Most AI Agents Not Yet Autonomous, but Simpler Solutions Deliver Good Value, ISG Study Shows

Despite high expectations for agentic AI to autonomously perform complex functions, nearly half of agentic production systems today can only trigger predefined actions, according to a new research report from Information Services Group (ISG) (Nasdaq: III), a global AI-focused technology research and advisory firm.

ISG will present the findings of its ISG State of Agentic AI Market report today at the 2025 ISG AI Impact Summit in Boston. The report, based on surveys of enterprises, service providers and software companies, analyzes the most recent developments in agentic AI capabilities to uncover what’s currently being delivered, what’s on the horizon and the key challenges ahead.

ISG finds that 43 percent of agentic systems in production today are limited to triggering predefined actions only under a restricted range of conditions. Moreover, while there is growing interest in AI agents that can operate on their own, most agentic AI solutions in use today have not removed humans from the loop, the report says. Only 25 percent of current solutions allow agents to operate independently, while 45 percent position AI as advisors supporting human decision-makers or requiring human review and approval for key workflows. Still, these systems are capable of handling challenges that previous generations of technology could not address, the report says.

Agentic AI is the next evolution of AI, introducing systems that can make complex decisions and pursue goals autonomously. Building on generative AI’s ability to mine knowledge, agentic AI will enable enterprises to harness the power of AI without direct human prompting. However, advanced applications are still emerging, and most enterprises need to make significant changes before successfully deploying the technology.

“While agentic AI technology is already capable of delivering enterprise value — and continues to mature, unlocking even greater potential — most enterprises are unable to realize that value due to challenges with data readiness and organizational requirements,” said Olga Kupriyanova, principal consultant, ISG AI Advisory and co-author of the report.

More than half of organizations struggle with legacy data, and 38 percent of companies engage service providers to improve data management, ISG says. Providers are using agentic AI to help clients meet their data challenges by bringing together disparate data sources. Adding to the challenge, AI does not consume data in the same way as humans or as other data-driven technologies, so enterprises need to change the basic data architecture widely accepted as the standard today.