Workday Illuminate, Agent System of Record, and Custom Agents: What Each Covers

Workday's AI architecture separates what an agent does, how it is built, and where it is governed into three distinct layers: Workday-built agents (delivered under the Workday Illuminate umbrella for standard HR, Finance, and implementation workflows), Custom Agents (built by your developers in Workday Extend AI using App Builder, .agent and .agentskills definitions, Agent-Ready Tools, and Custom Agent Actions), and the Agent System of Record (ASOR) (the tenant-wide governance control plane where administrators register, secure, and audit all five agent categories).

Side-by-side comparison of the Workday AI stack

When planning an AI rollout or answering architecture reviews, start by distinguishing the control plane (ASOR) from the agents registered inside it. Workday supports five agent categories inside ASOR: Workday-built agents, Custom agents (built in Workday Extend), External agents (also called Self-Built Agents, built on external platforms), Partner agents, and Sana User agents.

Layer 1: Workday-built agents (Workday Illuminate)

Workday-built agents are turnkey assistants delivered by Workday for specific functional domains. You do not write code or orchestrations to create them; instead, you discover eligible agents in the Find Unregistered Workday Agents report (on the Unregistered Workday Agent tab of the Agent Management Hub), register them, toggle on the specific skills you want to use, and assign security groups in the Available To prompt.

Workday's documented catalog of Workday-built agents includes:

• Employee & HR Service Delivery: Self-Service Agent (personal info updates, time off requests, goals, learning status, policy articles), Case Agent (case type suggestions, automatic routing, conversation summaries), and Talent Sentiment Agent (Peakon survey comment and score analysis).
• Finance, Payroll & Procurement: Payroll Agent, Financial Audit Agent (procure-to-pay audit sample requests), Revenue Contract Agent (creating and analyzing customer contracts from uploaded documents), Planning Agent and Decision Intelligence (conversational analysis and what-if scenarios in Adaptive Planning), Travel Agent, and Contingent Sourcing Agent (Workday VNDLY talent pool grading and rediscovery).
• Reporting & Tenant Operations: Core Reporting Agent (conversational report exploration and summarization), BP Optimize Agent (business process bottleneck analysis and step-by-step recommendations), Adoption Agent (Release Impact Analysis in Preview tenants), and Deployment Agent (RAG-assisted configuration inspection and troubleshooting in Implementation, Preview, Sandbox, and Customer Central non-production environments).

Layer 2: Agent System of Record (ASOR) and Agent Management Hub

Agent System of Record (ASOR) is not an AI assistant itself—it is Workday's centralized registry and security framework for managing your digital workforce. It is included in your core Workday subscription without an extra SKU, though you must enable and configure it separately in each tenant because ASOR configurations cannot currently be migrated across tenants.

Enabling ASOR in a tenant

1. Run Maintain Functional Areas, check Enabled for Agent System of Record, and run Activate Pending Security Policy Changes.
2. Open Domain Security Policies for Functional Area for Agent System of Record and enable six domains: Agent Compliance, Agent Management Hub, Manage: Agents, Reports: Agent Reporting, Setup: Agents, plus Reports: AI Agent Security (under the System functional area).
3. Use Maintain Permissions for Security Group to grant View and Modify permissions to unconstrained security groups (ASOR security policies currently support unconstrained security groups only) and activate pending security policy changes.

How ASOR enforces identity and runtime security

When you complete the Configure Agent task and activate an agent in the Agent Management Hub, Workday automatically creates an Agent System User (ASU) account (viewable in the All AI Agent Accounts report) and manages its private key OAuth 2.0 client credentials.

ASOR separates agent interaction access from tool execution access:

• For delegate mode skills, a user must belong to a security group listed in the skill's Available To prompt and hold domain or business process security permissions for the underlying Workday APIs that the skill's tools execute. Administrators verify this alignment using the View Security for Agent Skill report.
• For ambient mode skills (which run on a schedule or event without human interaction), Workday generates an Ambient Agent Security Group (ASG) that administrators must add directly to the required domain and business process security policies.

