Workday AI Agents: What They Do and What Is Available

Workday AI agents can handle routine HR and finance work today, but availability depends on the specific skill you need. Workday launched Sana Self-Service Agent with 300+ skills across areas such as pay, time, and absence; newer autopilot capabilities have separate rollout schedules. Start with an available skill that fits your existing Workday process, then confirm its entitlement and approval rules before enabling it. Workday's Sana launch.

Verified against Workday's product pages, public adoption catalog, administrator documentation, and newsroom on 26 October 2026. Availability below preserves Workday's labels; early access and announced delivery windows are not general availability.

Workday AI Agents: What They Do and What Is Available

Workday sells a portfolio of specialized agents, with different availability across their component skills. A product page describing an entire workflow does not establish that every step is available in your tenant, the organization's Workday environment.

Workday's catalog currently uses General Access for established skills and Early Access for limited rollouts. This guide keeps those exact labels. Where a dated newsroom announcement explicitly says generally available, that stronger wording is identified separately.

Workday organizes its agents around employee services, HR, finance, and IT. The practical unit to evaluate is the named skill underneath the agent family: recruiting, payroll data insights, contract review, or deployment guidance. Workday's agent portfolio.

Employee Services and HR

For an HR operations leader, begin with the repetitive request your staff already resolves successfully. A leave request and a reorganization proposal need very different approval boundaries, even when both start in a conversation.

The status labels in this table come from Workday's Agents for Everyone and Agents for HR catalogs. Product links explain the named capabilities. The autopilot and Concierge exceptions are dated below.

Within Talent Acquisition, Recruiting supports candidate prioritization and recruiter workflows; Candidate Experience supports candidate engagement through the hiring process; Talent Mobility helps employees find internal opportunities; and Contingent Sourcing recommends contractor talent. Choose the skill that addresses the bottleneck, rather than treating the family as one interchangeable hiring assistant.

Sana Concierge's launch status is narrower than its component-skill catalog might suggest. Workday's 13 October 2026 announcement says Concierge is available to early adopters, with general availability expected in the first calendar quarter of 2027. The same announcement says the new Workday Home experience is available, with administrators choosing when to enable it. Those are separate availability statements. Workday Home announcement.

For a shared-services manager, that makes Self-Service a sensible initial evaluation. A business case that requires the whole Concierge experience should instead be scoped as an early-adopter project, with a fallback to the existing case process.

Finance, Procurement, and IT

For finance, start with preparation and evidence where your existing reviewer can verify the result. For IT, start with a known configuration or business-process bottleneck. Neither requires assuming the entire department can operate autonomously.

Workday's finance catalog and IT and developer catalog supply the skill statuses below.

Several familiar names remain useful when speaking to your account team. Financial Audit Agent handles evidence collection and transaction testing; Revenue Contract Agent extracts agreement data and recommends downstream accounting treatment. Both adoption hubs list General Access and require Core Financials. Financial Audit, Revenue Contract.

Workday's main portfolio also names Financial Test Suite, describing tests on financial data and controls. Its presence in the portfolio does not establish a separate general-availability date; confirm the exact testing skill you are obtaining. Workday agent portfolio.

Likewise, the Admin Agent page still names Adoption Agent, BP Optimize Agent, and Deployment Agent individually. The first helps prioritize release features, the second identifies process bottlenecks, and the third gives configuration and troubleshooting guidance. An umbrella name is not a new entitlement for every feature underneath it.

Decision Intelligence has a documentation mismatch worth resolving before a rollout. The category catalogs still label its HR and finance entries Early Access. Workday's newer, dated 13 October 2026 release says Decision Intelligence is now available to Workday Human Capital Management (HCM) and Workday Financial Management customers, while Decision Intelligence for Adaptive remains available to Adaptive from Workday customers. Use that dated announcement for the portfolio status, then confirm the particular capability in your tenant. It supports governed analysis and scenario comparison across workforce and financial decisions. Decision Intelligence announcement.

Autopilot Is a Separate Availability Question

An existing agent's availability does not make its new background automation generally available. In its 13 October 2026 release, Workday states:

• Talent Acquisition Agent autopilot: available to early adopters, with general availability planned for early 2027.
• Payroll, Accounting, and Deployment Agent autopilot capabilities: planned for early-adopter availability in early calendar year 2027.

The recruiting example includes background pipeline monitoring and sourcing, while outreach still requires recruiter approval. Customers choose which actions run automatically, which need review, and how often agents run. Autopilot announcement.

If your payroll team needs help investigating missing inputs, evaluate the available Payroll skill. If the project assumes the new autopilot will continuously manage the wider process, put it on the roadmap until the exact capability is available to you. Do not remove a manual control based on a planned release.

Workday Illuminate AI: How to Read the Older Names

Illuminate is the branding in Workday's earlier agent announcements; the current catalog organizes the experience around Sana and broader agent families. Preserve the distinction between an old announcement and a current availability record.

