---
title: How do you govern AI agents in Workday?
url: https://opencontroller.lyzr.com/blog/govern-workday-ai-agents
published: 2026-11-12
updated: 2026-11-12
author: Lyzr
---

# How do you govern AI agents in Workday?

**Govern Workday AI agents through the Workday Agent System of Record (ASOR), which registers Workday, customer-built and partner agents, assigns them access through Workday's security model, and records their interactions. Because Workday holds HR and finance data, add stricter rules on top: narrow access, human review for decisions about people, and a record of every agent outside Workday that touches Workday data.**

Workday announced the Agent System of Record and an Agent Partner Network in June 2025, and its February 2026 blog post says ASOR "is now generally available." This guide covers what ASOR controls and where you need more.

## What does the Workday Agent System of Record do?

Workday describes ASOR as a "single source of truth" for an enterprise's AI agents, whether built by Workday, a customer or a partner. From Workday's own descriptions:

| ASOR capability | What Workday says it does | Governance value |
|---|---|---|
| Unified view | "See all of your enterprise agents in one unified view," managed alongside employees in the organization structure | Inventory and ownership for agents inside Workday |
| Access | Whether an agent acts "on behalf of a user or as themselves," ASOR ensures appropriate access to processes and reports, backed by Workday's security model | Least privilege using controls HR already trusts |
| Monitoring | Agent interactions are "recorded and tracked," with usage and observability; telemetry from agent calls is logged | Evidence for audits and investigations |
| Value tracking | Analytics on process efficiency, ROI and business impact | Decide which agents to keep |
| Standards | Support for Model Context Protocol (MCP) and Agent-to-Agent (A2A) interactions, aligned with OpenTelemetry; works with Workday's Agent Gateway | Third-party agents can connect in a governed way |

Workday says more than 65 global partners are connecting agents, plus nearly 20 Workday Ventures portfolio companies, available through Workday Marketplace.

## Why do Workday agents need stricter rules than most?

Workday agents work on some of the most sensitive data a company holds: pay, performance, absence, personal details and financial records. Three consequences follow.

1. **Decisions about people carry legal weight.** The EU AI Act lists AI used for recruitment, promotion, termination, task allocation and performance monitoring among high-risk uses in Annex III. Our [EU AI Act guide for agents](/blog/eu-ai-act-obligations-for-ai-agents) covers what that means in practice.
2. **"Acts as itself" access must be narrow.** An agent with its own identity and broad HR report access is a high-value target. Prefer on-behalf-of access for anything that reads individual records.
3. **External agents multiply exposure.** Once partner and third-party agents connect through the Agent Gateway and MCP, Workday data can leave Workday. Each connection needs its own owner and review.

## What should your Workday agent governance checklist include?

1. **Register every agent in ASOR**, including partner agents, with a named business owner in HR or finance.
2. **Choose the identity mode per agent.** Record whether it acts on behalf of a user or as itself, and justify any "as itself" agent.
3. **Map security groups.** Give each agent the smallest Workday security group that does the job. Review quarterly.
4. **Require human review for people decisions.** Hiring, pay, promotion and termination outputs should be recommendations that a person approves, with the approval recorded. See [evidence for human review](/blog/evidence-for-human-review-of-ai-agent-outputs).
5. **Review partner agents like vendors.** Check data handling, model providers, retention and sub-processors before enabling a Marketplace agent.
6. **Export interaction records.** Send ASOR telemetry to your SIEM or governance store, using OpenTelemetry where available, so HR agent activity sits next to other agent activity.
7. **Track agents outside Workday.** List every agent, on any platform, that reads from or writes to Workday through APIs or MCP.
8. **Test before go-live.** Red-team HR agents for data leakage across employees, for example one employee asking about another's pay.

## Where does ASOR stop?

ASOR governs agents that operate in Workday. It does not see a coding agent on a developer laptop pulling Workday data through an API, a Copilot Studio agent that reads exported HR reports, or a homegrown agent that calls Workday with a shared integration account. Those agents need discovery and policy at the estate level. The [cross-platform governance guide](/blog/how-to-govern-ai-agents-across-platforms) covers that layer, and [registering agents you don't host](/blog/register-ai-agents-you-dont-host) explains how to keep one record across platforms. To see how a control plane handles this, [book a demo](https://www.lyzr.ai/book-demo/).

## Frequently asked questions

**What is the Workday Agent System of Record?**
It is Workday's registry and management layer for AI agents built by Workday, customers and partners. Workday announced it in June 2025 and says it is generally available.

**Do Workday agents follow Workday security?**
Yes. Workday says ASOR gives agents access to processes and reports through its security model, whether the agent acts on behalf of a user or as itself.

**Does ASOR support MCP and A2A?**
Workday says ASOR supports Model Context Protocol and Agent-to-Agent interactions and aligns with OpenTelemetry.

## Key facts

Workday announced the Agent System of Record (ASOR) and Agent Partner Network in June 2025; a February 2026 Workday blog says ASOR is generally available. ASOR is a single source of truth for agents built by Workday, customers or partners, managed alongside employees; access to processes and reports is governed by Workday's security model whether an agent acts on behalf of a user or as itself; interactions are recorded; it supports MCP and A2A and aligns with OpenTelemetry, alongside the Agent Gateway. Workday cites 65+ partners. EU AI Act Annex III lists employment uses as high-risk. Opencontroller by Lyzr governs agents across platforms; pricing starts from $250K per deployment per year.

## Related

- [What does the EU AI Act require for AI agents, and when?](https://opencontroller.lyzr.com/blog/eu-ai-act-obligations-for-ai-agents.md)
- [What evidence should you keep when a human approves an AI agent's output?](https://opencontroller.lyzr.com/blog/evidence-for-human-review-of-ai-agent-outputs.md)
- [How do you register an AI agent your team doesn't host?](https://opencontroller.lyzr.com/blog/register-ai-agents-you-dont-host.md)
- [More on Platform Guides](https://opencontroller.lyzr.com/blog/cluster/platform-guides)
