Dive Brief:

• Most U.S. companies (65%) are experimenting with artificial intelligence agents — tools designed to perform workplace tasks in place of humans, up from 37% in the fourth quarter of last year, a KPMG study released Wednesday found.
• The Big Four accounting firm also found that generative AI budgets are expanding, with surveyed organizations planning to spend an average of nearly $114 million on the technology during the next 12 months, up from $89 million in the prior quarterly survey.
• Risk management is expected to be the biggest obstacle to companies’ generative AI strategies this year, according to the research.

Dive Insight:

The study’s findings come as CFOs and other business leaders are weighing the potential risks and rewards of AI agents as they sort through a flood of offerings from tech vendors, including the likes of Microsoft, Oracle, SAP and Workday.

The technology is still in the early stages of development, and analysts previously told CFO Dive it may be too soon for rapid adoption when it comes to functions such as corporate finance. Among other potential challenges, an agent could leak sensitive data or the technology might “hallucinate” and provide erroneous information. 

KPMG’s latest survey found that businesses piloting and deploying AI agents are using them in various ways, including for analysis of complex data sets (78%), administrative duties (66%), call centers (61%) and recruitment and sourcing of new employee candidates (26%).

Despite the rise in agents that are being explored and tested by companies, full-scale depoyment of the technology remained flat at 11%. 

More than eight in 10 respondents expected risk management to be the top hurdle to their generative AI strategy throughout the remainder of the year. Data privacy and security are now the most important decision points when choosing a generative AI provider, increasing from 43% to 73% since the fourth quarter, the research found.

Organizations also face hurdles in preparing their workforce for the technology, according to the study. More than 60% of respondents said they anticipate challenges in training employees to work with agents due to the potential complexity of the technology.

“[A]s the AI landscape evolves, organizations must overcome critical barriers in data infrastructure and workforce preparation to demonstrate measurable business outcomes,” KPMG said in an emailed press release.

The research findings also comes on the heels of a study by cybersecurity firm Delinea finding that digital entities such as bots and AI agents are fast emerging as prime targets for cyberattacks as organizations rapidly increase their reliance on them.

“While human identities remain a primary attack target, non-human identities (NHI) have quietly become an equally critical — and often overlooked — security risk,” a report on the Delinea findings said.