Workday entered the first full day of Workday Rising 2026 promoting its “new Workday,” with 6,000 customers running agents live. The company announced autopilot agents for HR, finance and IT, and said Sana became the Workday home starting Tuesday, October 13. Employees can ask for work in natural language, with Sana coordinating Workday agents.

Workday CEO Aneel Bhusri said its platform supports more than 80 million users and processed more than 1.4 trillion transactions last year. He cited more than 20 years of building a unified data model and customer trust as foundations for enterprise context. At the October 13 keynote, Bhusri said data is normalized across 80 million employees and thousands of data elements, possibly millions. Context means supplying an AI agent with relevant data when it generates an answer or action. Workday argues that this can improve agents’ usefulness while keeping private customer data out of AI providers’ models and training data. The article notes that relevant context can also sit in other vendors’ products, organizational memory, or external sources such as weather and supply chain data.

Workday describes governed agents as “lawful” and the controls around them as enterprise “rails.” Gerrit Kazmaier, Workday’s president of product and technology, said the platform combines frontier AI models with deterministic process execution and security guardrails. He cited policy adherence, audit logging and tracing as controls available to customers building extensions. Workday’s acquisition of C1.ai adds identity and access management: C1 CEO Alex Bovee said the platform automates access provisioning and gives agents just-in-time permissions, with C1 acting as a runtime enforcement layer.

The company is also offering Sana through other interfaces, including Microsoft Teams, Slack and Claude. In a described example, a business leader asks Sana inside Microsoft Copilot to compare planning options; Sana checks permissions, uses company data, saves the selected plan to Workday with an audit trail, and gathers context for any required approval. Workday says this lets employees get work done without leaving the conversation while companies retain control of data, policies and approvals. Bhusri said customers can choose Workday’s own interface or a headless option, though he prefers they use Workday’s interface. The article says Workday is also using Typesafe’s Jev to generate user interfaces on the fly.

Pricing and return on investment remain open questions at the event. Workday SVP of Product Max Wessel said the company has decoupled its internal token costs from agent pricing. A customer at a breakout session said its return on investment exceeded cost projections, while also stressing the need to measure AI costs. Bhusri said Workday is still learning consumption-based pricing and wants customers to spend no more than they intend; he warned that frequent Sana queries can drive up token costs and said the company must monitor agent use.

Workday is emphasizing agents that operate continuously in the background across several domains; the article says most stage demonstrations are shipped, in early release or nearing release, and remain subject to governance and configurable human approval. The author criticizes Workday for not explaining safeguards when demonstrating higher-risk agents in areas such as human resources, despite the company’s work on responsible AI. The excerpt ends as the author begins describing a positive shift in Workday’s framing of AI beyond cost and efficiency.

Workday entered the first full day of Workday Rising 2026 promoting its “new Workday,” with 6,000 customers running agents live. The company announced autopilot agents for HR, finance and IT, and said Sana became the Workday home starting Tuesday, October 13. Employees can ask for work in natural language, with Sana coordinating Workday agents.

Workday CEO Aneel Bhusri said its platform supports more than 80 million users and processed more than 1.4 trillion transactions last year. He cited more than 20 years of building a unified data model and customer trust as foundations for enterprise context. At the October 13 keynote, Bhusri said data is normalized across 80 million employees and thousands of data elements, possibly millions. Context means supplying an AI agent with relevant data when it generates an answer or action. Workday argues that this can improve agents’ usefulness while keeping private customer data out of AI providers’ models and training data. The article notes that relevant context can also sit in other vendors’ products, organizational memory, or external sources such as weather and supply chain data.

Workday describes governed agents as “lawful” and the controls around them as enterprise “rails.” Gerrit Kazmaier, Workday’s president of product and technology, said the platform combines frontier AI models with deterministic process execution and security guardrails. He cited policy adherence, audit logging and tracing as controls available to customers building extensions. Workday’s acquisition of C1.ai adds identity and access management: C1 CEO Alex Bovee said the platform automates access provisioning and gives agents just-in-time permissions, with C1 acting as a runtime enforcement layer.

The company is also offering Sana through other interfaces, including Microsoft Teams, Slack and Claude. In a described example, a business leader asks Sana inside Microsoft Copilot to compare planning options; Sana checks permissions, uses company data, saves the selected plan to Workday with an audit trail, and gathers context for any required approval. Workday says this lets employees get work done without leaving the conversation while companies retain control of data, policies and approvals. Bhusri said customers can choose Workday’s own interface or a headless option, though he prefers they use Workday’s interface. The article says Workday is also using Typesafe’s Jev to generate user interfaces on the fly.

Pricing and return on investment remain open questions at the event. Workday SVP of Product Max Wessel said the company has decoupled its internal token costs from agent pricing. A customer at a breakout session said its return on investment exceeded cost projections, while also stressing the need to measure AI costs. Bhusri said Workday is still learning consumption-based pricing and wants customers to spend no more than they intend; he warned that frequent Sana queries can drive up token costs and said the company must monitor agent use.

Workday is emphasizing agents that operate continuously in the background across several domains; the article says most stage demonstrations are shipped, in early release or nearing release, and remain subject to governance and configurable human approval. The author criticizes Workday for not explaining safeguards when demonstrating higher-risk agents in areas such as human resources, despite the company’s work on responsible AI. The excerpt ends as the author begins describing a positive shift in Workday’s framing of AI beyond cost and efficiency.