As AI agents move into production, workforce data takes centre stage

A year ago, most enterprise AI discussions centred on assistants and copilots. How do we build them? How do we connect them to enterprise systems? How quickly can we deploy them?

Today, different questions are beginning to dominate. What data can agents access? What actions can they take? Who authorises those actions? How are they monitored?

Several announcements from Workday this week reflect that shift. The company unveiled new tools for building AI agents, expanded access to workforce and financial data through its Data Cloud platform and introduced Agent Passport, a framework for testing, verifying and monitoring AI agents.

Taken together, the announcements suggest an emerging enterprise architecture for agentic AI. One layer provides the interface through which employees interact with AI. Another provides access to workforce, operational and business data. A third provides permissions, governance, security and auditability.

The learning industry has spent much of the past two years focused on the first layer. Vendors have introduced assistants, coaches, content generation tools and personalised learning experiences. The second and third layers are now attracting more attention.

Enterprise agents require context: they need access to information about people, skills, roles, reporting structures, policies and business processes. They also require controls that determine what actions can be taken and under what circumstances.

For HR technology vendors, this places workforce data in a different position. Information once viewed primarily through the lens of talent management increasingly has operational significance for AI systems. Skills data is one example. Questions about capability gaps, workforce planning, internal mobility and development opportunities all depend on accurate information about workforce capabilities. As agents begin to support or automate this, the quality and location of that data is key.

Workday’s acquisition of Sana last year was one element of a strategy centred on natural language interaction across enterprise systems. Recent product developments build on that approach, combining conversational interfaces, workforce context and governance capabilities within a single platform.

Learning platforms have traditionally managed learning histories, certifications, compliance records and, increasingly, skills information. As workforce intelligence becomes more closely connected to enterprise AI, questions emerge about where those data assets sit and how they are shared across the technology stack.

A year ago, the challenge was building AI assistants. Today, organisations are grappling with what those systems know, what they can do and who controls them.