Goldman Sachs CEO David Solomon, who calls AI causing mass unemployment fears as overblown, says: Bank will continue to…
Word count: 75.
Goldman Sachs will bring on roughly 2,400 to 2,500 interns this year, with a comparable cohort of permanent new joiners walking in come July. That’s the number CEO David Solomon kept circling back to on Bloomberg’s Odd Lots podcast this month, his answer to anyone convinced AI is about to gut the entry-level pipeline at investment banks. Hiring will “contract a little” over the next three years, he conceded—but nothing close to dramatic. For a bank that brought in north of 3,000 in 2021, the message is restrained, not apocalyptic. And Solomon, who used a late-May New York Times op-ed to call the jobs apocalypse overblown, seems determined to keep saying so.The shift in mix is real, just quieter than the headlines suggest. Goldman’s hiring has skewed more heavily toward engineering talent over the past decade, and Solomon expects that tilt to sharpen further as AI reshapes the work. This year’s intake lands roughly at pre-COVID levels.
The loudest AI jobs warnings aren’t coming from Wall Street right now—and that gap is starting to matter
Solomon’s read puts him at the opposite end of the table from Anthropic CEO Dario Amodei, who has warned of an entry-level wipe-out, and closer to Apollo chief economist Torsten Sløk, who sees no real evidence of AI-driven layoffs today. Uber COO Andrew Macdonald has separately argued the spending needed to automate tasks with AI is getting harder to justify. Solomon leans on work from Goldman’s own chief economist Jan Hatzius, who has flagged short-term dislocation without spotting a structurally higher long-run unemployment rate. “It’s no different this time,” Solomon argued in January, framing today’s disruption as another turn of the same wheel that’s been spinning for decades.Inside the firm, the AI push has a name: OneGS 3.0. An internal memo in October flagged a hiring slowdown and potential cuts, with Goldman’s usual 3% to 5% annual performance-based trim pulled forward to the second quarter. The programme rewrites six core processes—client onboarding and KYC, lending, regulatory reporting, and vendor management among them. Even so, the bank expects to close the year with a net increase in headcount, framing the AI effort as capacity expansion rather than a quiet cull.
The trickier problem isn’t who gets hired this year—it’s how you train an analyst who never had to do the math
That, Solomon said, is the real headache. He recalled starting out without minute-by-minute price feeds, comparing stocks by pulling figures from The Wall Street Journal, plotting them on graph paper, running the math longhand. Slow, but it built instincts. Now the answers arrive instantly, he noted, leaving open the question of whether anything actually gets learned along the way. His advice to juniors in the AI age sounded almost retro: pick up the phone. A call, he said, is ten times more valuable than a text or an email.
