Enterprise AI agents promise to simplify everyday work, but delivering that experience requires far more than embedding AI into existing enterprise applications.
Enterprise AI has expanded well beyond chatbots and copilots into workflows that affect hiring, finance and other business operations. As organizations move AI deeper into those processes, they also need ways to preserve the governance and accountability that enterprise systems already provide, according to Gabe Monroy (pictured), chief technology officer at Workday Inc.
“It’s really around how do you take the magic of these probabilistic systems and mirror it with the deterministic guarantees that you get out of an [enterprise resource planning] system like a Workday?” he said. “You can get the guarantees of people and money and the rules that are in such a system, but you can get the probabilistic reasoning that is provided by AI systems and those two things working together.”
Monroy spoke with theCUBE’s John Furrier during the Google Cloud: AI Agents in Action Series on theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed how Workday and Google Cloud are integrating AI into enterprise workflows while preserving the guardrails organizations require for mission-critical business processes. The segment included live demonstrations that showed how the partnership is designed to simplify everyday work while keeping AI agents aligned with enterprise policies and systems of record. (* Disclosure below.)
Rethinking the enterprise workflow
Many organizations already rely on Google Workspace as the center of employees’ daily work. Rather than asking customers to switch applications, Workday’s collaboration with Google Cloud focuses on extending enterprise AI agents into the tools users already know, according to Monroy.
“What we find is that a lot of customers are really bought into the Google Workspace and broader Gemini ecosystem,” he said. “We don’t want to ask customers to have to switch their front doors or log into Workday when they’re happy to use Gemini systems. Instead, we want to bring our agentic innovation through Google Cloud and allow customers to choose those front doors.”
During the first live demonstration, Monroy showed how an employee could begin a quarterly performance review directly from Gmail using Gemini Enterprise, automatically incorporate recent Workday feedback, initiate the review workflow and schedule time with a manager without leaving the existing workspace. The approach reflects a broader shift toward embedding AI into employees’ existing workflows instead of requiring them to navigate multiple enterprise applications.
“If you think about the perception of ERP systems like Workday, they’re considered clunky, slow,” Monroy said. “You’re clicking through a lot of screens; you’re navigating, and I think that perception’s fair. What you’re seeing here is what the next generation of ERP is intended to look like: It’s natural; it meets you where you are in email. It’s a structured workflow that’s allowing you to follow the guardrails … but the net result is something a lot more … joyful than the equivalent today.”
Building guardrails for enterprise AI agents
A second demonstration contrasted what Monroy described as “lawless” agent behavior that runs on the platform it’s governed by. While the first agent completed a worker onboarding request without required approvals or policy checks, the second followed compensation limits, background check requirements and approval workflows before proceeding. That difference illustrates why enterprise AI depends on deterministic guardrails backed by authoritative systems of record rather than AI reasoning alone, according to Monroy.
“In that first example, you almost didn’t know that it was lawless, because it wasn’t the agent’s fault; it was just goal-seeking in a vacuum,” he explained. “In this case, it’s running on the Workday platform that supplies those guardrails that are really difficult to supply when you’re not running on top of the system of record.”
Building those capabilities remains an evolving engineering discipline, Monroy added. Organizations developing AI agents independently must create their own approaches to context management, permissions, tool selection and governance, while Workday’s platform incorporates those controls into its runtime for enterprise workloads.
“We’re able to provide guardrails that are wildly differentiated,” Monroy said. “If you’re building on your own, you’re going to have to come up with your own versions of these things. There’s no solved answer to this. This is emerging technical best practice.”
As organizations expand enterprise AI agents into more sophisticated business processes, their deployment strategies should match the complexity of the task, Monroy noted. Simpler interactions may rely on protocols such as Model Context Protocol, with more advanced workflows often requiring orchestrated reasoning or richer user interfaces tailored to enterprise applications.
“Today, the best place to get that experience is inside of our Sana UI, where we’re able to control every pixel and provide a truly rich experience for customers,” Monroy said. “I think for folks who are trying to build AI solutions, you’ve got to think about where do you need to control just the MCP surface? Where do you need to control the reasoning chain as the second step versus where do you need to control the end-to-end UI and every pixel on the screen? That’s, I think, the guidance as best I’ve laid out.”
Here’s the complete video interview, part of SiliconANGLE’s and theCUBE’s coverage of the Google Cloud: AI Agents in Action Series:
(* Disclosure: TheCUBE is a paid media partner for the Google Cloud: AI Agents in Action Series. Neither Google Cloud, the sponsor of theCUBE’s event coverage, nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)
