US technology stocks fell sharply at the close of trading yesterday, with the NASDAQ Composite index down 1.4%. Among the biggest movers were Palantir, which dropped 9.4%, and Arm Holdings, down 5%. According to the Financial Times, it was the largest one-day fall since early August.
Some traders linked the decline to a research paper released by NANDA, an AI company that originated at the MIT Media Lab. The paper, now behind a survey wall, reported that only 5% of generative AI pilot projects reach full production and generate measurable business value.
Findings from the NANDA Study
- Research was based on 52 structured interviews, analysis of 300+ public AI initiatives, and a survey of 153 company leaders.
- The majority of generative AI projects had little impact on company profit and loss metrics.
- Most successful implementations were found in back-office workflows, such as reducing reliance on third-party agencies and BPOs.
- Front-office deployments, including customer-facing functions, showed lower measurable returns.
- Internal job reductions were minimal, despite automation claims.
Adoption Trends
- Around 40% of surveyed companies pay for LLM subscriptions.
- 90% of staff reported personal benefits from tools like ChatGPT, but these gains rarely translated into measurable enterprise-level value.
- Sales and marketing are the most common functions for AI pilots, while finance and procurement remain the least.
- Complex tasks, such as client management, are rarely trusted to AI, while routine tasks like report summarisation are more accepted.
Key barriers to success included:
- Poor contextual awareness of AI models.
- Lack of memory and adaptability.
- Excessive manual input required for every task.
Between 60% and 70% of respondents agreed with complaints such as “The AI doesn’t learn from our feedback” and “Too much manual context required.”
Industry Impact
Sectors benefiting most from generative AI included media and telecom, professional services, healthcare, retail, and financial services. Energy and materials showed negligible adoption rates.
However, the report’s language and conclusions raised concerns. It strongly promotes partnerships with vendors capable of providing adaptive, learning-capable AI systems—a service that NANDA itself offers. This has led some analysts to question the report’s neutrality, suggesting it may serve as much as marketing material as academic research.
Market Reaction
The report’s publication coincided with falling AI stock prices, though whether the drop was directly caused by NANDA’s findings is unclear. More likely, it reflects broader investor concerns about the practical effectiveness and commercial value of generative AI tools.
