Manila, Philippines – While AI agent adoption has become widespread among enterprises, confidence in the technology’s decision-making remains relatively low, with integration capabilities emerging as a key factor in building trust, according to a new study commissioned by Boomi and conducted by Forrester Consulting.
The survey found that 86% of organisations have progressed beyond the AI agent pilot stage. However, only 34% said they trust the actions their AI agents are taking.
The research also identified a divide between organisations categorised as having “agentic control”—those with stronger operational readiness across governance, integration, API and Model Context Protocol (MCP) management—and those experiencing “agentic chaos,” defined as the bottom quartile in operational readiness.
According to the study, 77% of organisations in the latter group are moving AI agents into production despite lower readiness, resulting in an estimated average of US$2.1 million in additional costs stemming from compliance penalties, customer losses, operational downtime, and rework.
Organisations with agentic control reported higher confidence in their AI agents, with 55% expressing trust in the actions and decisions made by their agents, compared with 22% among organisations classified as being in agentic chaos.
“This research confirms what we’re seeing everywhere: the trust problem with agentic AI is really a data problem,” said Steve Lucas, Chairman and CEO at Boomi. “Agents can only be trusted to act on data that’s been properly activated, connected, and governed, and most companies deployed agents before they did that work. The ones who did it first are the ones getting real value now.”
According to the report, integration infrastructure plays a significant role in enterprise confidence. Organisations with agentic control were three times more likely than those in agentic chaos to consider reliable and well-managed APIs as a determining factor before piloting AI use cases.
The adoption of integration platform as a service (iPaaS) also differed between the two groups. The study found that 46% of organisations with agentic control use iPaaS to support AI agent workflows, compared with 25% of organisations in agentic chaos. Meanwhile, 86% of organisations with agentic control viewed iPaaS and API management capabilities for building AI agents as important to operational readiness, versus 58% among organisations in agentic chaos.
The report also highlighted differing priorities between the two groups. Nearly half (47%) of organisations with agentic control identified improving integration with tools, APIs and applications as a key focus area, compared with 31% of organisations in agentic chaos, which remained more focused on improving AI models themselves.
As enterprises deploy larger numbers of AI agents—some operating as many as 200—the study pointed to governance as another differentiator. It found that 46% of organisations with agentic control had established centralised governance for MCP, compared with 32% among organisations in agentic chaos. Additionally, 48% of organisations with agentic control had aligned their AI and integration teams under a unified operating model, versus 37% of those in the lower-readiness group.
The research also examined business outcomes associated with stronger operational readiness. Among organisations with agentic control, 59% reported productivity improvements from AI agent deployments, while 51% cited increased innovation. Another 46% said they had developed reusable capabilities applicable across the business, and 45% reported automating repetitive tasks.
Based on the findings, Forrester recommended that organisations strengthen collaboration between AI and integration teams, establish governance through a centralised control plane for AI agents, and implement orchestration capabilities to enable agents to securely connect with enterprise data, applications and one another.
