Enterprise Apps as Systems of Action: Workday Pushing AI into Execution with Sana Launch

Workday is moving to collapse the gap between AI and enterprise execution, embedding a new conversational layer directly into its HR and finance platform so users can complete work, not just analyze it.

In a March 17 announcement, Workday introduced Sana, a unified AI interface, alongside a self-service agent with more than 300 skills and an “enterprise” layer that connects into systems like Salesforce, Outlook, and ServiceNow to orchestrate work across applications.

The launch targets a problem that has limited enterprise AI adoption over the past couple of years: Most copilots sit outside the systems that actually run the business. They can generate answers, but they cannot execute tasks within governed workflows, where permissions, data models, and audit requirements matter.

Workday’s position is that AI only becomes operationally useful when it is embedded inside those systems. “AI only works in the enterprise when it’s connected to trusted, deterministic systems,” CEO Aneel Bhusri said in the announcement, pointing to the need to ground AI in the same data and rules that govern HR and finance processes.

From Sidecar Copilots to Embedded Execution

In a media briefing preceding the launch, executives drew a clear distinction between the current generation of AI tools and what comes next.

They described most copilots as “side panels” attached to applications, useful for summarization or search but disconnected from execution. The shift with Sana is to bring AI into the core of deterministic business processes, where outcomes must be correct every time.

That architectural argument showed up repeatedly. Enterprise systems like payroll or financial close are deterministic, with defined start and end states. AI, by contrast, is probabilistic, designed to reason and suggest next steps. The value, executives said, comes from combining the two.

The result is a model where AI determines what should happen, but execution runs through the system of record.

Gerrit Kazmaier, Workday’s president of product and technology, described this as a move away from fragmented tools and interfaces toward a single conversational layer that can both interpret intent and carry out work across systems.

In practice, that means Sana can do more than retrieve information. It can update records, trigger workflows, generate reports, and coordinate tasks across applications in a single interaction.

Joel Hellermark, CEO at Sana and SVP and GM of AI at Workday, illustrated this during a demo, showing how processes that previously required dozens of steps across multiple systems can be reduced to a single prompt, with the agent handling the orchestration in the background.

Integration Depth Is Operational

The deeper play is not just a new interface, but a shift in where AI sits in the enterprise stack.

Workday is putting forth a control layer for AI execution, arguing that identity, permissions, and business process context must originate from systems of record. That gives it a structural advantage in ensuring accuracy and governance, particularly in regulated domains like HR and finance.

This is also how the company is addressing one of the biggest concerns around enterprise AI: hallucinations and reliability.

The executives argued that enterprise environments, with structured data and governed processes, reduce the likelihood of incorrect outputs compared to consumer AI use cases. The combination of clean data models and process constraints is intended to “engineer out” uncertainty by anchoring AI decisions to verifiable system context.

At the same time, Sana extends beyond Workday through connectors into third-party applications, allowing agents to act across systems while still using Workday as the source of identity and control.

That balance between openness and control is central to the strategy: orchestrate across the enterprise, but execute through governed systems.

From Licenses to Outcomes

The launch also signals a shift in how Workday plans to monetize AI.

Executives confirmed that Sana capabilities will be delivered through a consumption-based Flex Credits model, moving away from traditional seat-based pricing toward usage and outcomes.

That aligns with how Workday is framing AI value internally. Instead of selling access to software, the company is positioning AI around completed work, whether that is contracts analyzed, workflows automated, or support cases deflected.

The self-service agent reflects that shift. With hundreds of prebuilt skills, it automates routine HR and finance tasks, reducing manual effort and support overhead while increasing speed of execution.

This also has implications for workforce structure. The executives acknowledged that routine operational work will increasingly be handled by agents, while human roles shift toward higher-value activities such as planning, analysis, and strategy.

Bhusri noted that tasks like payroll processing are likely to be fully automated over time, while areas requiring judgment, such as workforce planning or financial decision-making, remain human-led.