Executive Summary:

Close out 2025 with a front-row seat to the future of

enterprise AI as Workday CTO Peter Bailis reveals

how one of enterprise software's most successful

platforms is reimagining work itself. In this season

finale of The Enterprise Edge, Bailis unpacks

Workday's bold "compound AI systems" architecture

- a sophisticated approach that's transforming how 11,000 companies worldwide manage their most critical resources: people and money. From introducing "agent systems of record" that give AI identities in your org chart to processing contracts at superhuman speed, Bailis explains why Workday's laser focus on HR and finance workflows positions them uniquely in an increasingly crowded market. Whether you're evaluating enterprise platforms, curious about practical AI implementation at scale, or simply fascinated by how distributed systems thinking translates from Stanford classrooms to managing billions of daily transactions, this conversation delivers the rare combination of technical depth and strategic insight that only a Cal-educated, Stanford-teaching CTO can provide. Hear why focus still matters even at hyperscale - and how the company governing your work location, manager relationships, and time-off approvals is betting everything on sticking to its core. Stream it now and be sure to LIKE, SHARE and COMMENT!

​

Transcript:​

Mark Vigoroso (00:01.875)
Greetings once again, everyone. This is the Enterprise Edge podcast. And this is Mark Figueroa, founder and CEO of the Enterprise Edge. I am thrilled to be with you on a Friday in mid December. I am joined by Peter Bayliss, CTO at Workday. I'm very happy to have him with us today. It's been a trick to coordinate calendars, but I am thrilled to have him before the year closes out. And this will in fact be the final Enterprise Edge podcast.

of calendar year 2025. So we intend to end with a flourish, which is what we will do. But before we jump in, let's give Peter a proper welcome. Peter, thanks for being here.

Peter Bailis (00:45.535)
Thank you so much for having me and delighted to close out the year, hopefully on a high note.

Mark Vigoroso (00:51.719)
That's right. That's right. We'll give them something to talk about over the holidays. Well, great. Well, guys, thanks for joining us, guys and gals. We will have a very short period of icebreakers, get to know Peter a little bit, and then dive into the matters at hand with regards to what Workday is doing and some of the areas where they're leading from an innovation and product development perspective. And then we'll end with a speed round, a little bit of fun before we wish everybody

a happy and healthy holiday season. So let's get started. So from a icebreaker perspective, I've done a little bit of research on Peter here and Peter, think you're based in the obviously San Francisco area. I think you have some pedigree with Stanford. I think you earned your PhD at Stanford, is that right?

Peter Bailis (01:45.434)
I did taught at Stanford for a years as faculty, but I was on the other side of the rivalry actually for grad school at Cal.

Mark Vigoroso (01:47.441)
Okay. it's faculty.

Mark Vigoroso (01:53.575)
That Cal, gotcha, gotcha, gotcha, gotcha. So you've had some academic experience in some of the highest sort of, call it technology inclined learning institutions, at least in our country. And now you're leading technology thought and execution at Workday, which is one of the sort of the success stories in the modern age when it comes to enterprise applications and...

the rapid evolution of enterprise apps in the AI era. And I'm curious when you think about that juxtaposition, you think about your academic experience, excuse me, and you think about now your sort of corporate experience, what are some things that you've learned in the belly of workday that academia didn't or couldn't prepare you for?

Peter Bailis (02:54.771)
Well, there's a lot different between, you know, academia and, you know, enterprise software like Workday. But I think the core commonality for me is the centrality of data. I used to do a lot of research in data management and systems for machine learning centered around data. And when you, what I've been pleasantly surprised to learn coming to Workday just about seven months ago is, you know, Workday is kind of like a big database.

Mark Vigoroso (03:01.702)
Yep.

Mark Vigoroso (03:08.242)
Mm.

Peter Bailis (03:22.967)
And it's got the, it's got all the data around people and money for 11,000 customers worldwide. And the magic is being able to govern and manage that data really well. And then also run the transactions on top of that data. And what's funny about the transactions is that in a lot of, mean, almost every application today has a database behind it. And the transactions on top of the workday database are the ones like literally define what work means. Like.

whether I'm approved to take time off or not, or where my work location is, or who my manager is, like it's a system of record. And it's very interesting to be inside of a system of record for such huge enterprise scale. And I think that the just magnitude of impact that you can have at that scale is one thing to read about in a textbook or to teach in, you know, introduction to databases or even grad databases, very different to see, you know, like you've got.

