How Workday's HR Team Is Reimagining Workflows With Sana

In this article we discuss:

• The Check-In Agent
• The Calibration Agent
• The Interview Questions Agent
• The Work Insights Agent
• 4 Tips For Building Agents in Sana

Every HR professional I talk to right now is asking the same question: How do I move beyond AI adoption and start seeing real value? I believe that success starts with concrete business problems, and no one understands those problems better than the people closest to the work. Our teams put that finding into practice with Sana from Workday, resulting in four agents that are improving performance feedback, accelerating calibration, scaling interview prep across more than 1,600 roles, and showing us how AI is actually reshaping work. 

There was no mandate from above, no pilot program, and no project tossed over the wall to a central engineering team. HR practitioners spotted a friction point in their own workflows and used Sana to fix it themselves. With Sana's natural-language authoring, in-platform testing, and prebuilt connectors to 21 systems, including Outlook, Google, Slack, Zoom, and Workday, they turned what they already knew about their processes into working agents in just weeks.

This is the essence of citizen development: giving employees with deep domain expertise the tools to design, deploy, and maintain intelligent agents tailored to their own work. The result isn't a proof of concept waiting to scale. It's a working model for how AI delivers real, measurable value when it's driven by expertise and urgency.

For those of us in HR leadership, that’s the real opportunity. The people on our teams who understand a workflow best are the ones best positioned to reinvent it, and with Sana, they can do it without writing a single line of code.

The Check-In Agent

Writing performance reviews used to mean manually stitching together scattered decks, notes, and metrics into written summaries, a process prone to recency bias and information overload. And the quality showed. When our team analyzed nearly 12,200 manager write-ups from performance check-ins against a rubric based on the Situation-Behavior-Impact (SBI) model, only 36%% received a "Good" rating. The rest were weighed down by boilerplate, goal recaps, or praise that offered no guidance for growth.

To close that gap, our talent management team built the Check-In Agent in Sana. The agent coaches employees and leaders to write clearer, more substantive check-ins. For our employees, it grounds each written review in what they actually accomplished, drawing on employee approved connections such as email, Slack messages, documents, meetings, workflow tools (Salesforce, Github, Jira etc), and Workday goals. For our managers, it pushes past generic praise toward specific, actionable feedback.

What's been most meaningful to me is that we're not just saving time. Feedback quality is getting better, too. High-quality feedback rose 50% in Q1 and 44% in Q2. Among agent users, the share of feedback rated "Good" climbed 16 percentage points, compared with +7 points for non-users.

"Using the agent reduced my check-in prep time by over 80%," said Jim Delia, SVP of global revenue operations. "I turned a complex [quarterly business review] deck into responses in 20 minutes—a task that previously took nearly two hours."

The Calibration Agent

Calibration, where leaders align ratings and distribution decisions across teams, was often conducted inconsistently. Guidance was in multiple places across our company intranet, slides, FAQs, and call recordings, so facilitators struggled to find answers quickly and relied on memory when questions came up live. And not every session had a HR business partner (HRBP) present. At lower levels, people leaders ran calibrations on their own, with no one to interpret the guidance, which made the inconsistency worse.

One HRBP decided to solve the problem herself, without having to submit an engineering request. She built a prototype in Sana, refined it with the talent management team, and trained it on the questions leaders and HRBPs were already asking in internal Slack channels. The resulting Calibration Agent now acts as a thinking partner across the full calibration cycle, assisting with preparation, live coaching, and resulting policy guidance.

Ask the Sana agent how to prepare with only a couple of hours' notice, and it returns a structured plan covering distribution guidelines, edge cases, promotion requests, and the questions leaders are likely to raise. Ask it a tough question mid-session, why the team distribution needs to change, and it responds the way a talent management partner would, explaining the reasoning, offering language for the conversation, and suggesting follow-up questions. 

The impact is clear. Eighty-three percent of our HR Business Partners said this was super useful to them, and they said it helped them make better-informed decisions. Half of users cut their prep time by as much as 75%.

"The Sana Calibration Agent became my thought partner, really boosting my confidence on both the content I was pushing out and the flow of the entire calibration session," said Laura Seto, principal people business partner.

People also found uses well beyond the original scope. A vice president in solution consulting used it as a live referee during a nine-box debate, noting that it made the conversation more data-driven. An HRBP in Singapore used it to resolve a leave-of-absence edge case in minutes, one that would otherwise have been escalated. 

Neither use case was on anyone's roadmap, and that's one of the lessons I'd pass along to any HR team rolling out AI: the most valuable applications are often the ones no one planned for. The work doesn't end at launch—we now pay closer attention to how people actually use a tool, not just how it was designed to be used.

