The Prompt: Your agents need enterprise truth to be successful

Business leaders are buzzing about generative AI. To help you keep up with this fast-moving, transformative topic, our regular column “The Prompt” brings you observations from the field, where Google Cloud leaders are working closely with customers and partners to define the future of AI. In this edition, Suraj Poozhiyil, AI product leader, shares how AI agents depend on what we call ‘enterprise truth.’

For decades, technology has pursued automation. We've written code to perform tasks, built workflows, and created applications to streamline operations.

An agent is the evolution of that “old-world” automation where the human does less, and the system does more. That's the core of it. But this approach, while incredibly valuable, has some limitations. It defines what we do, but not why we do it. It keeps the burden on humans to provide essential context and codify their ways of working.

This is where "enterprise truth" comes in – it's what we call your enterprise’s specific data, tools, constraints, policies, and processes that the agent needs to be successful.

Enterprises need to give agents the full picture of the organization’s context so agents can make better, more informed decisions. Connecting agents to data isn’t enough. Today, I'll explore what enterprise truth means, and how paired with agents, can unlock a smarter, more responsive, and more competitive enterprise.

Agent success is rooted in enterprise truth

We’ve heard it before – AI is only as good as the data you put into it. When it comes to agents, AI agents are only as good as the context you give them. Enterprise truth is the answer to questions like, "What is our company’s policy for creating a purchase order?" and "What is the specific workflow for approvals and compliance?"

Think about it: when a human joins a company, they bring general skills. They know how to make payments and how to open a PO. They know how to write a document. But they need specific training to understand how things are done within that organization. I remember when I joined Google, I had to learn the specific ways to operate as a Googler. I did our trainings, read documents, watched videos, or went to in-person classes that taught me how to follow the specific processes and use the specific tools available to me at Google. Similarly, AI agents may have general knowledge, but "enterprise truth" provides the specificity they need to operate effectively.

This is a profound shift from traditional automation. Traditional automation codifies the process (the "what"). Enterprise truth codifies the understanding that informs the process (the "why"). This has existential implications for the enterprise – how does an enterprise truly "know" what it knows? How do we ensure that this understanding is preserved and accessible, even as the workforce changes?

Enterprise truth is already a reality

This isn’t smoke and mirrors, or faraway magic. We’ve already started this journey with you. Last year, we launched Application Integration which orchestrates workflows across your business’ data starting with a prompt or a click.

We haven’t stopped there. What’s around the corner is an evolution that leverages all of your enterprise’s documents, workflows, APIs and data. Take a handful of examples:

Agentspace bridges the gap between a customer's enterprise truth and agent action. It offers pre-built connectors to your enterprise systems, real-time information from the Internet, Google-quality search, and Gemini’s multimodal intelligence for analysis — so both employees and agents can accomplish tasks based on your enterprise truth.

Another example is Vertex AI Agent Builder. It uses your enterprise context as the foundation for building agents by pulling from all your enterprise APIs. This way, your agents can be successful within the constraints of your enterprise’s governance.

So how can you harness enterprise truth for your work?

Balancing speed, quality, and trust

As with all new technologies, businesses are just grasping the possibilities of what they can do with agents. If you went back in time to the dawn of the low code market and approached enterprises saying: "I'd love to enable your business users to codify their knowledge as business process apps,” the first question you’d hear back is: “Why would a business user build an app? They don't have time to go learn how to code. If they did build apps, how would we govern these apps to manage risk?” A similar surprise and concern exists in the industry right now.

Successful adoption of AI is about the careful triangulation of speed, quality, and trust. If you approach AI adoption with the idea that you're going to overhaul your entire business at once, you're likely to undermine trust, and what seems like initial speed will ultimately lead to significant delays.

Looking ahead

In my work, I've always been driven by a singular purpose: "How do I make regular people feel super- powered through technology?" Every product I've built has been guided by this question. And in the realm of AI, agents – and especially autonomous agents – I see the biggest possibility.

To learn more about enterprise truth, check out this ultimate guide to Vertex AI and learn more about Google Agentspace.