Key Takeaways
- Five major executives spoke publicly about AI and job displacement between May 21–26, reaching five entirely different conclusions — all in the same week
- David Solomon (Goldman Sachs) called job fears "overblown" and said 25% automation equals reallocation, not elimination. Jamie Dimon (JPMorgan) disagreed, expecting fewer bankers and more AI specialists through attrition. Both are probably right at their own scale.
- Sam Altman (OpenAI) admitted on May 26 that he was "pretty wrong" about white-collar job displacement happening faster. In January, Dario Amodei (Anthropic) warned of 10–20% unemployment risk. In May, Amodei reversed to the Jevons Paradox: automation of 90% of a job creates demand for the remaining 10%.
- Jensen Huang (Nvidia) said the "lazy" narrative is that AI replaces people — the real story is people who use AI will replace people who don't. Uber's COO noted that AI cost-justification remains unclear in production, even with heavy adoption.
- The disagreement is real, not performative. The people most informed about AI scaling can't agree on labor market impact. That's the signal. For practitioners, the answer isn't in waiting for consensus.
The Same Week, Five Voices, Five Positions
Between May 21 and May 26, the CEOs and leaders of the institutions most invested in AI went public with their forecasts on AI and employment. They did not reach the same conclusion. What's remarkable isn't that they disagree — that's normal. What's remarkable is that the people most informed about AI scaling speeds, cost curves, and deployment friction all landed on different answers to the same question: Does AI eliminate jobs or create them?
David Solomon, Goldman Sachs CEO, went first. In a May 22 New York Times op-ed, he argued that fears about AI-driven job losses are "overblown." Automation handles 25% of finance tasks, which means reallocation and upskilling, not elimination. New roles emerge. The historical pattern holds. By Solomon's view, your firm is going through the same transition banks went through with electronic trading.
Jamie Dimon, JPMorgan CEO, disagreed in the same news cycle. Speaking on May 21–22, Dimon said he thinks AI "will reduce our jobs down the road" — fewer bankers required, but more AI specialists hired to replace them. It's not no job loss; it's job transformation. Your people leave through attrition. Your hiring mix changes.
On the same day, Jensen Huang (Nvidia) pushed back on the entire framing. The "lazy" narrative is that AI replaces people. The real story: "You won't lose your job to AI — you'll lose it to someone who uses AI." It's not a technology problem. It's a competitive advantage problem. Use the tool or be replaced by someone who does.
Then the people who actually build the AI changed their minds.
The Reversals
In January 2026, Dario Amodei, CEO of Anthropic, issued a stark warning. AI would displace 10–20% of white-collar workers. The pain would be "unusually painful." The window between 2026 and 2031 would be the hardest transition. Amodei wasn't being rhetorical. He was modeling the impact and warning the world.
In May — four months later — Amodei reversed. He pivoted to the Jevons Paradox, an economics principle that says when you automate 90% of a job, you create enough new demand for the remaining 10% that the total workforce expands. People still do the work. The lever just got longer. Amodei's January apocalypse became May's demand-expansion story. Both analyses came from the same data. The January forecast was wrong.
Sam Altman went even further. On May 26 — today — Altman posted that he was "delighted to be wrong" about white-collar job displacement. He'd predicted faster and deeper impact. The reality has been slower. He even tested GPT on his own Slack and "switched back" — meaning the tool didn't improve his workflow enough to justify the cognitive overhead. The man who built the most widely-used AI just admitted it's less transformative than he forecasted.
What changed between January and May? Not the technology. The technology got better. What changed is the operating reality of deployment: AI is harder to integrate into workflows than the models predicted. ROI is less automatic than the cheerleading suggested. Dislocation happens slower than the math implied. The two people most alarmed and most informed have both stepped back from their apocalypse forecasts. That's not reassurance. That's signal.
The Operating Reality Check
Uber's COO added one more data point. AI costs are "hard to justify" inside operating companies. Token usage doesn't map to useful output. The automation is real, but the value extraction is unpredictable. This is the ground-level truth that the macro debaters don't address: your AI vendor can show you a 40% task automation rate and you still won't know if you should deploy it.
Nexairi Dispatch
Get the next AI move before it turns into a workflow problem.
Join the free newsletter for concise AI news, practical checklists, and the decisions practitioners need to make next.
Free. No spam. Unsubscribe anytime.
What Actually Matters for Your Firm
Here's what I think is happening: the macro debate is real and unresolved. Solomon and Dimon are both right — it depends on your scale, your labor market, and your willingness to hire specialists. Huang is right — comparative advantage is ruthless. Altman and Amodei are right — the technology is scaling slower than the January hype suggested. Uber's COO is right — ROI is unclear.
The signal for a CPA firm is not to wait for consensus from Fortune 500 CEOs. The signal is to make your own firm-level decision about what work you want to automate, how you'll hire and upskill your team, and what you'll charge clients when your cost structure changes.
Gartner's models show net job creation from AI beginning in 2028. The World Economic Forum projects 78 million net new jobs by 2030 after 92 million are displaced. Both are probably true at the global macro level. At your firm level, the relevant questions are: (1) Which of my current workflows do I want to automate in the next 18 months? (2) What does my labor model look like if I do that? (3) Will I hire specialists to replace the displaced volume or will I upskill existing staff? (4) How does this affect my pricing and client engagement model? (5) What am I willing to sacrifice in service breadth to move faster on execution?
Those are not technology questions. They're strategy questions. And you don't need Goldman, JPMorgan, OpenAI or Anthropic to answer them. You need a clear answer from your own leadership about what your firm looks like on the other side of AI adoption.
The CEOs are still debating whether it's 2028 or 2032 when everything changes. Your firm's decision day is now.
The Real Signal
The fact that Altman reversed his forecast and Amodei stepped back from the apocalypse tells us something the cheerleaders don't want to admit: we're 18 months into GPT-4 scale and we still don't know if the labor market disruption will be fast or slow, permanent or cyclical, concentrated or distributed. The people most informed have stopped predicting with confidence. That's not a reason to panic. It's a reason to stop waiting for permission from tech CEOs and start deciding for your own firm.
Sources
- Fortune — Goldman Sachs CEO David Solomon on AI and Jobs (January 2026)
- Bloomberg — Jamie Dimon on JPMorgan AI Hiring Strategy (May 21, 2026)
- Euronews — Sam Altman: "I'm Delighted to Be Wrong" on AI Job Displacement (May 26, 2026)
- Fortune — Dario Amodei on the Jevons Paradox and AI Job Creation (May 2026)
- Fortune — Jensen Huang: "You Won't Lose Your Job to AI" (April 2026)
- Gartner — AI Will Create More Jobs Than It Eliminates Beginning in 2028 (May 2026)
Related Articles on Nexairi
Free Assessment
Is your firm ready for AI?
A 5-minute governance check for CPA firms using ChatGPT, Copilot or AI accounting software. Get your score and your top gaps — free.
Editor in chief at NEXAIRI, guiding reporting and long-form features. Previously led editorial teams at regional publications across the Southeast.


