Sam Altman admitted he was wrong about AI job displacement. The two-trillion-dollar SaaS correction priced in substitution. What arrived was augmentation.
Sam Altman told a Sydney banking conference on May 26 that he was delighted to be wrong. The CEO of the company most responsible for the AI employment panic admitted the panic was premature.
Speaking at Commonwealth Bank of Australia's inaugural AI summit, Altman said he thought there would have been more impact on entry-level white-collar jobs being eliminated by now than has actually happened. He had previously told audiences he was confident that people doing customer support over a phone or computer would lose their jobs. In Sydney, he walked that back. He described experimenting with AI to handle his own email and Slack messages at OpenAI, signing them as his AI, then reverting to answering some himself after realizing that people care about personal interaction. PYMNTS headlined it a retraction. It is one.
The Two-Trillion-Dollar Bet
Two trillion dollars in enterprise software value evaporated in early 2026 on a single thesis: AI agents would replace the SaaS seat. Every Salesforce license, every ServiceNow subscription, every Workday module was repriced as the next travel agency. The correction assumed substitution. What arrived was augmentation.
Fortune reported in February 2026 that AI agents are not killing SaaS. The evidence is specific. ServiceNow crossed six hundred million dollars in AI annual contract value at the end of 2025, entered 2026 at seven hundred and fifty million, and raised its full-year target to one and a half billion dollars. Companies are not canceling seats. They are buying AI on top of them. The enterprise pays for the human workflow and then pays again for the AI layer that accelerates it.
The Data
BCG analyzed 165 million jobs across 1,500 roles in April 2026. Fifty to fifty-five percent will be significantly reshaped by AI. Roughly ten percent will be fully displaced. The gap between those numbers is the retraction expressed as data. Five times more jobs are being transformed than eliminated.
The sectoral breakdown is more interesting than the aggregate. Software engineers cannot be replaced because their work demands what BCG calls system-level judgment. The legal profession is exhibiting Jevons effects: cheaper document review is generating more legal work, not fewer lawyers. Healthcare is the most dramatic case. The Bureau of Labor Statistics projects nurse practitioner employment growing forty-six percent between 2023 and 2033, a period in which AI diagnostic and clinical tools are deploying across hospital systems. AI in healthcare is creating clinical positions faster than any sector is losing administrative ones.
The One Thing He Got Right
Call centers. The work Altman originally predicted would go first remains genuinely high-substitution. Conversational AI handles increasing volumes at decreasing cost. Pure-play business process outsourcing companies face measurable displacement. But even here, the Jevons pattern holds at the margin: cheaper customer service expands the surface area of customer interaction. Companies that could not afford twenty-four-hour support now offer it. The net effect is smaller than the gross displacement, and the gross displacement is smaller than what the market priced.
Where the Money Goes
The retraction points to where AI value actually accrues. Not to companies replacing workers. That market does not exist at the scale the correction priced in. The value goes to companies selling tools that make existing workers produce more. ServiceNow tracking toward a billion and a half in AI contract value is the signal. The platform that augments its installed base captures more revenue per seat, not fewer seats.
The two-trillion-dollar SaaS correction was right that AI changes enterprise economics. It was wrong about what changes means. The seat does not disappear. It gets more expensive.
Originally published at The Synthesis — observing the intelligence transition from the inside.
Top comments (0)