You've got your whole world in your hands, right? When you've got that database of people and money, it's a super, super important data set to keep for seeing and also to build on top of.

Mark Vigoroso (04:26.31)
Yeah. Yeah. And you mentioned, Peter, you mentioned the research that you've done or been involved with around data and I think, you know, databases, but also what's been called distributed systems, if I'm not mistaken. And I'm curious, what would you give us, maybe a couple sentence explanation of what's the significance or importance of distributed systems?

Peter Bailis (04:41.046)
Mm-hmm.

Peter Bailis (04:52.052)
Yeah, it's a great question. One of my favorite descriptions of what a distributed system is, is by someone with the Turing Award in distributed systems, Leslie Lampert. And his definition of system is a system in which, you know, the failure of a computer you didn't even know about can render the entire system inoperable. So when you think about building software, distributed systems are a special type of

Mark Vigoroso (05:15.442)
Mm-mm.

Peter Bailis (05:21.492)
software where the two hard problems to solve are failures and then communication latency, like how long it takes to talk. And it turns out that, you know, especially as programmers and engineers, right? Like we think about the world in very synchronous terms. We experience it in, you know, live full feed format responded to the real world. And suddenly when you've write programs where they may not hear back for a long time, or they may not hear back ever if another note is down.

How do you make systems that are resilient to that type of behavior and know how to tell the difference between a failure and between communication latency? It's a, it's really hard. And it's kind of incredible that these are real problems that happen all over the world, you know, billions of times a day messages being delayed or dropped. And a lot of times servers go down and the fact that software runs as well as it does today, it's kind of a testament to the fact that the distributed systems concepts on how to layer over that and build abstractions on top of those, you know, sort of.

unreliable networks, unreliable machines. It's sort of a marvel of modern engineering.

Mark Vigoroso (06:24.668)
That's fascinating. Well, it's maybe we'll, we'll use that as our segue into some of the the business oriented questions. But I think, you know, we could certainly spend hours on that alone. But you know, there's, there's this idea that and I think you've done some work in this area as well, Peter, around the idea of compound AI systems where, where you have

multiple sort of components, right? To try to simplify this, right? You've got multiple AI components that are inter-operating. They're working together rather than sort of a single monolith. and I mean, more from my seat, you know, I'm working with a lot of different enterprise software companies and they're at various stages of developing and innovating and creating products in this era. And

And I believe, and you can keep me honest here, Peter, at Workday, I think you guys made a decision early on to sort of architect compound AI systems that, certainly talking about HCM, human capital and finance type workflows, right? But compound systems that rely on multiple interrelated components rather than.

sort of a single sort monolithic model. So first I wanna test that, make sure, is that supposition correct? that, would you say, that maybe an oversimplified version of the approach that you guys are taking?

Peter Bailis (08:00.327)
I think it's really well put. And the reason for this kind of compound approach is that, you know, models are really good when you give them very specific instructions. They've gotten much better in terms of what you call, technical terms, it's not funny, instruction following.

Mark Vigoroso (08:02.821)
Okay.

Mark Vigoroso (08:10.204)
Mm-hmm.

Peter Bailis (08:15.734)
in the last, um, you know, a couple of years, but the complexity of the instructions, any model or any, you LLM call can really follow is, is getting way better, but it's still relatively small. So when you want to go and build a complex system for automating something like payroll or making something like scanning contracts, I've gotten, figuring out payment terms. It's really hard to do that with one single.

Mark Vigoroso (08:15.857)
Yes.

Yep.

Mark Vigoroso (08:27.619)
Mm. Mm.

Mark Vigoroso (08:39.73)
Hmm.

Peter Bailis (08:45.138)
model or, or single prompt. And what you see in a lot of the systems, including ones that we build is you're kind of giving a model or what you call an agent access to a bunch of individual tools it can call. So in the case of say our contract intelligence agent, I might have something to go and scan the PDF. I might have a special tool to go and read a table because it's really hard to read tables. It turns out even with modern models, I want to have one to access my database of contracts. And I want to have one to do.

