Dive Brief:
- Agentic AI is gaining ground across businesses but agents remain siloed, according to a report from agentic AI company Leah. IDC surveyed more than 400 enterprise tech decision-makers on behalf of the provider.
- While two-thirds of organizations said they use AI agents in production, and plan to increase them sixfold by next year, only 29% of agents interact with each other, according to the Tuesday report. Despite challenges in agentic deployment, companies plan to double their budget for agentic AI over the next year, the report found.
- Governance, interoperability and cross-functional coordination are needed to manage AI agents effectively and scalably, the report found. “Right now, enterprise AI looks a lot like a thousand instruments playing at once without a conductor,” Anurag Malik, president and CTO of Leah, said in a statement.
Dive Insight:
Without productivity or cost-savings evidence — and often without proper technical foundations in place — enterprises continue to embed AI agents deeper into their workflows.
Only 1 in 5 leaders say their organization is prepared to redesign business processes to run autonomously with AI agents, an August Deloitte study found. But nearly three-quarters of U.S. company leaders expect roughly half of their business processes will be redesigned or rebuilt around AI agents by 2030.
Without cross-functional foundations, enterprises will continue to have issues with isolation and interoperability, Malik said. Just 7% of enterprises had advanced agent orchestration, which involves multiagent collaboration, the report found. More than half were using basic, or workflow-based handoffs.
AI agents should be viewed as coordinated, purpose-built systems that work together, rather than individual productivity tools, the report said.
A desire for visibility into AI actions is playing out across the enterprise, as some companies adopt specialized tools to monitor complex technology environments. One in four AI agents run unmonitored, according to New Relic data published this week.
“The next phase of enterprise AI will depend on creating harmony among those agents while maintaining the governance and oversight necessary to build trust and deliver meaningful business value,” Malik said in the report.
Most enterprises are moving from the experimentation phase with AI agents into execution, according to Malik. The survey suggested enterprises know the opportunity that AI could present, but they struggle to recognize all that’s required for successful deployment.
“Businesses must learn to coordinate thousands of decisions, workflows and agents as a single, governed system,” Malik said.
Complex business processes do not usually belong to just one department or system, said Neil Ward-Dutton, research VP, agentic automation and AI technologies at IDC, said in a statement. Enterprises should manage agentic AI and AI agents as a portfolio, rather than by-department projects, he added.
“By standardizing how agents are built and monitored, prioritizing core processes that cross functions, and measuring results at the outcome level instead of counting deployments, these enterprises will see actual ROI, rather than just a collection of pilots,” Ward-Dutton said.
