Workday ties its agents to Claude — and buys the layer that governs them
The News
At Workday Rising in Las Vegas, Workday announced a partnership with Anthropic that connects Claude's reasoning to Sana, its agentic orchestration layer, so HR and finance work can start in either environment and carry forward through the Model Context Protocol. Separately, Workday said it will acquire C1.ai, an agentic identity platform for governing access for both people and AI agents, with the deal expected to close by 31 January 2027. The through-line: Workday's existing permissions and rules should govern every agent action, wherever the work happens.
Why It Matters
This is the second time in a week that an HCM giant has bet its roadmap on agents and skills — SAP agreed to buy TechWolf only days earlier. For a TA leader the practical signal is that your core system vendor is becoming the governance layer for AI agents that will increasingly sit inside hiring workflows, from screening to internal mobility. The question is no longer what features ship next, but whether the vendor you already run on will control the guardrails for the agents your recruiters use.
Taking Action
All action is situated in your corporate context. These are starting points, not prescriptions — choose what fits.
Connect
Who to talk with in your enterprise
Your HRIS lead and your IT or identity team. Ask how agent access will map onto your existing HCM and ATS permissions, and whether your current contracts give you any say over which third-party models and agents can touch candidate and employee data.
Inspect
What to take a closer look at
Inventory the AI agents and integrations already connected to your ATS and HCM. For each, check who can access what, how access is revoked when a tool or person leaves, and whether you have a record of agent activity you could produce in an audit.
Reflect
What to think about more deeply
When your HCM vendor starts selling 'identity governance for agents,' is that evidence the agent problem is real — or evidence the vendor is building a moat around your data before the agents arrive?