Mark Vigoroso (08:47.623)
Yeah.

Mark Vigoroso (08:59.89)
Hmm.

Mark Vigoroso (09:09.522)
Hmm.

Peter Bailis (09:14.558)
redlining. So to redline this and you know, the first four are kind of traditional, almost like deterministic scan this PDF, giving the text back the redlining one though, that's like, that doesn't sound like, you know, a simple calculator and what you actually see in many cases, the tools themselves are also models. And so if you unpack, you know, the full tree essentially of what one of the agents is made of often it's made of other agents.

Mark Vigoroso (09:17.138)
Hmm.

Mark Vigoroso (09:21.778)
Mm. Mm.

Mark Vigoroso (09:40.85)
Hmm.

Peter Bailis (09:42.856)
And even today, the best models, you really can only handle a handful of tools at a time. So the only way to scale the task complexity is to essentially introduce hierarchy, which is one form of a compound system.

Mark Vigoroso (09:58.195)
Hmm. Yeah, it's fascinating. I wonder if we can, well, let's do this. If you think about the actual domain that you think about probably most of your days now is given where workday is, is sort of the finance and HR realm, right? And so you're thinking about the needs and requirements of sort of the

Excuse me, HR leaders, the finance leaders, the obviously the technology and the information system leaders. Excuse me, I'm gonna take a drink of water.

Mark Vigoroso (10:43.026)
And so my question is sorry, I'm gonna edit this part out because I'm we got a This is a coughing break I'm getting over a getting over a nasty cold this week. I thought it was gone, but no worries The beauty of post-production you'll you'll you'll never know

Peter Bailis (10:48.728)
Hey, I'm gonna do the same. I'm cough too, so... Perfect, I need my drink of water.

Peter Bailis (11:02.929)
you're all good. We're in the same boat, my friend, so.

Yeah, exactly.

Mark Vigoroso (11:12.74)
All right, rewind. Okay, got it. So in this world of the CFO, there's a lot of requirements that I think exist in those offices that have to do with transparency and auditability, behaviors that are very deterministic, you might say. There's not a lot of margin for creativity or error, if you will, or subjectivity.

When you're dealing with things like payroll, right? Or money, right? Money changing hands. It's not a lot of like, you know, margin for error. And so when you think about the fact that AI, a lot of the models are probabilistic and you're trying to apply value proposition to very deterministic environments. How do you reconcile that when you're building solutions, right?

Peter Bailis (11:58.806)
Thanks.

Mark Vigoroso (12:11.398)
when you're trying to apply this rapidly evolving technology category to a highly deterministic environment, very little margin for error, is that something that, I mean, it's probably a lot of what you grapple with, but I'm curious of your perspective on that, and maybe there is a simple way that you think about it, but I'd love to hear your thoughts on that.

Peter Bailis (12:31.358)
This is one of the hardest parts about building AI for people and money is that on like a web search.

Mark Vigoroso (12:35.931)
Yep.

Peter Bailis (12:37.845)
where even a lot of times the right answer, what's the best web page to go and show you? It's completely subjective here. You know, I'm going to close my books. They better be, they better be correct. Or I'm going to go and, make a employee, you know, data change that better be correct that someone's livelihood. Right. And ultimately that trust is also kind of core to who workday is. We started, you know, 20 years ago, the radical idea you should run your HR and finance software in the cloud, starting with HR.

Mark Vigoroso (12:43.111)
Yeah.

Mark Vigoroso (13:06.695)
Yeah.

Peter Bailis (13:07.433)
And that that was radical. know, people were like, I need to have an escrowed version of the source code. was like so, heretical. So trust was a big deal and still is a big deal for us in terms of why our customers choose workday. So when you build AI, you know, how do you, how do you work around that? Where these models are not perfect and they ultimately, you know, our operating domains when they need to be right. I think there's really two key things that we, that we focus on. One is.

Mark Vigoroso (13:12.86)
Yeah.

Mark Vigoroso (13:20.615)
Boom.

Peter Bailis (13:35.882)
recouple the models with the core business processes inside of the platform. So when I want to run payroll, for example, there is a payroll process and what payroll means my company is defined by my company, it's configured, it's very specific. And a lot of the AI comes down to how do I make it easier to understand which business processes to run in the first place? And second, how do I go and orchestrate them? So composing and combining many of these. And so when you look at what we've done,

Mark Vigoroso (14:01.884)
Hmm.

