Workday is in a race to re-define vertical SaaS, says its CTO. How's it going?

AI is forcing massive, rapid change on established enterprise application vendors, and their ability to survive this disruption depends on several factors. Will their value proposition remain competitive? Can they change fast enough? And how will they bring their customers with them as they evolve? At Workday Rising EMEA in November I caught up with Peter Bailis, Workday's CTO, and our conversation touched on several aspects of the vendor's response to these factors.

One element of what Workday sees as its unique value proposition helped persuade Bailis to make the leap from a pivotal role at Google Cloud to join Workday nine months ago. Its single data model contrasted with his experience at Google, as he explains:

I was coming from the data space. I ran the coding models and Gen AI and assistance for the data services at Google Cloud. In a [hyperscaler] cloud environment, every customer is completely different — completely different schemas, completely different data. What does revenue mean to you? It's completely different than someone else. What does an employee mean here versus employee there? It's Wild West and very, very challenging — because at the end of the day, this is not an AI problem, to identify what data means what. It's actually a business problem. You need the organizational context...

One thing that just blew me away [at Workday] is that across 11,000 customers, we have one data model and we have one process model. There's an employee, there's a journal entry, there's a payroll event. You can customize each of these, because your time-off rules are different than my time-off rules and so on. But it's a single data model, single process model.

What he describes goes beyond having a single dataset. The semantic context around that data is the real advantage for Workday, and also spans the application layer too, where processes use that context to act on the data appropriately. AI in the enterprise needs this System of Knowledge to build meaningful insights and recommend appropriate actions. Established vendors who figure out how to surface that embedded context to help structure AI queries and responses will have a significant advantage. Bailis goes on:

What that lets you do from an AI perspective is so deep. And I think that when you look at the action in AI, it really is happening at two ends of the stack. One, foundation model provider. Two, system of record and vertical applications. I think that we have an opportunity at Workday to really redefine what vertical SaaS means, what great looks like in this agentic era.

It's clearly a strong asset for any established vendor, especially when as in Workday's case the native dataset extends across both people and finance. But does it translate into an advantage for Workday's customers, whose data and processes are spread across many different vendor landscapes?

Security model

This is where a second part of the value proposition comes in — Workday's role in defining access permissions to data and processes. Workday's history as a cloud vendor means it already had to build a robust security model to reassure enterprise customers that their data was safe in its infrastructure. When combined with modern integration technology, this security model for access to its own applications is easily extended to access data across multiple applications. Bailis explains:

When you were trying to get people to move your people and money data to the cloud, it's like, this thing has got to be secure. You want to run it through an app... You need to really build that trust.

The evolution is, with some of the more modern technology, like zero-copy analytics, we can now offer access to this data, but in a secure way. And I think the opportunity that I see Workday really leading with and customers really appreciating, is that when you get the security model right, and you can adopt open standards — everything from our API coverage, more REST and GraphQL, down to MCP — suddenly that [integration] overhead is much, much lower.

This shows up, he goes on, in Workday's partnerships with the likes of Okta and Azure, where Workday serves as the directory not only of an organization's people but also of their place in the organization, which in turn maps to their data access rights and permissions. The next step Workday has taken is to extend that permission model to the universe of AI agents, through the Agent System of Record it launched last year. While agents are not a like-for-like replacement for people, he explains that it's helpful at the moment to think of them performing familiar roles and workflows roles because that makes it easier to understand where they fit in. Similarly, extending the org chart to include agents helps enterprises manage and control the data and processes they can access, in the same way they manage access by employees. He elaborates:

I'm going to have these agents going and running skills. They're running across all the different data. They're going and taking multi-step actions, sometimes autonomously. How do I control who has access to what?

By using the same organizational principle around the org chart, which is how you do it for people today — 'I'm in the engineering organization, I have these grants and these permissions' — we can do that in a way where an agent has almost the same [access]. It's not just personifying the agent from a conceptual perspective and a cognitive perspective for people, it also fits in the security framework way better. That's really the jump that I see, it's mapping to the established security framework in the directory inside of the Oktas, the Azures, the identity providers of the world.

This permissions model combined with the data model provides a strong value proposition that should help persuade customers to stay with Workday, especially after its acquisition of Sana, with the intention of becoming 'the front door to work' across the enterprise. But Bailis concedes that the vendor still has to work hard to stay competitive at every layer, while remaining interoperable with other platforms. He tells me:

In the end, there's going to be many front doors. There's going to be many systems of record. There's going to be many data lakes. We want to be in all of them, right? And it's customers' decision. Do they want to use Sana? Do they want to use Gemini Enterprise? Both of these should work.

