AGI Society AI and Employment Claims evaluation

The AGI Society has an educational commitment to provide reliable insights into AI’s future, as well as evaluating claims about AGI and AI broadly based on reliable, well sourced independent data. This special report evaluates notable public claims made between January 2023 to September 2026 about AI’s impact on employment. Notably, while this report does not evaluate a specific AGI claim, the topic of impacts on the labor force from advanced narrow AI overlap considerably enough to be relevant to the discussion of AGI progress.

Conclusion 1: Public claims of AI impacts on labor are not supported by evidence

Technology executives, like Marc Benioff, Drew Houston, Sam Altman and Arvind Krishna warned of AI replacing jobs as early as 2023, but these forecasts lacked measurable outcomes and more recent data reveal that these claims were either misleading or were later reversed. Both OpenAI CEO Sam Altman and Salesforce’s CEO, Marc Benioff were early to claim AI tools would reduce engineering headcount by up to 30%, yet subsequent data showed few actual layoffs or sustained hiring freezes. Drew Houston cited AI as a reason for 16% workforce cuts at Dropbox, but later reporting acknowledged that AI was one of many factors, not a direct cause. These examples highlight that in most cases, speculative claims lack any credible evidence whatsoever, and when compared over multiple years, the introduction of AI technologies has not demonstrated a lasting reduction in workforce or elimination of roles.

Conclusion 2: AI impacts are credibly concentrated in changing rather than displacing work

AI technologies can be credibly demonstrated to have reshaped tasks but claims of job elimination have no credible data backing them. IBM’s CEO, Arvind Krishna (2023), said 30% of back-office roles were going to be replaced with AI, while increasing hiring in software and sales. The 2026 Stanford/ADP study found that workers aged 22–25 in AI-exposed roles faced a 19% employment gap compared to peers, suggesting AI is altering skill needs rather than removing jobs outright. These shifts align with the OECD’s 2023 report, which noted that 27% of jobs are at high automation risk, but labor demand has not collapsed or measurably reduced.

Conclusion 3: Early-career workers face the largest impacts from AI

AI’s impact is uneven, with young workers disproportionately affected. The 2026 PwC report found that AI-exposed jobs increasingly require senior-level skills, creating a “skill gap” for entry-level roles across both IT and non-IT related roles. This mirrors the 2026 Stanford/ADP study, which showed a 19% employment gap for workers aged 22–25 in AI-related fields. These findings suggest AI are likely shifting opportunities and requirements for new entrants while also shaping role expectations for experienced professionals rather than eliminating roles.

Conclusion 4: Most AI augmentations are around existing tasks, not new ones

AI tools have demonstrably changed workers productivity, but have not demonstrated sustained replacement of tasks or roles outright. A 2023 MIT study found that ChatGPT reduced writing task completion time by 40% for professionals, while a Harvard/BCG experiment (2023) showed AI boosted productivity for consultants by 25%. These results align with the 2026 PwC report, which noted that AI-exposed jobs saw 40% higher productivity growth. However, these gains rarely translate to fewer workers and instead raise expectations for role output.

Conclusion 5: Companies have reversed experiments with AI based labor reductions

Klarna (2025) initially cut 16% of its workforce claiming AI automation would fulfill the gap, but later reversed the decision, citing customer satisfaction concerns. Meta (2026) scaled back its claimed AI-induced workforce transformation due to employee resistance and lack of human-level capabilities. In September 2026, Apple was reported to have considered eliminating approximately 5,000 AppleCare customer support positions based on the expectation that AI-powered phone and web agents could assume portions of their work. Despite already deploying generative AI into customer support, Apple subsequently placed the workforce reduction plan on hold indefinitely and was no longer pursuing the cuts. These cases show how anticipated labor reductions attributed to AI frequently fail to survive actual deployment, highlighting the expansive gap between forecasts of workforce replacement and on-the-ground company operations.

While CEOs and companies make bold claims, the data reveals a nuanced reality between those claims and evidence of the impacts on labor markets from the introduction of AI technologies. The evidence suggests AI’s impact on labor is not aligned with public statements and actions from relevant company heads. AI technologies have proven they reliably reshape role-specific tasks across a variety of technical and non-technical roles, increase pressure on early-career workers to adapt quickly, and boost productivity for senior workers, but offer no evidence that they are the primary cause for role elimination or sustained workforce reductions.

The AGI Society concludes that the available data does not support claims that adoption of generative AI has induced market-wide hiring reductions. Evidence suggests that workforce reductions attributed to anticipated AI capabilities have repeatedly been reversed, scaled back, or canceled outright. Apple’s September 2026 decision to suspend a reported plan affecting approximately 5,000 AppleCare employees is particularly notable because the company had already deployed generative AI customer-support agents and nevertheless declined to proceed with the anticipated labor reduction. This indicates that public statements and projections about AI-driven employment reduction should not be treated as reliable evidence of AI capabilities or AGI progress without independently measurable labor outcomes.