Peter Bailis (14:05.681)
In payroll, as an example, you're essentially allowing a payroll administrator to go and perform bulk actions and go and run their payroll and go and make, say, minimum wage compliance related changes very quickly. But they're still in control. And the core substrate is really still running those core business processes that are deterministic. And you can think about the job of the agent to go make it easier to get to that outcome faster, but leaving the human in control.

Mark Vigoroso (14:17.97)
Hmm.

Peter Bailis (14:33.215)
The other part about this that I think is really interesting is we spend a lot of time, even just on the compliance side. So when you have a lot of the take financial audit, for example, a lot of folks, you know, you'll hire an audit firm to go run an audit once a year, and they look at a handful of records and then they're done. Well, you know, the tests that an auditor would run, you know, you can actually start to, you know, create deterministic versions of those that run all of the time. And if you can combine that with agentic logic that determines what tests are run and, and, and, and when.

And then takes those tests results and can give you a synthesized explanation of what's going on to avoid the equivalent of kind of like alert fatigue, but for an accountant, you know, that's like a really powerful solution. And you're kind of unlocking this use case where who wouldn't want to have an audit. Every single day of the year from a financial compliance perspective, it was just cost, you know, prohibitive to do so otherwise. And with an agent, I do have that audit. can run it literally anytime I want.

Mark Vigoroso (15:12.646)
Yeah.

Mark Vigoroso (15:22.642)
you

Mark Vigoroso (15:28.242)
You're right.

Peter Bailis (15:33.173)
And I have an agent going and interpreting those results and making sure that I understand what could be a deterrent of audit results coming back and help me prioritize the ones that matter.

Mark Vigoroso (15:33.34)
Right.

Mark Vigoroso (15:46.173)
Yeah, no, it's fascinating. mean, think we are seeing, I mean, you just nailed one of the use cases that I think I'm seeing in a lot of different forms, which is decisions have been made historically to do things or not do things based on the level of effort and level of return, right? mean, whether or not you realize it, I think most business people make cost benefit decisions all the time.

you know, almost without knowing it, right? And that logic, you know, people are trained to do. And what I'm finding is that, you know, in a good way, companies are able to, and people, sort of unwind some of the presuppositions about...

about cost, right, to your point. Well, we would never do that because it takes too much time, effort, money, resources, and therefore we're just living with the result, which is we only get that particular output once a quarter. It's like, well, what if we were to challenge all of those underlying assumptions and say with a GENTIC, you can do that with minimal cost, minimal effort, right? It's like, okay, then that means we could get this particular report or outcome or framework every week.

It's like, okay, well, what does that mean? And it creates this sort of cascade of almost like new white space where you can say, okay, well, what would that enable us to do? What value would that create? And that's where these sort of ROI, some of the ROI models that when I talk to end users and companies that are actually trying to implement and scale AI, they're not often thinking about the ROI.

Peter Bailis (17:31.253)
Okay.

Mark Vigoroso (17:33.907)
in all of those multiple cascading effects, right, what did you just unlock by being able to perform this audit flow or whatever on a weekly basis instead of a quarterly basis? What does that mean to you financially? And they don't always logically just come up with this, well here are the 17 layers of value. It's often understated.

And I think it's fascinating because in the world of finance, it has been so labor intensive for so many decades that what you guys are doing, and maybe this is a question, a roundabout way of asking the question is, you helping your finance stakeholders see some of that value?

that might accrue in areas that might not be readily evident right away, or are they seeing it on their own? Or how are they thinking about ROI in the first place? Maybe that's an easier way to ask the question.

Peter Bailis (18:37.539)
Yeah, it's a great, you know, I think I'd say the industry. Except in very select domains. Most folks are figuring out how do I actually get any ROI from, from AI and the arc I think we're on is people saw this happen and you know, chat GPT and the consumer era.

Mark Vigoroso (18:42.993)
Yeah.

Hmm.

Mark Vigoroso (18:49.809)
Yeah.