Workday's September 2025 announcement also named Business Process Copilot Agent for process setup, Document Intelligence for Contingent Labor Agent for statements of work and obligations, Cost & Profitability Agent for allocation rules, Academic Requirements Agent for academic setup, and Student Administration Agent for student administration. It gave the new agents a planned 2026 availability window. That statement alone does not confirm today's general availability for those individual names. Illuminate announcement.

The older Case, Employee Sentiment, and Performance agents described service administration, feedback analysis, and review assistance. Job Architecture described job structures, Financial Close close automation, and Financial Test continuous financial testing. Planning Agent was described as early access in that release. Current skill pages and dated availability notices are more useful than counting historical announcements as separate, fully released products.

Document Driven Accounting Agent, described as extracting contracts for accounting workflows, appeared in the May 2025 announcement with planned early-2026 general availability. The current catalog describes Revenue Contract under Accounting, but an older name and a similar function are not enough to establish an entitlement or a formal rename. Ask for the current skill identifier. May 2025 announcement.

How Sana and Agent System of Record Fit Together

Sana is the experience people use; Agent System of Record is the place to establish agent identity, access, and accountability. An agent is the task-performing component, while Workday remains the business system holding the underlying records and processes.

Think of a working harbor. Sana is the dispatcher interpreting the request, agents are the service boats, and Workday is the controlled dock where the work happens. Agent System of Record, or ASOR, is the harbor master's registry and control office. Giving a boat a name does not authorize it to enter every dock.

Workday explicitly reported ASOR as generally available in its 21 May 2026 results. Its current product page describes an agent lifecycle of registration, configuration, activation, deactivation, and archiving. Each agent has an Agent System User, a distinct identity used for authentication, authorization, and auditing, with least-privilege access and traceable actions. ASOR availability, ASOR product page.

Workday's ASOR datasheet extends that management model to Workday-built, customer-built, partner, and third-party agents. It describes visibility into ownership, utilization, efficiency, and value, with Agent Gateway handling external communications. The governance promise depends on the relevant agents and interactions being connected to that system. Registering an agent does not, by itself, prove that every action it takes in an unrelated application is governed by Workday. ASOR datasheet.

For an IT leader, require evidence at the action level. An agent proposing a worker transfer should show whose authority it used, what changed, and which approval applied. A report that merely shows the agent ran is insufficient for that review. Our AI governance framework explains how to assign an accountable owner and define the evidence that owner needs.

Permissions Are Checked Beneath the Conversation

Access to the agent and access to its underlying actions are separate checks. Workday's administrator guide says it evaluates both at runtime. Administrators enable individual skills, choose the security groups allowed to use them, and align those groups with the permissions on the APIs the skills call. Agent registration and configuration.

Consider a manager who may request a change for direct reports but cannot see another department's compensation. The evaluation should include both a permitted request and a request outside that manager's authority. A helpful answer to the first does not demonstrate that the second is correctly blocked.

This is also why an agent should inherit an intentionally designed business process. It cannot repair a poorly defined approval policy simply by having better language understanding. For workflows crossing systems, the agent orchestration guide covers handoffs, retries, and where human approvals belong.

How to Get Workday Agents Into Your Tenant

Access starts with your agreement, product entitlement, and skill configuration. Flex Credits address consumption; they do not replace the required underlying Workday products.

Workday's administrator guide requires the UMSA subscription agreement for Workday-built agents in both production and non-production tenants. Customers on the older MSA must opt into UMSA and enable the feature through Innovation Services. Only agents belonging to products your organization has obtained can be registered. Administrator requirements.

The public adoption hubs make the dependencies concrete:

• Self-Service Agent: Core HCM. Its regional line excludes FedRAMP environments. Self-Service requirements.
• Recruiting Agent: Core HCM, Recruiting, and HiredScore AI for Talent. Recruiting requirements.
• Payroll Agent: Core HCM and native Workday Payroll coverage, with the hub describing availability across native Workday Payroll countries. Payroll requirements.

These are examples of why an agent being publicly available is different from your organization being ready to activate it. Check geography and environment restrictions for each selected skill, especially if a regulated deployment is involved.

Workday Flex Credits are an annual bulk-purchased, usage-based model with a complimentary allocation; eligible skills consume credits as defined on Workday's rate card. Workday says most agents can be tested free in non-production and describes planning, monitoring, and budget alerts in the Platform Consumption Console. Use your applicable rate card and pilot usage to budget, rather than assuming one universal price per agent. Flex Credits.

Enable a Narrow Skill Set First

Configure a pilot around a specific job and a limited user group. Workday documents the following activation path; the scope and acceptance criteria are your operating decisions.

1. 1

   Confirm the eligible skill

   Ask your Workday administrator to verify the agreement, required product, region, environment, and exact skill status. Record whether the project depends on General Access, early access, or a future capability.

2. 2

   Register and configure

   In Agent Management Hub, open Unregistered Workday Agent and register the eligible agent. Open it in Agent Registry, choose Configure Agent, and enable only the skills needed for the pilot.

3. 3

   Set the allowed users and actions

   Populate Available To with the intended security groups. Use View Security for Agent Skill to check the underlying security policies. Keep approvals for consequential changes in the business process.

