Artificial intelligence (AI) agents are quickly becoming part of enterprise operations, but most organizations still do not trust the technology enough to let it make business decisions on its own.
Research by Boomi, the data activation company for AI, found that 86% of enterprises have already moved beyond testing AI agents and are using them in real business operations. However, only 34% said they trust the decisions and actions their AI agents are making.
“This research confirms what we’re seeing everywhere: the trust problem with agentic AI is really a data problem,” said Steve Lucas, chair 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.”
The study, conducted by Forrester Consulting for Boomi, surveyed 409 IT and technology decision-makers across North America, Europe, and Asia-Pacific.
The findings suggest that many organizations are deploying AI agents before putting the necessary data, governance, and integration systems in place. According to the study, companies with weak AI controls face an average of $2.1 million in added costs from compliance penalties, customer losses, system downtime, and rework.
The research found that companies with strong AI governance were far more confident in their AI agents. About 55% said they trust their agents’ decisions, compared with only 22% among organizations with poor operational readiness.
The biggest difference between the two groups was system integration. Companies with better AI governance were nearly twice as likely to use integration platform as a service (iPaaS) to connect AI agents with business applications and data. They also placed greater importance on API management, which allows AI agents to securely interact with enterprise systems.
Forrester noted that AI models alone are not enough to deliver business value. Without connections to business systems, AI agents remain limited, comparing the technology to a brain that needs limbs to take action.
The study also found that some organizations are already managing as many as 200 AI agents. Without centralized oversight, however, these deployments can create what Forrester described as “pilot purgatory,” where AI projects fail to move beyond experimentation because they cannot securely connect with enterprise systems.
Organizations with stronger AI controls reported better business outcomes. Nearly six in 10 (59%) said AI agents improved productivity, while 51% reported greater innovation. Others said AI helped automate repetitive work (45%) and created reusable capabilities across different parts of the business (46%).
For enterprises in the Philippines accelerating AI adoption, the findings highlight that successful AI projects depend not only on the AI models themselves but also on secure data management, system integration, and governance.