Upside Isn’t Automatic
Why heavy AI investment isn’t producing the productivity gains companies expected. And Jeff Bezos thinks AI will grow the job market.
Welcome back from a long holiday weekend. We gave you an extra day to catch up before sending this out. Memorial Day snuck up on us this year, and we’re struggling to reconcile the official start of Summer.
While you were grilling hotdogs, you probably weren’t reading the latest buzz on whether the AI bubble is in fact a bubble that will pop dotcom-style. At the end of last week, there were many public announcements from large organizations that have invested heavily in the tech, but have not seen the productivity gains they expected.
Uber’s president and COO, Andrew Macdonald, said in an interview released Saturday that he can’t connect Uber’s rising AI spend to real productivity. “That link is not there yet,” he said.
What gives the comment credibility is the fact that in Uber’s first-quarter earnings, it reported 95% monthly adoption of AI coding tools across its engineering workforce, with AI agents writing more than one in ten lines of code. Back in April, its CTO said Uber had already burned through its entire Claude Code budget for 2026, four months into the year. The tools are everywhere inside Uber, but the COO still can’t draw a straight line to output. Sounds like there are some alignment issues brewing for sure.
So the popular commentary this week will certainly may be that the bubble is starting to deflate. We think that there is actually a much more useful story to be told, and it’s focused on culture. For Uber the technology worked and the adoption happened. The problem is that buying powerful tools and getting people to use them is only half the job, and most organizations treat it as the whole thing.
Productivity isn’t something AI produces on its own. There’s a reason Captain Kirk and crew were on the Enterprise. For AI to be productive, it has to have people who believe in what they are doing and are inspired to do it. A company can hand every employee the best tools in the world, but will see zero change if the conditions around those people are quietly working against them.
Just last month, Uber’s CEO told investors the company is slowing hiring to help offset its AI investments. Now picture being an engineer there, with leadership asking you to “lean into” the tools while telling Wall Street those same tools are the reason your team isn’t growing. There’s undoubtedly a trust problem. Trust is one of the eight conditions we measure in the NICH Employee Experience Framework. It’s actually the most critical. Low-trust environments rarely succeed.
Unfortunately, organizations feel that they are in an arms race with AI. That kind of urgency means they budget for licenses and training, but assume the human side will sort itself out. It doesn’t. The organizations that pull ahead over the next two years won’t be the ones who spent the most on AI. They’ll be the ones who did the harder, less glamorous work of getting their people genuinely ready for it, starting with giving employees a real reason to believe the technology is there to make their work better rather than to replace them.
Also, last week, Jeff Bezos went on CNBC from Blue Origin’s Rocket Park and told American workers that if you’ve been digging out a basement with a shovel and someone hands you a bulldozer, you should be thrilled. It’s not exacly the communication strategy we would be giving Jeff. But more importantly, (and his point) he called the people predicting AI-driven mass unemployment “dead wrong.” Admittedly, that’s a tough message to deliver the same month Amazon announced roughly 30,000 corporate job cuts, and we get why it makes people wince.
But on the substance, we think he’s right, and so does the data. The World Economic Forum sets one scenario that 92 million roles displaced by 2030 and 170 million created, a net gain of 78 million. MIT economist David Autor’s research found that roughly 60% of the current U.S. workforce is employed in occupations that did not exist in 1940. The economy has run this play before, and it has reliably invented more work than it destroyed. We believe the long-term effect is greater opportunity for many more people if organizations are strategic and prepared.
Productivity gains land where leadership puts them, and new roles get filled by people who were given the runway to grow into them. The macro chart trends up. The micro experience is what you build.
Links
Uber’s COO says the AI-productivity link “is not there yet” (Yahoo Finance) The interview behind Part One, including the 95% engineering adoption figure that makes the productivity gap so notable.
Jeff Bezos calls AI job-loss fears “dead wrong” (eWeek) Bezos’s full bulldozer argument from Blue Origin’s Rocket Park, in his own words.
Bezos says AI will elevate workers as Amazon cuts 30,000 roles (IBTimes) Coverage of Amazon’s roughly 30,000 corporate cuts, the backdrop to the timing tension in Part Two.
Why AI may not kill jobs, per the Jevons paradox (Fortune) Apollo’s Torsten Slok argues that cheaper professional work means more of it, the mechanism underneath the net-job-creation case.
How AI will affect work across industries (World Economic Forum) The projection behind the 92 million displaced and 170 million created figures, broken out by industry.
Generative AI could affect 300 million jobs globally (Nexford) Goldman Sachs’s estimate that AI could touch 300 million jobs worldwide, paired with its forecast that the real unemployment effect stays small and mostly temporary.