford was already recruiting experienced white-collar engineers to its own workforce to cover for the shortcomings of their AI system when the CEO made this declaration on the stage at Aspen about millions of jobs being lost to AI.

Nationally, Challenger, Gray & Christmas record the reason each American employer gives when it announces job cuts. Employers attributed 101,743 announced cuts to artificial intelligence across the first half of 2026, compared with 54,836 for the whole of 2025. It led every other stated reason for four consecutive months to June, and its share climbed from 7% of announcements in January to close to 40% in May.

Oracle's 10-K, filed on 22 June 2026, was one such announcement in corporate form: staff numbers fell from about 162,000 to 141,000 across the financial year and the filing stated that the adoption and deployment of AI across its operations "have resulted, and may continue to result, in reductions to our workforce".

We can see the reasons employers gave for making these cuts, but not know on what basis they acquired the conviction that prompted them. Predictions such as Farley's were public from the moment he spoke, yet the accompanying account of the repair work needed at Ford to correct what the AI had got wrong was kept inside the company until much later. And figures are emerging that make the scale of the overreach visible. Robert Half, a recruitment company, reports that a third of American hiring managers who cut a role primarily because of AI have since started to rehire for it. Orgvue, which sells workforce planning software, reports that 55% of leaders who made redundancies during an AI roll-out later called the decision wrong. In Australia, the Commonwealth Bank installed an AI voice system, cut around 45 customer service roles, reversed the decision as call volumes rose, and apologised to the people it had let go.

An ideas festival asks of its speakers a remark that the audience will carry out of the room, and the prediction that artificial intelligence would replace "literally half of all white-collar workers" in the United States is a sentence of precisely that kind, shaped to lodge in the memory. Ford's own account of the same technology, gathered from inside its own walls, was of an altogether different order. The tools had fallen short of what the company expected, and experienced engineers, some of them former employees, others recruited from suppliers, had been needed to repair the damage. It required three years and a formal press briefing before that version of events existed at all. Throughout that whole period, an employer deciding what to do about its own AI programme had access to the prediction delivered from a festival stage, but to none of the private experience that would later emerge. The people whose jobs those decisions affected were working from exactly the same incomplete information.