Our job at each layer — application, agent, platform, data — is to make it so good that you pick Workday. But if you want to use it at a different level, knock yourself out, right? We should not be the arbiters of our customers' data on the end. It's not our data, it's our customers' data. We're here to serve our customers, and I think, let the best products win.

Keeping pace

What about Workday's ability to adapt and change fast enough to keep pace with all of these challenges? There have been big changes in how the vendor develops its products to focus more on a cross-functional platform, together with a move away from a seat-based business model to more of a consumption model. Bailis says he's found the company "very adaptable and extremely hungry to adapt." He elaborates:

On my personal teams, I run the AI teams of the company. I also run the UX teams — because these are one and the same. A lot of the platform work is saying, for all of these user experiences, all these tasks, all of these journeys that we allow customers to go through, how do we make that into a great set of APIs and a great platform layer, so our AI teams can build on them? And the same technologies our AI teams are building on, make that available to customers.

It takes a while, and organization change is obviously hard, but I think people — especially the core technologists and the product visionaries that got us where we are today — there's a lot of hunger to evolve and adapt.

That readiness to adapt is a tribute to the culture that co-founders Aneel Bhusri and Dave Duffield instilled in the company from the start, he believes:

Aneel and Dave founded the company saying, 'We want this to be a place where people have fun, where they take care of each other, where we really value innovation.' It makes a lot of the conversation on where we need to go much less territorial, much less, 'We did it this way before,' much more curious, much more kind, and I think a little more benefit of the doubt, 'Let's see where this goes.'

We talk about the greater good as one of our leadership principles, right? And people just really show up for the greater good here, because of the hard work that Aneel and Dave and our workmates have done for 20 years.

He senses a similar eagerness among customers to explore the new capabilities Workday is bringing to market. At the same time, they remain free to choose the speed at which they move, with Workday's licensing giving the option to switch into and out of consumption-based pricing at any time. Bailis says:

You have the flexibility. You're using our APIs, and you decide in six months, I'd actually really rather integrate via Agent-to-Agent protocol with the Workday self-service agent or the Workday payroll agent. You can go and do that. It's all on one, also commercial, model, which is included in your base SKU, and it's pay-as-you-go. There's no commitment that you have to go and make to any individual agent or platform capability. It's all in one model. And so ... when you look at the constellation of vendors that a CIO has to navigate, we're kind of easy to deal with, because you're already a Workday customer, or you should be a Workday customer, and how you want to plug us in, it can evolve.

Where that ends up in terms of the future shape of Workday product remains to be seen, he adds:

The customers do the work of advancing what SaaS looks like. It's our job to help provide that platform. That's why I'm not in the lab anymore. It's so much fun to actually just build the stuff, and with software you can build anything.

My take

There have been big changes in Workday's product and tech leadership over the past year-and-a-half, including the arrival of Gerrit Kazmaier as President, Product & Technology, Bailis as CTO, and others. These have been accompanied by rapid changes in how the vendor builds its platform and products, along with a flurry of acquisitions such as Sana, agent builder Flowise, and most recently agent integration platform Pipedream. At the same time, Bailis points out to me there's continuity too, with individuals who joined with the acquisition of enterprise service bus vendor Cape Clear almost two decades ago still playing key roles in the tech platform today. The result has been an impressive pace of innovation that shows no sign of slowing just yet.

While Workday scores well on this metric, bringing customers along — and partners for that matter — is a different challenge. There will always be eager early adopters, but shepherding the majority of Workday's customers into this new era of enterprise tech while retaining their loyalty will be a big test. When I asked about this at Rising in Barcelona, both Gerrit Kazmaier and Workday CEO Carl Eschenbach spoke about the importance of fielding consultants that can understand the customer's business and advise on how to deploy the technology to best effect. Eschenbach told me:

Every customer I've talked to, they've asked for exactly what you said, Phil, 'How can you help us on the people side of this transition to AI? Not the technology, because we know that will ultimately work.' So we're going to build service offerings, and we already have partners out there that do this as well.

This, like the future shape of the product itself, remains a work in progress, part of a collaborative journey for both the vendor and its customers.