The host of Moonshots, Peter Diamandis, lined up two job stories before handing the floor over. The first was a Washington Post piece arguing that the job apocalypse is unlikely: historically, from the loom to the automobile to the internet, technology transitions created more jobs than they destroyed, displacing specific tasks rather than whole occupations. Its thesis, as he summarized it, was that AI is replacing tasks inside jobs, and that people who learn to use it become more productive rather than less employed.
The second cut the other way. Goldman Sachs had warned that professional services firms — consulting, law, accounting — face an existential risk as AI improves faster than their partners adapt, opening a gap in which junior staff who use AI outperform senior partners who refuse to. The dividing line, in that framing, is not education but AI fluency.
Salim Ismail's first response was a plea rather than an argument: be evidentiary and data-driven about how you see the world. He named the people he thinks fail that test, starting with Bill Gates, who had said two days earlier that the jobs were going. Ismail called himself a massive fan of Gates's philanthropy, but said that as a futurist Gates has "somewhat of a dismal track record" — he missed the internet, he missed mobile. He put Ray Dalio, Larry Fink and Yuval Harari in the same category: unbelievably good at framing the past, questionable on the future, because they do not really understand exponential change.
What the survey measured
Then he called for his slide. The data came from Principal Financial Group, which sells health, benefit and retirement plans to small and medium-sized employers, and which Ismail described as having 130,000 customers and a cross-section of the country.
The survey itself is narrower than that customer base. According to Principal's August release, it was an online survey of 1,000 business owners, decision-makers and leaders at employers with between 2 and 10,000 employees, fielded from 22 June to 13 July 2026 through a third-party research panel rather than a census of Principal's own clients.
The headline number Ismail seized on: 1.4%. He described it as the share of companies that had cut staff and blamed AI or automation for it; in the release, 1.4% is the share of all surveyed businesses reporting a staffing decrease they attributed to AI or automation. "This is so inverse," Diamandis said. "It's so inverted to the current BS tropes of radical unemployment."
The other figures are expectations, not events. Asked about the next 12 to 24 months, 4% of employers expected AI to reduce both staffing and wages. Thirty-one percent expected both to rise, and a further 24% expected higher wages with staffing unchanged. Over the preceding year, the share of employers who said AI simply did not apply to their business fell from 19% to 10% — on the podcast, Ismail described that group as the companies not using AI.
He added a claim of his own about why this particular slice of the economy matters: over the past 50 or 60 years, he said, 100% of job growth has come from small and medium-sized companies, with large firms growing while getting more efficient, for a net contribution of zero. Reading further from the slide, he said that over the last three months 52% of these employers had increased staff, 30% had held it steady and 12% had reduced it.
On the Goldman warning, Ismail deeply disagreed. As the world gets more volatile, he argued, companies need more help, not less — which is what he says he is seeing. What has to change is the billing: firms will move from selling hours to selling outcomes, something he said is already starting to happen. He did not think that was a particularly big deal.
Above the loop
The question Ismail says he gets most often is what human beings will actually be doing. His answer was a hundred-year tour through accounting. It starts with double-entry bookkeeping by hand: credits on one side, debits on the other, in pencil with a big eraser, across tabular columns and stacks of ledgers. Calculators and slide rules made the arithmetic faster, but the entries were still made by hand. Today the software reads the bank account and makes the entries itself.
The person, in his phrase, is no longer in the loop but above it — categorizing transactions and vendors, sorting types of incoming revenue, chasing reconciliation gaps, looking at the process flows. He thinks that is more valuable work than the rote entry it replaced, and he reached for the economist Erik Brynjolfsson's term for what is being removed: white-collar drudgery. In big companies, Ismail said, an enormous amount of activity is massaging information — taking something from a sales report and putting it somewhere else — and that is the part he expects to disappear, leaving curatorial judgment and experience applied to the work rather than the work itself.
His own evidence was three books. The first took three years and was "just horrible," because there was no help: every line had to be written and checked by hand. The second, written with Diamandis, took two and a half years and was in his telling even worse, because of the constant question of how much of the first book to carry over. ("It was an absolute joy working with you," Diamandis interjected, "so don't you dare say anything like that.") The third has taken six months of what Ismail called unadulterated joy, because he can tell the model that Bill Gates or Satya Nadella just said something, have it read through the book and, acting as a developmental editor, tell him where the inserts should go. People forget, he said, how much more fun the work is when something else is doing the cognitive drudgery — and how much more creative you can be once the thinking left to you is judgment.
The people who aren't in the room
Dave's contribution was a story about the people who opt out. He and his brother begged their parents for an Apple II, the first PC anyone they knew had. It changed his life. He showed his father a word processor called Super Scrib — no more handwriting your notes. His father, then in his forties, said he had done it this way his whole life and did not need to be a computer person, and finished his career at GE without becoming one. Dave half-forgives him: the machines broke constantly, the printer was a nightmare. This, he said, is not hard by comparison, and it rewards builders and creators regardless of technical skill.
Diamandis's advice for anyone in that position was to treat the model as the most patient, most capable teacher available: start at zero, ask it to build you a curriculum from not understanding AI to being able to use it, and spend at least 30 minutes a day in conversation with it. It is free, he said, and already on your phone.
Alex closed on a different claim, about individuals rather than payrolls. He traced the segment's premise back to Brynjolfsson, and said many people do not realize they have permission to be high agency — that agency is one of the few human traits actively rewarded in an era of superintelligence, and not something reserved for superheroes in Hollywood movies. Anyone can exhibit it now, he argued, and it is one of the few things left that can still produce extreme vertical mobility. His warning was about timing: there is a window, and he does not know how long it lasts. "Maybe only a handful of years. Do it now. Otherwise, do it never."
That forecast is doing something different from the slide. The survey records what a sample of employers did with their staffing and what they expect to do next; Alex's window is a claim about how far an individual can move in the meantime.