GenAI FOMO: 95 percent of AI investments yield zero returns

American companies have invested between $35 billion and $40 billion in generative artificial intelligence (GenAI) projects, yet so far, almost none of this has produced tangible results. According to a report from MIT’s NANDA (Networked Agents and Decentralized AI) project, 95 percent of organizations have not achieved any return on their AI investments, while only five percent successfully utilize AI tools in full production.

The report is based on 52 structured interviews with business leaders, an analysis of over 300 public AI projects and publications, and a survey of 153 business professionals. The authors of the report, Aditya Challapally, Chris Pease, Ramesh Raskar, and Pradyumna Chari, have termed this issue the ‘GenAI divide’ and believe it does not stem from a lack of infrastructure, knowledge, or talent, but rather from the inability of AI systems to retain data, adapt, and learn over time.

– Although AI chatbots succeed due to their simplicity and flexibility, they fail in key business processes due to a lack of memory and adaptability – states the report.

– We have seen dozens of demos this year. Perhaps one or two are truly useful. Everything else is just wrappers or scientific projects – said one unnamed CEO.

The research confirms a trend of declining trust among business leaders in GenAI projects. The chances of GenAI bringing significant changes to most sectors are slim, with exceptions being technology, media, and telecommunications, where a material impact has been observed. For other industries such as healthcare, pharmaceuticals, retail, financial services, advanced manufacturing, and energy, GenAI currently has no significant effect.

AI as Justification

One change is evident in employment within the affected sectors. In the technology and media sector, more than 80 percent of executives expect a reduction in new hiring over the next two years. Additionally, cuts have most often pertained to non-core business functions that are frequently outsourced, such as customer support, administrative processing, or standardized software development.

Examples like the recent layoffs at Oracle suggest that companies are using AI investments as a justification for cost-cutting, while some IBM employees claim that AI has been used to shift jobs overseas. According to the report, nearly 50 percent of AI budgets go to marketing and sales, and the authors recommend redirecting investments to activities that yield real business results, from qualifying potential clients and retaining customers to optimizing outsourcing and financial processes.

The report also highlights that generic tools like ChatGPT often outperform specialized enterprise tools, even when using the same AI models. The reason is simple: employees are already familiar with the ChatGPT interface and use it more, while more expensive and rigid corporate tools are less well-received.

The authors conclude that companies that successfully bridge the ‘GenAI divide’ do not view AI merely as software, but as a partnership.