Peter Bailis (18:56.914)
And it's incredible for knowledge search and public information retrieval. And these models are trained on the public internet and cancer or every book cancer, any question. And then you bring it to work. And like, there's a massive fall off because, you know, these models know nothing about my data. they're not super good at a lot of back office tasks or knowledge work tasks. There's a recent benchmark that came out this past week, looking at even just basic.

Mark Vigoroso (19:12.775)
Yep.

Peter Bailis (19:23.124)
you know, data manipulation and back off like what, what a knowledge work would look like. And these models are so, so early there, uh, even the frontier models. And then I think what a lot of people actually built from a market perspective was just like simple prompt wrappers. And it was not very sophisticated and the results were as a result, not great either. So what do you do? Right. Uh, our approach is really simple. We look at the core functions that we serve across people and money.

So for example, in finance, you've got a planning department and there's a bunch of people in FP &A. You say, what do they spend all their time doing today? And okay, they do a ton of work, you know, running different types of analyses and building forecasts and doing sensitivity, you know, analysis and scenario planning. And what we basically do is we look at those like the day in the life and we say, okay, we own the business process for this, for say a scenario plan. If we took that business process and the core data inputs and then the core

Mark Vigoroso (19:53.842)
Yep. Yep.

Mark Vigoroso (20:06.193)
Yeah.

Peter Bailis (20:18.663)
You know, business processes we have built in to run a, a scenario. How do I go and make that way easier to go and set up? Way easier to go and scale. And in the case where there's compliance angle, just, you know, easier to drive more correct results. And what that ends up doing is, you know, for those FPNA folks who are just spending tons and tons of time and spreadsheets building out scenarios or our, you know, data modeling. They are actually freed up to do a lot more work. Right. And so the types of ROI that we see, which we work a lot with on our customer, with our customers.

You know, you get a 30 % reduction in time doing manual analysis. It's not like that FPNA professional. It's just, you know, you know, kicked back and, you know, hang on the break room. They're actually spending more time on what they can uniquely do, which is thinking about strategy versus building yet another scenario breakout or tweaking another parameter in a model. And that's where I think the bull case is on at least AI for work. We kind of fit the work we can do into the hours in the day. There's always more work to do.

And you take a business process lens on this and you say, we're going automate the business processes or major parts of those business processes. You actually free up the parts that are less mechanical and much more around what people are actually really good at, which is frankly bringing a lot more context that's not in the data to the analysis. So for that scenario, the FP &A analyst is great at thinking about what are the scenarios I should run? And if they spend less time on, on, you know, actually building them out because the model is going in.

creating the scenarios on their behalf and they're interpreting the results, that's like net positive and ultimately leads to better decisions.

Mark Vigoroso (21:52.825)
I love it. love it. Well, that's a good bridge to the next thought here, which probably gets to a lot of decisions you're asked to make, you and your team. When you think about building product, and you said 11,000, right? You have 11,000 customers about, is that right? Yeah, so obviously a massive base of data, very sensitive data, right, around employees and finances.

Peter Bailis (22:14.471)
Mm-hmm. Mm-hmm.

Mark Vigoroso (22:22.578)
very sensitive data and the models are evolving very, very quickly. The sophistication of reasoning that's capable in some of these tools is improving rapidly. And so with that rapidly moving landscape, how do you think about

Peter Bailis (22:32.051)
you

Mark Vigoroso (22:45.926)
Decisions like intellectual property like what should we be building as? Workday as opposed to what should we be maybe acquiring? Right or maybe what we should be partnering to gain access to Right. It's classic build versus buy versus partner right against a rapidly evolving and shifting landscape and that's part of the probably the challenge and the thrill of your job, right, but

Peter Bailis (23:13.375)
Yeah.

Mark Vigoroso (23:14.615)
Tell me how you're thinking through that and how do you navigate that?

Peter Bailis (23:19.58)
And we always try to think about what are we best at and where do want to maintain that edge? And for us, you know, we're the system of record for people and money. And we also have a ton of context around how that's used. So it's not just the database, but there's a trillion transactions that say, what does work mean for customer A versus B versus C? We build all of our AI around that premise and it's all our customers data. So significant amount of work on responsible AI, data control, so on.

