Morgan Stanley’s New Report on AI and Jobs Reaches a Conclusion That Will Surprise Both Optimists and Pessimists

After reading enough research notes from large Wall Street banks, you begin to notice a certain tone. They hardly ever give their all. After hedging once more, they sneak a genuine opinion into the seventh paragraph. That pattern is followed in Morgan Stanley’s mid-April report on AI and the labor market, but the conclusion it draws is genuinely strange—so strange that both AI utopians and doomers will find something to dislike. Essentially, the bank claims that not much has occurred thus far. which neither side has been asserting.

The main conclusion of the report is nearly uninteresting when first read. Employment is still high across U.S. payrolls in the sectors most exposed to artificial intelligence. Workers who are most exposed to the technology have seen a slight increase in unemployment, while workers in positions like analyst, accountant, and judicial clerk between the ages of 22 and 27 have seen a somewhat greater increase. However, the AI effect decreases when you take into account how various professions react to regular economic cycles. It does not vanish. It simply no longer appears to be the end of the world that everyone was predicting two years ago.

The example that resonated with me was the legal sector. According to the headlines, junior attorneys were supposed to have been replaced by tools that could draft contracts in a matter of seconds and review documents more quickly than a first-year associate could. However, law firms continue to hire. The people are still there, but the work has changed. Similar to spreadsheets in the 1980s, AI has integrated itself into the workflow without eliminating it. The Morgan Stanley economist who oversaw most of the analysis, Diego Anzoategui, points out that a single tool can both replace and enhance a worker, sometimes even within the same job description. That seems easy. In actuality, it is the central point of contention.

The historical sweep is the part of the report I’d recommend reading in full, because it does something modern AI commentary almost never does: it puts the current moment next to five previous innovation waves and looks at what actually happened. Factory workers replaced artisans as a result of the Industrial Revolution. Railroads created completely new types of white-collar labor while ending some occupations. The internet, internal combustion engines, electricity, and postwar electronics all caused disruptions but eventually increased employment. Michael Gapen, the chief U.S. economist, frames it bluntly: innovation waves are disruptive, capital-intensive, and often volatile, but over time they raise productivity and restructure labor markets in ways that, when institutions adapt, lift living standards broadly. The phrase “when institutions adapt” is doing a lot of work in that sentence.

What makes the report genuinely surprising is the productivity number buried in earlier Morgan Stanley research — companies that have used AI for more than a year report average productivity gains of 11.5%. That is not insignificant. In fact, it’s the kind of number that, in any other industrial cycle, would already be reshaping wages and hiring patterns. The fact that it isn’t, at least not visibly, suggests the labor market hasn’t fully metabolized what’s happening yet. Corporate earnings calls have started referencing “displacement” more often than “job creation,” a shift Morgan Stanley reads as directional rather than conclusive. Investors appear to hold one of two beliefs: either the gains will eventually translate into widespread employment growth, or they will concentrate at the top in ways that have been observed in the past.

As this plays out, it’s difficult to avoid thinking that the loudest voices on both sides might have been mistaken about the timing rather than the destination. Although it is small, the displacement is real. Although uneven, the boom is real. Furthermore, policy decisions regarding education, retraining, and the distribution of productivity gains will likely have a greater impact than any benchmark a foundation model meets in the upcoming year. That isn’t stated explicitly in the Morgan Stanley report. After the data, hedges, and historical comparisons, all that’s left is the subtle hint that the AI revolution may ultimately resemble all previous revolutions. disruptive in certain areas. In others, generative. And it’s not at all like the version you were sold.