4. 4

   Activate and inspect outcomes

   Activate the agent after configuration. Run representative requests and deliberately unauthorized ones. Inspect the resulting records and approvals, then compare completed work, corrections, escalations, and credit consumption before expanding access.

The registration, security, and activation controls above come from Workday's setup instructions. A successful pilot should show useful work completed within the intended authority, not merely a high number of conversations.

When to Use a Native Agent, Build Your Own, or Connect an Assistant

Use the packaged skill when the task fits Workday's existing process. Build when the workflow itself is distinctive. Connect an existing assistant when the main problem is where users start the work. These routes can overlap: a custom workflow can use Workday's governed tools and still appear in another assistant.

Use a Native Workday Skill for Standard Work

For an HR service lead answering recurring leave and benefits questions, evaluate Self-Service first. For a controller spending time assembling supporting records, evaluate Financial Audit. The case for a native skill is strongest when the necessary data, permission rules, and review process already live in Workday.

Skip a custom build whose main purpose is to reproduce an available skill. You would inherit maintenance, evaluation, and support work without necessarily improving the employee experience. Conversely, do not force a native skill to cover a materially different process merely because its label sounds close.

Build for Business Logic the Packaged Skill Does Not Cover

Workday's Developer Agent, Custom Agents, and Agent-Ready Tools are generally available in Workday Extend AI under its 13 October 2026 platform announcement. Agent Harness, the environment that supplies context and governs agent actions, is also generally available. This gives a Workday customer a native custom-development route as well as the option of connecting an external application. Trusted Agent Platform announcement.

There is a packaging detail to settle with your account team: Developer Agent's product FAQ still describes availability through Workday Extend Professional, while the newer announcement says Workday Extend AI. Do not assume those labels establish the same entitlement in your contract. Developer Agent FAQ.

A mid-market IT leader might need a worker-change process that checks Workday, obtains equipment approval, and coordinates another internal system. Build around that distinctive sequence, with clear ownership of failures and approvals. A solo builder should first prove one authorized action and its failure path before adding conversational planning.

Workday distinguishes its existing REST, SOAP, GraphQL, and RaaS APIs, interfaces for system-to-system integration, from Agent-Ready Tools, which expose actions through Model Context Protocol (MCP), a standard way for an AI application to discover and call tools. Its product page describes user-scoped security and schemas designed for agents. Agent-Ready Tools and MCP.

Choose a direct API integration when you need explicit application logic and a predictable sequence. Consider MCP when an agent needs to discover and select approved actions during a conversation. Either way, your design still needs a human owner, bounded permissions, safe retries, and a record of what happened. Our AI agent platforms comparison helps assess the surrounding orchestration options.

Connect an Existing Assistant When the Entry Point Is the Problem

Anthropic Claude is a concrete example of an external AI assistant becoming an entry point to Workday. Workday's 13 October announcement describes a Sana in Claude plugin that retains Workday permissions and audit controls, with users confirming changes. It is early access, with general availability expected in late 2026. The same release says Agent-Ready Tools are generally available through Extend AI, while access through the plugin comes later in the year. Workday and Anthropic announcement.

Choose that route when employees already work in the assistant and the available connector exposes the actions they need. An assistant subscription alone is not evidence that the Workday integration, product entitlement, or write actions are enabled. Confirm whether the specific connection can only retrieve information or can execute the approved transaction.

A founder at a company already running Workday should ask a practical question: is the missing piece a new workflow, or simply a better place to request an existing one? A connector can solve the latter without justifying a separate agent application.

What Is Overhyped, and What to Do Next

The risky assumption is that a broad product promise equals a fully deployed autonomous worker. Workday's own availability notices support a narrower, more useful adoption decision.

First, avoid blanket claims that every agent skill is generally available. Financial Close is still marked Coming Soon, Concierge has an early-adopter schedule, and autopilot extensions have their own dates. The catalog has useful available work without requiring those future capabilities.

Second, do not treat Workday's governance language as proof of error-free decisions. Permissions and audit trails provide boundaries and evidence. They do not establish that a candidate recommendation is fair, an accounting judgment is correct, or a generated explanation is complete. Those are outcomes your process owner must evaluate.

Third, distinguish existing governance from additional announced controls. Workday's June 2026 Agent Passport announcement describes signed testing and verification records, but projects general availability before the end of 2026; that announcement is not confirmation that it is generally available now. Agent Passport announcement. The October platform release likewise gives future early-access windows for Agent Context Service and Grounded Reasoning. Do not make a control you have not enabled part of your current assurance claim.

Your Monday move: the HR, finance, or IT process owner should bring one frequent task to the Workday administrator. Write down the exact skill, required products, permitted users, approval boundary, and evidence of success. Pilot it with a small group, review exceptions alongside completed work, and expand only when the workflow is both useful and controllable.

Act now where an available skill fits a well-understood process. Wait where the business case depends on an announced capability. Keep a reliable existing automation when conversation or adaptive reasoning would add little value.

For practical guidance on adopting AI in business workflows, join the newsletter.