Investing.com -- Barclays (LON:BARC) said in a note to clients Tuesday that it has observed a decline in the deployment of AI projects and in those showing "meaningful" ROI, with both metrics hovering around 45-50%, according to the firm's latest AI survey.
"The percentage of AI projects that are deployed is falling, and of those deployed projects, fewer are showing 'meaningful' ROI on average," the bank wrote.
Despite enthusiasm for AI's long-term potential, the bank said large enterprises show limited urgency in expanding their deployments, largely due to high costs and early-stage ROI challenges.
Enterprises are said to be cautious about investing heavily in generative AI, as the average preparation cost, including data cleanup and training, is estimated at $75 million.
Barclays explains that many organizations are waiting for "best-of-breed" AI vendors to emerge before they invest.
Challenges like data security, handling structured and unstructured data, and limited experience with large language models (LLMs) further slow adoption.
However, Barclays notes that AI budgets are increasingly separated from general IT budgets, with 46% of respondents planning budget increases and 15% reallocating funds from software budgets.
Barclays also sees a 1-10% anticipated growth in generative AI investments over the medium term (2-5 years).
On a brighter note, AI awareness and usage continue to grow, particularly among small and mid-sized companies, according to the bank.
Key applications, such as customer service and productivity enhancements, remain top priorities.
Still, bottlenecks in sourcing and cleaning data reportedly increased 10 percentage points year-over-year to 48%, presenting a significant hurdle.
Barclays' findings suggest cautious optimism for AI adoption, contingent on overcoming cost and deployment challenges.