Workday Rising 2026: Workday wants to be the system of record behind every AI assistant
Follow Us
No one uses the HR application of their own free will. They have to do it either because there is a form to fill, a payslip to verify, or something to approve from a manager’s point of view. Workday has been working on creating interfaces to help with all those tasks for twenty years. This year at Workday Rising 2026, Workday made a different decision.
SurveyAlso read: Workday opens its platform to AI agents: Autopilot Agents and Headless Access explained
Sana leaves home
The Workday agentic experience and orchestration engine Sana has now become the Workday Home, where employees submit requests in plain language, and Sana orchestrates Workday agents to perform actions on their behalf. Even more interesting than that is where else Sana will go. According to Workday, Sana will be used not only in Workday but also in Claude, Google’s Gemini Enterprise, Microsoft Copilot, Slack and Teams, based on self-service capabilities which are currently available.
This is an interesting move from the company which made its product its interface, but not an entirely new one. Back in May, Workday integrated Sana Self-Service Agent into Gemini Enterprise and made Gemini the default underlying model for Sana.
The Anthropic partnership
Anthropic makes headlines in this cycle. Workday and Anthropic have joined hands to create integrations between their platforms and agentic experiences. It combines Claude’s reasoning capabilities with Sana agents and Workday data & business context in HR and Finance. According to Workday, the experience will be available either through Sana or Claude, depending on the work that needs to get done. However, there is no information about the date of release and commercial terms of the deal in the announcement summary.
The connection, however, is more profound than that. Workday’s Agent-Ready Tools use Model Context Protocol, allowing an MCP-enabled agent to get an official permission to access Workday data and take action. As per Anthropic’s Ankur Sinha, the protocol allows agents a standardized method to interact with the systems where work gets done, and Workday’s Agent-Ready Tools offer this protocol in HR and Finance domains without compromising the existing business permissions and approvals. Also, Claude becomes another model choice for building agents along with Gemini, Llama, Mistral, GPT, and NVIDIA’s Nemotron.
Why the rules matter more than the interface
According to Workday’s view, a good model is not enough. Its research shows that large language models could misinterpret HR policies even if the correct policy was available for them. Payroll that is very close to being right is still wrong payroll and leads to an incorrectly paid individual. Thus, any agents from whatever source, be it Workday itself, its clients or partners, operate using Agent Harness with equal permissions, approvals and audit logging of actions. As CTO of Workday Gabe Monroy put it: “Headless can’t mean lawless”.
The same reasoning applies to the new Autopilot Agents that operate behind the scenes in the process of hiring, payroll and accounting according to the limits set by the customers. Today only Talent Acquisition Agent operates in the autopilot mode with others coming early in 2027.
What to watch
There are three questions that deserve particular attention. The first question relates to the depth of integrations in third-party assistants, because the presence in Claude or Copilot is different from the place where things get done. The second question is whether the customers would permit a single vendor to be the manager for agents created using the tools of another vendor. The third one relates to India, where the large GCCs and IT services companies have big HR and finance back offices, and would like to know when these features, including payroll compliance, will be available locally. So far, Workday has not said anything.
Also read: Epic Games AI Disney game: Marvel and Star Wars are now the test case for AI made art