Note that built-in usage analytics reports in the Agent Management Hub (Total Agent Sessions, Total Unique Users, Skill Interactions, and input/output compliance reports like Agent Input and Output Request Details) are documented as available only for Workday-built agents, whereas Active Agents, All AI Agent Accounts, and View AI Agent User Audit Trail cover all registered agents.

Layer 3: Custom Agents in Workday Extend AI

When Workday-built agents do not cover a company-specific workflow—or when a workflow spans custom business objects and third-party APIs—developers build Custom Agents in App Builder on the Workday Developer Site.

How Custom Agents are structured

A custom agent consists of two file types in the open AgentSkills format:

• An agent definition file (.agent) defining the agent's persona, instructions, and behavior.
• One or more skill files (.agentskills) declaring the tools the agent is authorized to invoke.

You can package these files inside an existing Extend app or inside a dedicated Agent app (which holds only Agents, Skills, Attributes, and Orchestrations, excludes PMD pages and Extend Model components, and does not count against your Extend app entitlement limit). Inside App Builder, you can write these files manually or use Developer Agent (specifically its Custom Agent Creator subagent) to discover tools and generate the definitions.

What tools a Custom Agent can execute

Custom agents run in delegate mode only and cannot make arbitrary, undeclared API calls. Every action must go through one of three documented mechanisms:

1. Agent-Ready Tools: Workday's curated library of pre-built tools for reading and writing Workday data, categorized in the tools explorer as Basic (read-only), Advanced (write-only), and Prompt (read-only lookup tools required by certain Advanced tools).
2. Custom Agent Actions: Orchestrations built from the Custom Agent Action template in your app. These orchestrations can call Workday REST and SOAP APIs, query WQL, and invoke external connectors. Each action is marked with an effect level: Read-Only, Idempotent, Destructive, or Open World. Every non-Read-Only action prompts the invoking user for explicit approval at runtime.
3. Custom scripts: Scripts bundled in the skill or generated from skill instructions that execute in an isolated sandbox with no network access and no direct access to Workday APIs.

Once deployed to a tenant and activated in ASOR, employees can invoke a custom agent from the Self-Service Agent chat panel, Sana, business processes, or external orchestrators connected over Agent to Agent (A2A) or Model Context Protocol.

How kiweely fits

Deciding whether a business request is already covered by a Workday-built agent skill or requires a custom Extend AI agent—and wiring the right Custom Agent Action orchestrations and domain security policies—requires inspecting live tenant metadata. Kiweely connects to your Workday environment, maps the required REST, SOAP, and WQL operations against your tenant's security model, and builds verified Extend apps and orchestrations without hallucinated endpoints. Kiweely Personal is free for your Workday personal developer account and its tenant. Any other tenant needs a Personal licence, $99 a month, or a seat in your firm.

Frequently asked questions

Do I need an additional SKU to use the Agent System of Record (ASOR)?

No. ASOR is included in your core Workday subscription at no additional SKU cost. However, to register agents in a Production tenant (or to use Workday-built agents in any tenant), your organization must be on the Universal Main Subscription Agreement (UMSA), and running agents consumes Workday Flex Credits according to your contract.

Why does a custom agent skill fail at runtime even though the user's security group is in Available To?

Adding a security group to a skill's Available To prompt in the Agent Management Hub only grants permission to interact with the skill. Because custom agents and delegate skills run on behalf of the signed-in user, the user must also belong to a security group that has View or Modify permissions on the underlying domain and business process security policies for every API tool the skill calls. Run View Security for Agent Skill to identify the missing API policy permission.

Can I migrate Agent System of Record configurations from Sandbox to Production?

No. Workday documents that there is currently no migration tool for ASOR configurations. You must enable the Agent System of Record functional area, enable the ASOR security domains, and register, configure, and activate agents in each tenant individually.