AI stocks took a hit on Wall Street yesterday, triggering a broader tech sell-off. The NASDAQ Composite slumped 1.4%, with notable declines from Palantir, which shed 9.4%, and Arm Holdings, down 5%. According to the Financial Times, this marks the steepest single-day market drop since early August.
While market jitters are common, some analysts attribute this particular downturn to a newly released report [PDF] from AI firm NANDA. The report, originating from the MIT Media Lab, paints a less-than-rosy picture of generative AI adoption, highlighting a surprisingly high failure rate in commercial deployments.
NANDA’s research indicates that a mere 5% of generative AI pilot projects actually make it to production and deliver tangible financial returns. The vast majority, according to their findings, fail to significantly impact a company’s bottom line. The analysis encompasses 52 in-depth interviews with enterprise decision-makers, assessments of over 300 publicly announced AI initiatives, and a survey of 153 company executives. The study focused on ROI within the first six months after a project exited the pilot phase.
Interestingly, the report suggests that while AI often gets deployed in high-profile, customer-facing roles, the most successful applications tend to be in the unglamorous world of back-office workflows. These operational efficiencies, driven largely by reduced reliance on third-party agencies and BPOs, are where the real cost savings are being realized. The survey also found that AI implementation had minimal impact on overall internal staffing levels.
Despite 90% of employees reporting personal benefits from using publicly available AI tools like ChatGPT, these individual productivity gains aren’t necessarily translating into institutional-level benefits. Around 40% of surveyed companies are paying for subscriptions to large language models.
A recurring theme among failed projects was generative AI’s “contextual awareness” problem – the inability to adapt to evolving circumstances, “remember” past interactions, and learn from feedback. NANDA argues that partnering with a vendor capable of providing a learning-capable, deeply integrated AI system, tailored to a company’s specific needs, is crucial for success. The report cites interview feedback, with 60-70% of respondents echoing sentiments like “[The AI system] doesn’t learn from our feedback” and “Too much manual context required each time.”
The media & telecom sector appears to be reaping the most benefits from generative AI, followed by professional services, healthcare & pharma, consumer & retail, and financial services. Energy & materials, however, have seen negligible generative AI project launches. Sales & marketing are currently the most popular departments for AI experimentation, while finance & procurement lag behind.
Echoing this same sentiment, AI is more likely to be deployed for Sales & Marketing purposes than for Finance and procurement. When it comes to task assignments for generative AI models, routine tasks such as summarizing a report or writing an email were more likely to be automated, whereas complex tasks such as client management were rarely assigned.
However, the report’s marketing-driven undertones cannot be ignored. The paper’s authors urge for strategic partnerships with a knowledgeable vendor to increase the chances of generative AI projects’ success, a partnership which NANDA is positioned to capitalize on. There are “unprecedented opportunities for vendors who can deliver learning-capable, deeply integrated AI systems,” the paper’s conclusions state.
The headlines may have spooked some companies tasked with generative AI implementations, yet the claims within the report should be viewed with healthy skepticism, given the source’s vested interests. While this week’s stock performance may be partially influenced by the NANDA publication, it’s just as likely that the findings simply validate pre-existing concerns about the practical effectiveness of generative AI as a genuine business tool. In other words, is this just a self-serving prophecy, or a legitimate warning sign for the AI hype cycle?
Original article, Author: Samuel Thompson. If you wish to reprint this article, please indicate the source:https://aicnbc.com/7740.html