AI came a long way in 2026

getty

Data – the essential fuel of AI – may be running short. a new report out of Stanford University warns. AI and large language model adoption is soaring, yet it’s possible that available real data for AI training models may be depleted within the next six years.

That’s the word from Stanford University’s 2026 AI Index Report, which looks at major developments in AI over the past year. Overall, the report’s authors see mixed indicators on the impact of AI on jobs and productivity.

Alarm over the impending data shortage was raised in last year’s report, and there has been no change in circumstances for the current report, which aggregates leading industry and academic studies. “AI researchers have publicly claimed that the available pool of high-quality human text and web data for training large models has been exhausted, a state often referred to as ‘peak data,’" the Stanford co-authors relayed. "This has continued to raise industry-wide concerns about the sustainability of scaling laws, which have historically depended on ever-larger datasets."

Employing synthetic data may not fully resolve data shortages and even be a drag on AI performance, they add. “Limits on the availability of real-world data may be less consequential if synthetic data – data generated by AI systems – can be used to improve the performance of subsequent models,” the co-authors state. The risk is synthetic data may not adequately support AI, except for more narrow use cases. Real data still needs to remain part of training sets. However, “simply adding more data does not necessarily lead to performance gain. There is still no definitive evidence that synthetic data can fully offset real-data depletion in pre-training contexts. ”

The Productivity Question

The study’s authors also question AI’s productivity impact to date. Productivity gains “are measurable within narrow tasks, but the evidence at the macro level remains early and mixed,” they state. “The AI economy is scaling quickly, but how widely and how fairly that growth translates into real economic value is still an open question.”

Productivity gains are real but uneven: AI shows 14 to 26% improvements in customer support and software development, but is weak when it comes to tasks requiring business judgement. Productivity gains are seen in tasks that are “structured, language heavy, or supported by clear feedback loops,” the study shows. “In others, gains are marginal or even negative when tools are poorly matched to the task.”

Productivity gains from AI are appearing in fields where entry-level employment is starting to decline, suggesting a relationship between AI adoption and workforce reduction. “Studies show productivity gains of 14% to 26% in customer support and software development, with weaker or negative effects in tasks requiring more judgment.”

The Jobs Question

TAI may be eating into entry-level jobs, but the data is not so clear on the trend or where it’s heading. “Employment among the youngest workers in AI-exposed occupations has already declined since 2024, even as headcount for older workers keeps growing,” the report states. “One-third of organizations expect AI to reduce their workforce in the coming year, even though broad displacement hasn’t yet appeared in aggregate labor data."

Anticipated reductions due to AI displacement are highest in service operations, supply chain, and software engineering, the report continues. Still, they concede, "it is challenging to measure AI’s impact on employment, particularly because the technology is still in the early stages of widespread deployment. So far, effects on the workforce appear to be uneven, initially showing up in hiring pipelines, among younger workers, and within specific business functions. The evidence does not point to broad, uniform displacement."

AI’s impact on employee headcounts is an open question, but the technology is fueling a boom in entrepreneurial activity. “The U.S. led in entrepreneurial activity with 1,953 newly funded AI companies in 2025, more than 10 times the next closest country."

The Agents Question

TAI agents – the latest variation of AI being promoted across markets – deployment remains in single digits across nearly all business functions, the co-authors report. “AI capabilities are advancing rapidly in the lab, but real-world reliability consistently falls short of what benchmarks. 2025 was declared the year of AI agents, but models still struggle to reliably execute multi-step tasks, with performance on agent benchmarks remaining well under 50%."