Workday has introduced Sana, an AI governance solution, to address the challenges enterprise AI agents face due to permissioning issues rather than model performance. This initiative recognizes that many organizations struggle with maintaining integrity in approvals and security when using external solutions. According to Gerrit Kazmaier, Workday’s president for product and technology, the integration of governance directly within the system of record is critical for ensuring that agents operate safely and effectively, especially in areas like HR and finance where accuracy is paramount. Errors in these domains can lead to significant consequences, emphasizing the need for AI systems to function within established organizational structures to avoid potential chaos.

Würk: Würk operates in the HR and finance space as a practitioner-focused organization. Its director of product highlighted that AI agent permissions must reside in the core system of record to maintain control in regulated environments, warning that external definitions lead to failure.
Google: Google has expanded its partnership with Workday to integrate Sana as the governance layer for AI agents within Gemini Enterprise. This allows agents built on Sana to leverage Google’s reasoning capabilities while maintaining Workday’s business process logic and verification models.
Workday: Workday is an enterprise software company focused on human resources and finance applications that serve as systems of record for organizations. It launched Sana in March to embed governance, approvals, and security models directly into its platform for AI agents. Workday expanded its Google partnership to make Sana-built agents discoverable in Gemini Enterprise, helping customers avoid overly broad results from DIY agent setups.
Compance.AI: Compance.AI develops AI solutions with an emphasis on agent management and enterprise deployment. Its leadership stressed the risks of chaos arising from undefined agent ownership, performance tracking, costs, and actions outside proper governance frameworks.
Dan Obendorfer: Dan Obendorfer is director of product at Würk. He has stated that governance for AI agents must live inside the system of record, as permissions defined externally undermine the integrity of where data and rules actually reside.
Gerrit Kazmaier: Gerrit Kazmaier serves as president for product and technology at Workday. In a recent interview, he explained that Sana preserves the integrity of permissions and security models for agents, addressing common customer struggles with external data access that loses organizational context. He stressed that near-accurate outputs are unacceptable for critical HR and finance tasks like payroll and scheduling.
Kadan Stadelmann: Kadan Stadelmann is chief technology officer and co-founder of Compance.AI. He has underscored that effective AI agent systems require clear ownership of performance, costs, and actions to prevent operational disorder.

`json
{
“Accuracy Requirements”: “In HR and finance, AI outputs require high reliability because errors in payroll, scheduling, or financial processes can cause immediate damage without easy correction loops.”,
“Governance Integration”: “Embedding AI agent controls directly in the enterprise system of record ensures adherence to approvals, security models, and organizational hierarchies that external solutions can often fail to maintain.”,
“Enterprise AI Bottleneck”: “Permissioning and governance, rather than model performance, are the primary challenges for enterprise AI agents operating in production workflows.”
}
`