Bill Gates’s latest essay argues that businesses face a difficult stretch as AI matures, and that many leaders aren’t preparing for it. Most of the reaction has focused on jobs because for anyone whose work is changing, that question matters most to them.

The essay also raises another question: how organizations decide where and how to use AI, and whether they’re doing it well. Stanford’s 2026 AI Index, a new study that’s currently one of the most comprehensive independent reviews of the field, suggests it’s mostly a question of leadership, and it impacts employees as much as it does shareholders.

‍

What the employment data does and doesn’t show

So far, the AI Index finds no sign of large-scale job losses in overall employment data. But, while the evidence does not point to broad, uniform displacement, it is an average, and averages offer little comfort to the people impacted. 

The report finds that AI’s labor effects are showing up unevenly. They appear first in hiring pipelines, among younger workers, and within specific business functions.

The clearest signal is among people at the start of their careers. In two of the occupations most exposed to AI, software development and customer support, employment for workers aged 22 to 25 has declined since 2022, even as demand for more experienced workers in those jobs kept growing. For young software developers, the drop was close to 20 percent from the 2022 peak. 

The clearest signal is among people at the start of their careers. Across the occupations most exposed to AI, employment for workers aged 22 to 25 has fallen roughly 16 percent relative to the least exposed, a gap that began widening in mid-2024. Software development shows the pattern most acutely, with employment for young developers dropping close to 20 percent from its 2022 peak, even as demand for more experienced developers kept growing. Customer support has followed a similar path. 

Much of the decline seems to come from slower hiring, not layoffs, meaning there are fewer doors opening for people trying to get a foot in.

Those experiences should be taken seriously on their own terms, not netted out against gains elsewhere. They also make clear why leadership decisions matter so much. One-third of organizations in a McKinsey survey cited by the report expect AI to reduce their workforce in the coming year, while 43 percent expect little or no change. Leaders decide which work gets redesigned, who gets trained, and how changes are explained, and those choices largely determine whether a role evolves or simply erodes.

‍

High adoption, uneven results

The data shows a second pattern, this one inside organizations. Organizational AI adoption reached 88 percent in 2025, and generative AI is now used in at least one business function at 70 percent of organizations. But deeper integration is still rare, with scaled use of AI agents remaining in the single digits across nearly every business function..

Results are uneven as well. Controlled studies show productivity gains of 14 to 26 percent in customer support and software development, with the largest gains often going to less experienced workers. In tasks requiring more judgment, the effects are weaker and sometimes negative. In one study, developers using AI tools were actually 19 percent slower, despite believing the tools were helping.

At the economy-wide level, the evidence is earlier and more mixed. A survey of 6,000 executives found widespread adoption but minimal realized productivity gains so far. The report suggests this gap may reflect how early most organizations are in integrating AI. Economists describe this pattern as a J-curve: organizations must absorb the costs of adoption before the larger gains appear.

‍

Preparedness starts with measurement

So what does a prepared company do differently? Usually, it starts with gaining visibility. The AI Index itself notes that the data infrastructure needed to track AI’s impact is struggling to keep pace with the technology. Inside companies, the same problem shows up as leaders who know AI is being used but can’t see whether it’s working.

That points to three priorities for leadership teams:

• Measure where value is actually being created. Adoption figures and productivity results often diverge. Treating one as the other leads to misplaced confidence.
• Track the difference between adoption and results. High usage without matching performance gains signals a problem, not progress.
• Invest in people alongside the tools. A study of 12,000 European firms cited by the report found that AI adoption raised labor productivity by about 4 percent without reducing short-run employment, and that training significantly strengthened those gains. Being clear about how roles are expected to shift, and investing in the skills those shifts require, is part of the work, not an afterthought.

None of this requires a dramatic initiative or a public announcement. It requires the discipline good leadership has always applied to a new capability: understand where you stand before deciding where to go.

‍

A gap in the conversation

The report highlights a striking divide in expectations. Seventy-three percent of AI experts expect AI to have a positive effect on how people do their jobs, compared with just 23 percent of the public. That 50-point gap reflects real uncertainty among the people whose work is changing, and it won’t close through reassurance alone.

It may start to close as organizations show, with evidence, that they’re adopting AI in ways that make work better and treat people fairly along the way. That question gets answered inside individual companies, by leadership teams willing to look honestly at what’s working and what isn’t.

‍

Where this goes from here

The organizations that handle this transition well won’t necessarily be the ones that moved first or spent the most. They’ll be the ones that understood where value was actually being created, corrected course when it wasn’t, and brought their people with them. For employees and businesses alike, that kind of leadership will shape how this period is remembered.

‍