Your browser doesn’t support HTML5 audio
Today brings a new AI-powered user experience (UX) to enterprise finance and people management platform Workday, as it rolls out general availability of Workday for Sana to its customers in what co-Founder and CEO Aneel Bhusri acknowledges as "Workday's next chapter" in its history. The fruit of its acquisition of Stockholm-based AI start-up Sana last year, the new UX provides a conversational AI experience that replaces the traditional menus and navigation that have previously been characteristic of enterprise applications, and reaches beyond Workday's own platform to find information and perform actions across a wide range of third-party enterprise applications.
Workday is introducing three new components today. The first is Sana for Workday, the application vendor's new AI interface to its core HR and finance applications. Gerrit Kazmaier, the vendor's President, Product & Technology, explains that this:
[R]eimagines how people in HR and finance get their work done when working with Workday — a beautiful experience that seamlessly integrates with the Workday agents.
The second component is accessed via this new UX. The Sana Self-Service Agent automates a wide range of HR and finance workflows that users previously would typically have gone to an HR helpdesk or shared service center to fulfil. It has a wide range of skills and capabilities, from answering simple queries, such as 'How many vacation days do I have left?' to performing complex actions, such as 'Build me a dashboard showing recruitment pipeline stage and interview feedback.'
Finally, Sana Enterprise extends both the reach of the UX and the domain of the Self-Service Agent to connect to third-party applications, providing a single access point to find, organize and automate work across potentially all of a user's enterprise systems and applications. Kazmaier continues:
Sana Enterprise basically takes all of that potential and scales it out across the entire enterprise application ecosystem, the same AI native conversational experience, but unlocks it across all the applications beyond Workday.
This connection into other applications means that the aim is for Sana for Workday is to become the primary way people access their work, drawing on a deep awareness of all of their context as it does so. He goes on:
The way we think about that is that the Sana applications are the last software that people have to learn, because it's the first software that's going to learn *you* and help you how to get work done every day...
We believe there is a shift in how knowledge work gets done, from a world where people learn systems, tools and tasks, and follow instructions, to a model where you actually find information, where you are able to take action on this information.
This is the big change with Sana, right? Sana is not just offering you insights, it's right in the core of the business process, which means it helps you actually then acting on this information and getting work done, [to] build agents and workflows and orchestrations on the fly that are specific to an individual worker's need, and then automating them for you in the background and doing this across a large amount of enterprise applications.
Multi-step workflows
Connecting to third-party apps makes it possible to set up multi-step workflows such as a monthly workflow that reviews the user's email inbox for receipts, checks them against policy, compiles an expense report for approval and then submits the approved report — and then repeat this as an automated process every month. As of today, the full list of available connectors encompasses Box, Confluence, Gmail, Google Calendar, Google Contacts, Google Drive, Google Tasks, Jira, Linear, Miro, Microsoft Outlook Calendar, Microsoft Outlook Email, Notion, Salesforce, ServiceNow, SharePoint, Slack, and Zoom. More are planned to join the list later this year.
Workday is emphasizing that Sana runs inside Workday's existing security, permissions and audit framework, inheriting the same controls customers already trust when users access their sensitive HR and finance data. Bhusri explains that this framework is founded on Workday's mapping of the people in an organization:
I do think we have to remind folks that identity, both for agents and for humans, comes out of the HR system, or the agentic system of record. Identity has to come from somewhere — what you can see from a data perspective, what you can see from a business process, or how you interact with the business process. The identity companies out there, they get their identity source from the HR system with all of our customers. So we have a better point of entry into all the other applications than anyone else does, since we are the source of identity.
User context
Joel Hellermark, SVP and GM of AI at Workday, and formerly the co-founder CEO of Sana, explains the importance of bringing the user's full context along as the AI interacts with external systems:
When it goes out and searches SharePoint, or when it goes and does an action in Salesforce, it has all of the context about you and who you are and what you've done in that system in the past, and can use that to drive much higher accuracy in the tasks that it solves...
What's interesting with this approach [is that] you can apply judgment to this too. So for example, if you ask for your parental leave policy and you have two conflicting documents, the agent will look at those documents and then it will say, 'Okay, Joel is based in Stockholm. This is his tenure, etc. This document is outdated. This other document applies to Stockholm employees,' and then it will use that document. So it can gather based on that context, it can apply judgment to a greater extent.
Sana is available through Workday Flex Credits, which is an addition to the base subscription provided you're among one of the 4,000 Workday customers on the current Terms of Service. Introducing a billing model that’s based on consumption moves in the direction of charging customers based on the outcome the software helps to deliver, although Kazmaier notes that there are limitations to applying this model to multi-vendor scenarios:
It really changes from a per-seat based subscription, per user per month, to a usage-based metric that captures the value at an outcome level. On top of that, if you really think that through, the consequence is — and you see this with Self-Service Agent, is that it actually shifts way more towards delivering the business outcome directly, than the technology.
So it's not only a shift in how you monetize a seat license versus an outcome-based metric, it also shifts the customer delivery model. You're moving from 'Hey, I'm selling you software, or I'm selling you access to software,' to 'I'm actually selling and delivering you an outcome.' That's the step change Workday has led with Flex credits...
Obviously in some areas, outcome and activity are hard to parse out from each other. There are some clear ones, like contract intelligence and case deflections. There are some other scenarios... in orchestrating work about other systems where we actually don't know the outcome, because you're working across 10 different systems, and then it's going more on an activity metric, like the [amount] of reasoning capacity you have spent on that. So there's always going to be a duality. But for Workday, our North Star is that we deliver outcomes.
My take
The sea change in Workday’s product and platform development organization over the past two years now bears fruit in the delivery of an all-new, AI-led user experience that helps this SaaS leader stay relevant in the AI era. Bhusri remains adamant that enterprises still value the deterministic certainty of Workday’s underlying system of record, but recognizes the importance of teaming this with the probabilistic reasoning that AI can layer on top.
For customers there’s now a big change management exercise ahead to introduce users to this new way of working — and a leadership imperative to calm nerves about the massive disruption to existing employee roles, particularly in HR helpdesks and other administrative roles. The Workday team had plenty to say about that, too, which we’ll cover in a companion article to this one.
But questions also remain about the mechanics of hooking up processes and agents across multiple systems. Workday has a huge asset here in its org mapping, which provides a foundation for understanding permissions and roles, as Bhusri alludes to. But I haven’t heard enough about how other aspects of business and industry context are shared between agents coming from different vendor environments. I suspect that there’s still a learning curve that needs to be negotiated here, and the realities may produce more stumbling blocks than today’s slick demos suggest.
Nevertheless today marks a big milestone that will help Workday’s customers see a way forward to introducing some of the benefits of AI into their operations.