Workday Ships Sana as Its Front Door and Agent Harness for Enterprise AI

30-Second Summary

Workday replaced its home screen with Sana, a natural-language agentic interface for HR, finance, and IT. Employees ask for outcomes in plain language, and Sana routes the request to agents or people. At the same time, Workday released its Agent Harness to general availability. The Harness lets organisations build custom AI agents that run inside Workday with memory, guardrails, data sovereignty controls, and a choice of underlying model. Anthropic Claude is available as one of those models through a new partnership, with deeper integration rolling out in stages.

At a Glance

• Topic: Enterprise AI, Agentic Platforms
• Company: Workday
• Date: October 13, 2026 (announced at Workday Rising, Austin)
• Announcement: Sana Home generally available as the new Workday front end. Agent Harness, Agent-Ready Tools, Workday Data Cloud, Developer Agent, and Custom Agents all reached general availability simultaneously.
• What Changed: Workday's home screen is now an agentic interface. Enterprises can build and run their own lawful agents inside Workday for the first time using the Agent Harness.
• Why It Matters: Workday sits at the centre of how large and mid-sized organisations manage people, money, and operations. Moving the default interface to an agentic model affects every user immediately, not just technical teams.
• Who Should Care: HR leaders, CFOs, IT directors, and operators at any organisation using Workday.

Key Facts

• Sana Home is generally available and priced under a Flex Credits consumption model.
• The Agent Harness provides memory, guardrails, sovereign routing (data residency controls), model choice, and step-by-step audit traces.
• Workday's Trusted Agent Platform now includes headless access to HR and finance data, letting external agents and systems query Workday programmatically.
• Anthropic Claude is integrated into Sana agents for HR and finance workflows, with the Sana Cowork product in early access.
• Features still coming: Agent Context Service (early access late 2026, general availability early 2027), Sana Concierge (general availability Q1 2027), and Talent Acquisition autopilot (general availability early 2027).
• Cross-platform Sana is being built into Claude, Gemini Enterprise, Microsoft Copilot, Teams, and Slack, with staged rollouts from late 2026.

What Happened

On October 13, 2026, Workday used its annual customer conference, Workday Rising in Austin, to announce the largest change to its product interface in several years. Sana, which Workday introduced as a concept at an earlier event, is now the default home screen for all Workday users. Rather than navigating menus, employees type what they need in plain language, and Sana routes the request to an agent, a person, or a workflow.

At the same time, Workday released its Agent Harness to general availability. The Harness is the technical infrastructure that every Workday agent runs on, and it gives enterprise customers the ability to build their own custom agents for the first time. The Harness includes memory so agents can maintain context across interactions, guardrails to prevent agents from taking actions outside defined policies, sovereign routing to keep data in specified geographic regions, and model selection so organisations can choose which AI model powers their agents.

The release also made several related products generally available: Agent-Ready Tools, which expose Workday's HR and finance capabilities to external agents; Workday Data Cloud; Developer Agent; and Custom Agents. Taken together, this represents Workday opening its platform to the broader AI agent ecosystem while maintaining its own compliance and governance controls.

The Anthropic partnership gives Sana agents access to Claude for tasks in HR and finance. The Sana Cowork product, which puts a Claude-powered interface directly inside Workday's agentic layer, is in early access now.

Why It Matters

Workday's scale makes this structural, not optional. Workday serves more than 10,000 organisations and processes payroll, benefits, hiring, budgeting, and procurement for tens of millions of workers. A change to the default interface is not a feature release. It is a change to how work happens, at scale, for a large share of the global enterprise workforce.

The Agent Harness closes the governance gap. The single largest reason enterprises have avoided deploying AI agents in HR and finance is the absence of reliable controls. An agent that can query payroll data, approve leave requests, or initiate purchase orders needs guardrails, audit trails, and data residency controls before it is lawful in most jurisdictions. The Harness ships with all of those. It does not solve every compliance question, but it removes the most common blockers.

Headless access changes what partners and customers can build. Workday's Trusted Agent Platform now provides an API layer that lets external systems query HR and finance data programmatically, without a human navigating the Workday interface. This is significant for operators who run AI agents outside Workday, such as a sales operations agent that needs to check headcount data, or a custom onboarding workflow that needs to create a worker record.

The Claude integration is a direct connection, not a wrapper. Workday's partnership with Anthropic connects Claude to Sana's agent layer, meaning Claude can read and act on Workday data through sanctioned tools rather than through a screen-scraping or chat interface. For operators already using Claude for other workflows, this is the path to connecting those workflows to HR and finance data.

The roadmap signals where Workday is headed. The features still coming, including the Agent Context Service, Sana Concierge, and Talent Acquisition autopilot, suggest Workday is building toward a model where most routine HR and finance interactions are completed by agents without human involvement. The October 13 release is the foundation, not the finished building.

The David and Goliath View

Workday Rising was not primarily a technology conference. It was Workday telling its 10,000 customers that the product they bought has changed. Sana Home is not an optional add-on. It is the new front door. Operators who are not already thinking about how their teams will use it are already behind.

The Agent Harness is the more significant of the two releases for organisations that want to build rather than just use. A governed, auditable, data-sovereign agent runtime inside the world's most widely adopted HR and finance platform is the infrastructure that most AI agent projects in the enterprise have been missing. It will accelerate the deployment of AI in functions that have been cautious precisely because the guardrails did not exist.

For organisations using Claude, this is the clearest integration path to date for connecting Anthropic's models to HR and finance workflows. The Sana Cowork early access programme is worth applying for now.

Where This Fits in the AI Stack

• Model layer: Anthropic Claude (via Workday partnership), plus model choice built into the Agent Harness
• Agent runtime: Workday Agent Harness (memory, guardrails, sovereign routing, audit traces)
• Data layer: Workday Data Cloud, headless API access via Trusted Agent Platform
• Interface: Sana Home (natural language, multi-agent orchestration)
• Cross-platform: Sana inside Claude, Gemini Enterprise, Copilot, Teams, Slack (staged rollout)

Questions Operators Are Asking

Do I need to do anything to get Sana Home? Workday has made Sana Home generally available to existing customers under the Flex Credits consumption model. Check with your Workday account team to confirm whether your licence includes it and whether it has been switched on for your organisation.

Can I stop Sana from making changes without human approval? Yes. The Agent Harness includes guardrails that define what agents can do autonomously and what requires human confirmation. The default configuration is conservative, meaning agents surface recommendations rather than taking autonomous action unless you explicitly expand their permissions.

What data can Sana and the agents access? Agents built on the Harness access only the Workday data they are explicitly granted access to through the Agent-Ready Tools layer. Sovereign routing controls mean you can specify that data does not leave a particular geographic region. No agent has unrestricted access to your Workday tenant.

Is the Claude integration secure? The Sana Cowork product, which uses Claude, runs inside Workday's own security perimeter through the Agent Harness. Workday has not published its full data-sharing agreement with Anthropic, but the architecture routes requests through Workday's agent layer rather than sending raw Workday data to Claude directly.

When does Sana Concierge come out? Workday announced early-adopter availability now for Sana Concierge, with general availability expected in Q1 2027. Contact your Workday account team to apply for early access.

Citable Summary

Workday made Sana its default agentic interface on October 13, 2026, and released its Agent Harness, Agent-Ready Tools, Developer Agent, and Custom Agents to general availability at Workday Rising. The Agent Harness provides memory, guardrails, sovereign routing, and model choice for enterprise AI agents running inside Workday's HR, finance, and IT platform. An Anthropic partnership connects Claude to Sana agents, with deeper integration available through the Sana Cowork early access programme.