Google adds agent control layer, and lock-in debate begins

Last week, Google held its annual Google Cloud Next customer conference in Las Vegas. There were announcements aplenty, from new TPUs to new security and data products, but perhaps the biggest was Google’s expanded Gemini Enterprise Agent Platform, which the company is positioning as a more complete orchestration and management layer for enterprise agents.

As Google described it in a blog post, the platform repackages and extends several existing pieces, including Gemini Enterprise, Vertex AI and Google Workspace capabilities, into a more unified platform. It is also adding new app‑layer and management features such as Agent Runtime, Memory Bank and Agent Studio, the latter giving developers a tool to build and refine agents. All of these components make the platform a one-stop shop for enterprise agent development and management.

I spoke to three analysts to get their take on the new platform, and while they all agreed that Google did a good thing in filling in a critical missing piece in the Google Cloud AI platform, they differed on whether the platform's Google-centric design was positive or not. We'll get into that more shortly, but it's worth pointing out that I asked Google about this a week ago after a press event. They were tied up with the conference and weren't able to respond.

Jason Anderson, an analyst at Moor Insights & Strategy who covers all things cloud and AI, told me that this was clearly a long-needed control plane for agents, and its absence had been discouraging enterprises from going all in. "They had some good tools and good systems of engagement. They also have really good infrastructure, but without the Agent Platform, there was a missing middle, so to speak, which was really impeding enterprise adoption," Anderson told FastForward.

Jim Rowan, head of AI at Deloitte, agrees that this piece is critical and long overdue. He frames it using a term we're hearing a lot lately, the AI harness, which is basically the control and orchestration layer for agents, much like Kubernetes for containers. "Robust infrastructure acts as this harness: the operational wrapper around the model handling logic, memory, tool execution and physical infrastructure the software resides on," Rowan said.

He believes that this kind of platform is critical for enterprise users building agents, not only from a management standpoint, but also in a way that could reduce costs. "By layering capabilities into the harness rather than relying on the core model for heavy lifting, enterprises theoretically reduce token consumption and compute waste," he said.

Lock-in or choice

As we noted earlier, while they all agree that this is a long-needed step, the platform appears to be designed primarily for Google customers who are bought into Google's vision and are less concerned about vendor lock-in. Holger Mueller, an analyst at Constellation Research, dismisses lock-in concerns. To him, companies want an integrated solution more than they want flexibility.

"Of course the lock-in argument is there. But in the long run, suites always win. Thomas Kurian has delivered the complete TPU-to-agent stack from one vendor, vertically and horizontally integrated," Mueller said.

But Rowan sees this as more of a move to cross-platform integration that started with the Agent2Agent protocol announcement at last year's Google Next. "The evolution of AI agents and how they interact across AI ecosystems is one of the most integral parts of agentic workflow success," he said. "Rather than remaining a walled-garden, the Gemini Agent Platform is actively bridging ecosystems. [Last week's] announcements regarding new integrations with outside agents demonstrate that cross-vendor collaboration is happening right now, with Gemini acting as a central hub."

Anderson has a different viewpoint. He believes that most of these kinds of solutions tend to favor the vendor, and this one is no exception. "But in terms of "openness" I'd say that while most of these platforms are biased, Google's may be slightly more so right now," he said.

Google has finally filled in a crucial piece in the agent stack. The question is whether that comes at the cost of vendor lock‑in. Companies will have to decide how much they value flexibility versus the convenience of the suite. The success of this platform will depend on how they answer that.

This article builds on a news brief originally published in FastForward #62.