The Interview Questions Agent

Workday has more than 800 job-critical skills across 1,600+ roles, far too many combinations to write tailored interview questions for each. As a result, hiring teams sometimes fell back on generic behavioral questions that didn’t always allow for deeper discussions. Talent Acquisition needed a more scalable way to help hiring managers assess whether candidates could apply the right skills in the situations each role actually demands.

A manager on our talent acquisition team built the Interview Questions Agent in Sana to close that gap. Interviewers simply share the role's job description and the skills they've been assigned to assess. The agent combines those inputs with skill descriptions and guidance on writing behavioral questions, then generates two tailored questions per skill. Every interviewer, regardless of experience, walks in ready to assess each candidate meaningfully.

Workday now has tailored interview questions for 100% of its skills and roles. What stood out to me most is the scale. No small team could have written tailored interview questions for more than 800 skills across 1,600 roles, which evolve constantly. Now, an agent built by one person does it for all of them.

The Work Insights Agent

Until recently, we had no clear picture of how AI was reshaping work. Job descriptions reflected what roles were supposed to do, not the tasks people actually performed, and employees had plenty of AI tools but little sense of which ones fit their work.

Our work intelligence team built the Work Insights Agent in Sana to give both leaders and employees that missing view. The agent reviews employee approved connections—such as emails, meetings, documents, workflow tools (Salesforce, GitHub, Jira etc.) and calendar—to infer the tasks they actually perform. Employees then validate that list in a short conversation, get AI-powered recommendations, and receive a personal Work Insights Profile, a shareable summary of their tasks, time, AI opportunities, goals, and ready-to-use prompts they can take back to their day-to-day work. Validated task data also flows to the work intelligence team to inform role modernization, workforce insights, and capacity planning. Essentially an agentic Time-in-Motion study at scale for every employee - reducing traditional HR research work from months to minutes.

Eighty-four percent of employees rated the agent's task-capture extremely accurate, 93% gave the conversational review the same rating, and 64% believe it will significantly change their daily work.

"Sana is the perfect tool for Workmates to filter their digital exhaust into accurate task profiles, validate those profiles, and collaborate in modernizing those tasks—all in one single agentic experience," said Carl Adams, Principal, Business Intelligence.

And we're already putting that data to work. As more of Workday's coding shifts to AI agents, we're bringing our software testers into the same organization as the engineers who build our products. By looking at what the agent tracks alongside the team's day-to-day coding activity, leaders can see testers' output catching up to engineers' in real time and know the move is working.

The data also showed us something bigger: AI accelerates technical work far more than it accelerates coordination and collaboration, which still fill much of the day. Engineers who use AI most see the biggest gains in output and morale, yet their meeting time barely changes. In other words, adopting AI doesn't automatically transform a role. That's why modernization needs to start with real, employee-validated task data, the kind the Work Insights Agent now provides. In other words, the true value of work intelligence isn't that we know more about work; it's that it gives us a map of how work is evolving.

4 Tips For Building Agents in Sana   

Based on our internal use cases, we learned a lot. Here are our biggest recommendations for other businesses building agents with Sana:

1. Build a team around the builder: The strongest agents emerged where a hands-on builder partnered with subject-matter experts and an executive sponsor. Impactful agents don’t require a large engineering team; they require the right mix of domain expertise, sponsorship, and engaged testers.

2. Define success up front: Across every agent, the biggest challenge was quantifying value after we built the agent. Establishing baselines and standardized assessments before launch protects the integrity of the metrics that ultimately prove an agent's value.

3. Keep agents narrowly focused and grounded in your standards: Our four Sana agents were highly relevant to users, driving true business outcomes because they encoded internal standards, stayed tightly scoped, and understood domain context. Well-bounded agents outperform broad ones.

4. Think about governance early: At Workday, responsible AI, legal, and privacy partners review every agent through our HR AI Hub, keeping the company safe without slowing innovation. The hub also prevents duplicate work, so builders focus on problems no one else is currently solving.

What’s Next 

These four agents are just the beginning of our journey. As we scale, these citizen-developed solutions serve as our front-line blueprint, validating what works so we can build, integrate, and deploy with confidence on the Sana enterprise platform. 

In this next phase, we will deepen these connections, while continuing to standardize how we measure impact to ensure every new build starts with clear, data-backed metrics all deployed responsibly.

And no one is better suited to solve business problems than the people closest to the work. I believe HR practitioners are uniquely positioned to close the gap between AI adoption and realized value in the workplace. As leaders, it's on us to give them the tools to do it. The companies that win in the AI era will be the ones that weave AI deep into how they work and invest just as much in their people.

The next time you have a pain point, ask yourself one question: What if I built an agent for this? Learn how to build agents with Sana from Workday.