Workday rethinks the enterprise workflow for AI agents
SiliconANGLE Victoria Gayton
Workday is pushing AI agents into Gmail and Google Workspace, not making users jump into Workday. The twist: it’s trying to keep enterprise guardrails intact while the agents do the busywork.
Based on reporting by SiliconANGLE, Victoria Gayton — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Workday wants enterprise AI to feel less like a chatbot bolted onto old software and more like work flowing through the tools people already use. That’s the pitch Gabe Monroy, Workday’s chief technology officer, laid out during the Google Cloud: AI Agents in Action Series on theCUBE. The company is working with Google Cloud to bring AI agents into existing workflows without losing the governance, approvals and record-keeping that keep enterprise systems usable for serious business tasks.
The emphasis is on front doors. Monroy said many customers are already comfortable in Google Workspace and the broader Gemini ecosystem, so Workday does not want to force them to start their day by logging into Workday if they’d rather stay in Gmail or Gemini systems. In a live demo, an employee began a quarterly performance review from Gmail through Gemini Enterprise, pulled in recent Workday feedback, kicked off the review workflow and set up time with a manager without leaving the workspace. It is a neat little demo, but also a telling one: the software is supposed to disappear into the flow of work, not become another destination.
Monroy framed the problem as a mismatch between two kinds of systems. AI is probabilistic; enterprise software is supposed to be deterministic, with rules around people, money and process. That tension shows up fast when an agent is left on its own. In another demo, one onboarding agent ran ahead without the required approvals or policy checks. A second agent, running on Workday’s platform, followed compensation limits, background check requirements and approvals before moving on. Same goal. Very different result.
His point was blunt: if the AI is not sitting on top of the system of record, the guardrails are hard to recreate. Workday says its platform bakes in context management, permissions, tool selection and governance at runtime, while companies building agents on their own have to invent those controls themselves. Monroy called that an emerging technical best practice, which is a polite way of saying the industry is still figuring out how to keep agents from freelancing in places where freelancing is a bad idea.
Workday is also thinking in layers. Simpler jobs can use protocols like Model Context Protocol, while more complex workflows may need orchestrated reasoning or a richer interface. For now, Monroy said the best place for that richer experience is Workday’s Sana UI, where the company can control “every pixel.” That is a very enterprise sentence, but the direction is clear: AI agents are only useful in the back office if they can be trusted to stay in the lines.
My take — AI-written commentary, not fact-checked reporting
This is the right fight. Enterprise AI is not impressive because it talks back; it’s useful when it follows rules without turning every workflow into a small compliance incident. The industry keeps acting like the answer is smarter agents, when the real prize is boring control over permissions, approvals and system-of-record truth.
Read more about this at: SiliconANGLE