Enterprises hoping to extract real value from AI and deploy it successfully should concentrate on the practical, systematic governance required to build AI workflows grounded in trust, transparency and interoperability, according to a Google Cloud executive.

The constant stream of new AI models arriving each week can make the field feel like one long publicity cycle — a confusing landscape for organizations trying to work out what they actually need to make their AI workflows effective. To move beyond the promise of AI models, tools and agents, businesses should avoid placing unrealistic expectations on the technology and instead establish the right foundation for building practical AI applications based on trust, transparency and interoperability.

"Interoperability and openness are two really key concepts, and transparency as well," said Michael Clark, director of product management at Google Cloud, on the Targeting AI podcast from AI Business.

The emphasis on openness reflects a broader industry shift: as enterprises move from piloting individual AI tools to deploying multi-agent systems, vendors including Google have backed open interoperability standards such as the Agent2Agent protocol, which Google introduced in 2025 with dozens of partner companies to let AI agents from different vendors exchange information.

According to Clark, systems that can connect with other systems are central to the success of generative and agentic AI applications. For an AI agent to succeed in a customer service setting, for example, enterprises need interoperability with disparate systems.

"In a lot of call centers, people do things very manually, not using APIs or not using other things when they're solving customer problems," Clark said. "So, it meant having either agents or having other context about what's happening in other parts of the business and being able to take actions."

Context can also be a key differentiator between an agentic application that works and one that does not, Clark added.

"We see context being audio and video, even in some cases, things like lidar and other kinds of information, financial streaming data and other kinds of things to help agents make decisions daily," he said.

Beyond interoperability and context, enterprises also need governance and trust — a theme that has grown more pressing as regulators in the European Union and elsewhere introduce AI rules that place obligations on companies deploying high-risk AI systems.

"As a business, understand your outcomes really clearly and understand your risk tolerance," Clark said. "Once you've identified that risk, putting a human in the loop for anything that's above that."