SINGAPORE — Enterprise technology company Boomi has introduced an Agent Control Plane designed to give organizations greater control over artificial intelligence agents, including how they access business systems, use data and consume computing resources.
Boomi announced the platform innovations as part of its efforts to address governance, security, cost and data-sovereignty concerns that are emerging as enterprises move AI agents from experimentation into production. The Agent Control Plane is described as AI-native infrastructure that connects agents to core business systems, governs their activities and provides controls over AI-related spending.
The company said the control plane is vendor- and model-neutral, allowing organizations to manage agents, tools and models across public cloud, private cloud and on-premise environments. It also supports bring-your-own-model deployments and specialized small language models.
The Agent Control Plane is not a standalone new product but a solution powered by the Boomi Enterprise Platform. It brings together capabilities for connecting, building, contextualizing, governing and running AI agents.
Boomi said the system is intended to address a growing governance gap as enterprises deploy agents that can interact directly with operational systems such as Salesforce, SAP, Oracle and Workday. The company said unrestricted access could expose organizations to security, compliance and financial risks.
The platform provides a centralized control point for agents and tools, with the Boomi AI Gateway serving as an enforcement layer. According to the company, the gateway can inspect traffic, apply identity and rate limits, manage token consumption and require human approval for high-risk transactions.
Boomi also highlighted its ability to connect agents to enterprise systems through Boomi Connect, which provides more than 1,000 pre-built Model Context Protocol tools and servers. Existing Boomi integration processes can be exposed to agents as governed tools without requiring organizations to rebuild their systems.
The company’s release cited Gartner’s prediction that 40 percent of enterprises will demote or decommission autonomous AI agents by 2027 because of governance gaps identified after production incidents. It also cited a 2026 FinOps Foundation survey in which 98 percent of respondents said they were managing AI spending, up from 31 percent two years earlier.
Boomi said its hybrid runtime architecture allows agents, tools and models to operate within private networks and regional boundaries. The approach is intended to help organizations comply with data-sovereignty requirements, keep sensitive intellectual property behind corporate firewalls and manage computing costs.
The issue of where AI workloads should run also emerged during Boomi’s World Forum regional media event in Singapore on Sept. 10, where executives said on-premise infrastructure remains relevant despite the broader shift toward cloud computing.
Amy McNee, Boomi senior vice president for Global Solution Engineering and general manager for Data Management GTM, said organizations are increasingly taking a more selective approach to infrastructure.
“It’s not fully away from cloud, it’s really more of a rationalization of what belongs in the cloud versus what is better managed on-premise,” McNee said.
She said regulatory and compliance requirements and cost were among the factors driving demand for hybrid deployments. McNee cited financial services companies that have retained infrastructure in their own environments to manage costs or because AI workloads containing personally identifiable information cannot be run in the cloud.
“And so they have to have a hybrid strategy. And we’re seeing more and more of that, where people want choice. They want to be able to choose to put some things in one area and the other, and they need tools that can cross over both of those,” McNee said.
Asked by The Manila Times whether data sovereignty was also contributing to the continued demand for on-premise systems, McNee said the factor was becoming more significant amid geopolitical developments.
“I think a lot of the geopolitical things that we’re seeing right now are driving it faster now. A couple years ago, the regulatory and compliance and cost were driving it, but we definitely see more sensitivity to it now,” she said.
David Irecki, Boomi chief technology officer for Asia-Pacific and Japan, said organizations are evaluating infrastructure according to the sensitivity of their systems and data rather than adopting a single deployment model.
Using Ampol, Australia’s largest transport energy distributor and retailer as an example, Irecki described a three-tier approach involving on-premise systems, private cloud infrastructure and software-as-a-service data running on Microsoft Azure.
“They look at what their the systems are, what the data is, what the use cases are, and they have a three-tier model,” Irecki said. “So they have some systems on-premise, and integration is done on-premise because of the sensitivity of the data they have.”
Irecki said the same flexibility is being extended to AI workloads, including smaller and open-weight models.
“They want to have agents on-premise, working wrapped around their SLMs, as an example, or their open-weight models. They may want them running in a in a private cloud or equally in a public cloud as well,” Irecki said.
He also linked the deployment question to the value of enterprise data.
“Organizations are starting to understand the data behind their four walls. It’s their competitive advantage against everyone,” Irecki said.
Beyond deployment location, Boomi’s presentation emphasized the importance of governing the context available to AI agents. McNee said governance extends beyond conventional access and security controls to include the business definitions and data provided to agents.
“Context is governance, right?” she said. “So if your agent doesn’t have the right context, they’re not operating in the right way, so that’s also a part of it.”
Irecki said governance should be established before AI systems are deployed rather than added afterward.
“When you’re looking at the AI discussion, you know, all of those concepts around policy enforcement, governance, etc., apply. But you don’t want to bolt that on as an afterthought. You want to have governance in place first before you go down that journey,” Irecki said.
Boomi is also positioning agentic workflows as an alternative to placing entire business processes inside autonomous agents. McNee said complex processes can be divided between probabilistic AI tasks and deterministic steps, with human oversight incorporated where necessary.
The broader approach reflects Boomi’s push to position enterprise AI as an infrastructure and governance problem as much as a model or application problem. Its Agent Control Plane is designed to allow organizations to deploy agents across different clouds, models and infrastructure while maintaining centralized visibility, policy enforcement and auditability.