Artificial intelligence has rapidly become part of the language of business. Executives talk about AI as a source of productivity, innovation, efficiency, and competitive advantage. But as organizations race to demonstrate that they are embracing AI, there is another question leaders should be asking: what is happening to their people?
In a recent study with my co-author Diego Ravenda, we analyzed the prepared remarks of 71,433 quarterly earnings calls from 4,327 U.S.-listed companies between 2019 and 2024 — the very period when AI moved from a niche topic to a fixture of corporate discourse. Using machine-learning tools to identify AI-related and employee-related paragraphs and to measure the tone in which managers discuss them, we tracked how the volume and positivity of AI talk relates to companies’ subsequent workforce and social performance, as captured by independent ESG ratings.
AI discourse was still relatively rare through 2022: in more than half of the firm-years in our sample, AI was not mentioned at all in the prepared remarks. That changed abruptly from 2023 onward, when AI-related language overtook employee-related language in earnings calls for the first time in our six-year window.
The results reveal an uncomfortable pattern. Companies that talk more intensively, and especially more enthusiastically, about AI tend to see lower subsequent workforce and social performance, even after accounting for firm size, profitability, and other differences across companies and time. Firms whose earnings calls fall in the top 10% of AI intensity within their industry see their workforce scores — which capture issues such as employee satisfaction, health and safety, diversity, and development opportunities — decline by roughly 2 points in the year of the announcement, and by nearly 5 points within two years. The pattern survives statistical tests designed to rule out reverse causality, such as instrumenting a firm’s AI talk with what its industry peers were saying a year earlier.
But perhaps the most interesting finding is what happens at the moment companies announce major AI initiatives. At precisely the same time that managers increase their emphasis on AI, they also become more positive in the way they talk about employees — highlighting training, well-being, culture, and other people-related initiatives. Yet the subsequent workforce indicators move in the opposite direction. In other words, the rhetoric becomes more employee-friendly just as the outcomes become less so.
Interestingly, this rhetorical cushioning is weakest exactly where AI-related disruption tends to be greatest. In technology companies, an AI announcement is not accompanied by any meaningful increase in employee-friendly language — arguably because investors already expect tech firms to talk about AI, so there is less to gain, in legitimacy terms, from reassuring rhetoric. Yet workforce scores fall roughly twice as much in tech firms as elsewhere. The sector most associated with cutting-edge AI adoption is also the one where the gap between narrative and outcome is starkest.
To test whether this decline reflects something more than a change in how companies report on their workforce, we also examined a harder measure: announced layoffs relative to total headcount. Here, too, the pattern is unmistakable. In the year a company delivers an unusually AI-intensive earnings call, announced layoffs rise significantly, and the effect does not fade — if anything, it grows over the following years. This suggests that an AI-heavy earnings call is often not a one-off announcement but the opening chapter of a longer restructuring story.
We describe this pattern as consistent with what we call “AI-washing”: organizations pairing enthusiastic messages about technological transformation with reassuring messages about their commitment to employees, without those messages being matched by the workforce outcomes that follow. This does not necessarily mean that managers are deliberately misleading stakeholders. Leaders may genuinely intend to protect their people and respond by publicly emphasizing that commitment. But intentions and communication are not the same as outcomes — and our evidence suggests that, on average, they diverge.
For leaders, the implications are important. Responsible AI should not start with the communication strategy around AI; it should start with how AI is implemented. When announcing major AI initiatives, leaders should ask three questions. What will this technology actually change for our employees? What are we concretely doing to manage those consequences? And what evidence will let us demonstrate, a year or two down the line, that our commitments were delivered? This means complementing investment in technology with investment in people: reskilling employees whose tasks are changing, redesigning jobs rather than simply eliminating them, monitoring the effects of AI on employee well-being and trust, and establishing clear safeguards around issues such as algorithmic monitoring and surveillance. It also means tracking hard workforce indicators rather than relying on statements of intent.
There is also a lesson for investors, boards, and other stakeholders. When a company enthusiastically announces its latest AI initiative and, in the same breath, emphasizes how much it cares about its people, that pairing itself is worth noticing. But it is worth looking beyond the narrative: commitments to training, well-being, inclusion, or responsible AI matter most when they are backed by measurable outcomes, not simply by the tone of the next earnings call.
AI can undoubtedly create enormous value for organizations. The challenge for responsible leaders is to ensure that some of that value is also captured by the people whose work AI is transforming. The question should therefore not simply be “How much are we investing in AI?” It should also be “How much are we investing in the people affected by it?”
Have a Great and Impactful Week!
João Cotter Salvado, Academic Director of the Católica-Lisbon Entrepreneurship Center