Workday Risks Becoming Data Infrastructure as External AI Agents Handle Customer Workflows

Workday, one of enterprise software's most dominant players, faces an existential threat: as customers increasingly deploy external AI agents from Anthropic, Microsoft, and others, the HR and finance platform risks becoming merely the data layer beneath more valuable AI-driven work. According to consultants managing over 2,000 Workday accounts, clients are routing their agentic AI initiatives through third-party agents rather than Workday's native tools, despite signing usage-based AI agreements. At consultancies tracking 400+ Workday implementations, approximately 50% of customers committed to usage-based AI pricing, yet only 10–20% actually use Workday AI, and virtually none deploy Sana Enterprise, Workday's flagship agent platform acquired to accelerate its agentic transformation. This gap between commitment and actual adoption reveals the fundamental risk: companies like NinjaOne and Druva are building external agents that extract and orchestrate Workday data, positioning themselves as intermediaries that capture workflow value while Workday supplies the underlying infrastructure. The concern echoes through enterprise software broadly. If customers route AI work through general-purpose agents built on Claude, Copilot, or other foundation models, SaaS platforms risk transformation from centers of innovation to plumbing—critical but increasingly undifferentiated. Workday argues that AI cannot replicate deterministic enterprise processes: payroll cannot run on reasoning alone, financial close requires approval chains, and hiring workflows need segregation of duties enforced over decades of configuration. Yet the mismatch between Workday's vision of embedded agents and customer behavior suggests the market sees differently. In response, Workday launched its Agent System of Record (ASOR) in mid-2026, a registry that governs not just Workday-built agents but external ones too—allowing third-party agents to connect through standard interfaces like Model Context Protocol (MCP) and A2A to access Workday data. The company also began offering free AI credits and waived API charges through February 2027, a signal of urgency around adoption. The question is whether governance and pricing incentives can reverse a structural shift: as autonomous agents become the primary interface for enterprise work, the value flows toward the agents themselves, not the systems that house the data.