Workday and Google Cloud put Workday's Agent System of Record together with Gemini Enterprise, so agents from Workday, Google and third parties can work HR and finance tasks under one governed roof. The phrase is the tell: agent sprawl leads somewhere specific, you will need a system of record for the agents themselves.
On 28 May, Workday and Google Cloud announced an expanded partnership that is more interesting for its vocabulary than its features. They combined Workday's Agent System of Record with Google Cloud's agent platform and Gemini, so that agents from Workday, Google and third parties can collaborate on real HR and finance workflows under one governed roof. The first concrete piece is Workday's Sana self-service agent running inside Gemini Enterprise, with Gemini as its default model, letting an employee ask a question and get a personalised answer from Workday data while every permission and policy stays intact. "Our customers want HR and finance at their fingertips, not scattered across a dozen applications," said Workday's Gerrit Kazmaier.
What an Agent System of Record actually claims
Read the name slowly, because it is a real idea dressed as a product line. An ERP is the system of record for transactions; an HR platform is the system of record for employees. Workday is asserting that there now needs to be a system of record for the agents: a single authoritative register of which agents exist, what each is permitted to do, whose data it may touch, and how it hands work to the next one. That is not a feature so much as an admission. If agents are going to act across your business, something has to know they exist and hold them accountable, the same way a ledger holds transactions. Workday is trying to be that ledger.
The concrete integration underneath the concept
The idea is anchored by shipping detail. The Sana agent is available inside Gemini Enterprise; Gemini is the default model but customers can use others. The partnership enables Agent-to-Agent, Agent-to-UI and Model Context Protocol handoffs, so a task can pass between agents and interfaces inside a single workflow rather than dead-ending. The use cases are the daily grind of HR and finance: checking leave balances and payslips, updating personal information, bulk timesheet approvals, initiating performance reviews, and answering expense and travel-policy questions with guided help. And the rollout leans on the large integrators, Accenture, Deloitte and KPMG, to put it into customers' hands. This is not a demo; it is an operating model for agents doing back-office work.
It is the same story every vendor is now telling
Step back and Workday's register is one more instance of a pattern that has defined the year. Microsoft gave Dynamics 365 an Immersive Home to manage and monitor fleets of ERP agents; SAP built an AI Agent Hub and got ISO-certified to govern its agents; Microsoft and SAP wired their agents together so they could hand tasks across vendors. Now Workday names the layer outright: a system of record for the agents. The whole industry has arrived at the same realisation within months of each other. Capability without a register, without a governed place that knows which agents run and what they may do, is not power. It is sprawl, and sprawl in a finance system is a liability.
A register is not a strategy
Here is the distinction to keep, because it is easy to be dazzled by a clean governance story. A system of record for agents tells you, authoritatively, which agents exist and what they did. It does not tell you whether what they did was the best thing to do. Knowing that you have a payroll agent, a collections agent and an expense agent, each scoped, permissioned and logged, is necessary and it is genuinely valuable, but it is not the same as knowing that the collections agent chased the right accounts in the right order, or that the policy it enforced was the one that actually protects your cash. The register keeps order. Order is not advantage. A perfectly catalogued fleet of agents can still be executing mediocre decisions with impeccable governance.
The multi-vendor future it quietly assumes
The most honest thing about the Agent System of Record is what it assumes: that no single vendor's agents will run everything. Its whole reason to exist is that agents from Workday, Google and third parties need a shared, governed place to coexist and hand work to one another. That is almost certainly the real future, a business running agents from several makers at once, and it raises the bar twice over. It makes governance harder, because the register now spans vendors, and it makes decision quality harder, because the optimal call has to hold no matter which vendor's agent happens to execute it. A register that spans vendors is useful precisely because the world is going to be messy. It does not make the world's decisions any wiser.
The layer above the register, for Dynamics teams
The pattern is the same whichever platform you run. The agents execute the transaction; a register like Workday's, or an Immersive Home in Dynamics, keeps them accounted for and governed; and above both sits the system of intelligence, the layer that computes the decision that is optimal for your specific business rather than the one the agent would reach on its defaults. On Microsoft Dynamics that runs as a companion app in tandem on Dataverse and Power Platform, shipped through AppSource, inheriting the platform's governance instead of working around it, and it is where Cognilium works. Workday just gave agent sprawl a ledger, which every serious operation will need. The register tells you the agents are accounted for. Whether they are deciding well is a separate question, and a more valuable one to answer, the same thread we followed across Business Central's agents and the move to running whole fleets of them.
Share this article
Weekly AI engineering brief
One email a week. New model releases, agent patterns, and lessons from production systems we ship.
No spam, no client data sales. Unsubscribe any time.

Ali Ahmed
AI Business Analyst & Product Owner, Cognilium AI
Ali Ahmed
AI Business Analyst & Product Owner, Cognilium AI
Ali Ahmed is an AI Business Analyst and Product Owner at Cognilium AI, where he owns the product…