So that, you know, it's really your data, you know, your business. And when we think about what do we, what do we build? think there's enduring value in embedding the AI and infusing AI directly in that process model. And that's why, you know, we have a lot of great partners who will build on the Workday platform. We're opening up more of the platform to make it easier to build on Workday with standard APIs. When you really want to go deep, deep, deep into something like employee self-service, you know,

Mark Vigoroso (23:59.987)
Mm.

Mark Vigoroso (24:09.042)
Hmm.

Peter Bailis (24:14.067)
wants to take time off or employ, know, or alternatively on the case management side, I want to go have, you know, HR specialist manage caseload better. Like that's just such a core part of our business. We have so much data around that, that it's, you know, our right to win. And then by embedding that AI so deeply and infusing it in the process, how does the model get better and more capable than the whole flywheel? It just gets, it gets even better, right? We have done a fair number of acquisitions this year. We've acquired Paradox, which is a frontline recruiting solution, which is really cool.

Mark Vigoroso (24:14.438)
Hmm. Hmm.

Mark Vigoroso (24:26.918)
Yeah.

Mark Vigoroso (24:35.271)
Got it.

Peter Bailis (24:43.123)
Sana, which is an incredible knowledge management and learning platform, flow wise, which is an agent builder. I think in each of these cases, you know, it's really about, you know, what lets us move faster towards our vision. Paradox, you know, we know we want to do more on, on recruiting and AI for recruiting and their close partner. And they were in many, many accounts and they said, we love this. It's such an incredible platform. Awesome for Sana.

really incredible user experience, leapfrog experience for our users that we'll be bringing to the Workday platform, which is an area we get a lot of feedback on. And this is an area where we can just have Scandinavian design ethos infused into the Workday platform and bring it to everyone. Very excited about that. And on partners, think Workday's historically been very closed as an ecosystem, owing to its roots where going to the cloud was heretical, right? Now that we're in a world where we're part of a constellation of SaaS providers,

Mark Vigoroso (25:17.874)
Yeah.

Peter Bailis (25:33.957)
AI tools depend on other AI tools and really, you it's, you know, you want to pull together data from many different places. We want to make it so you can get access to your workday data while applying that, you know, security and governance model. So we're actually investing heavily, heavily in an open ecosystem using open standards applied in our governance model. And we've actually seen the number of people building on workday, both from a developer population perspective and just from partners building businesses on workday, like meaningful scale businesses has gone up a bunch as well.

So it's all about the core. It's like, we'll be the center for people money, the transactions around those. And we'll try to give you the best agents for people with money. But if you want to build your own or you're the partners, that's great too.

Mark Vigoroso (26:17.958)
Yeah, you know, it's interesting. mean, even though the landscape is different, I mean, the logic that you have always applied to the classic build by partner decisions is the same, right? know, stick to your, stick to what makes you you, stick to what makes you different and stick to what makes you, you know, who you are and who you've been. And I think it still applies even if I'm interpreting what you're saying correctly. I think that's good wisdom.

Peter Bailis (26:40.082)
Mm-hmm.

Mark Vigoroso (26:44.806)
Well, great. Well, Peter, this is awesome. And like I told you, it's going to go fast and it did, but we're not done yet. We're not done yet. We're going to, we're going to have a little bit of fun with, with a speed round. And we're going to ask hopefully very simple questions and, we're going to learn a bit about you as well. All right. Let's see if we can get through five of them. All right. First one. So you've most likely participated in, or maybe even hired.

Peter Bailis (26:58.322)
you

Mark Vigoroso (27:14.034)
some engineers yourself. What is the one question in an interview that tells you more about a candidate than their entire resume?

Peter Bailis (27:29.284)
I think asking someone to walk through a recent technical problem they've solved end to end tells you tons. And it usually makes for a very fun conversation. You can make that a three hour interview if you want to. Not that I've gone that long, but it, you know, it's very interesting because you get a sense of what are they, where do they start answering that question? You know, how do they talk about the customers that they're working with? How do talk about the challenges they have? It's, it's an incredible jumping off. And I think on surface level, it's kind of a very easy question to answer.

Mark Vigoroso (27:35.751)
Mm-hmm.

Mark Vigoroso (27:40.506)
Yeah.

Peter Bailis (27:57.907)
but it actually contains multitudes when you really think about the many layers of it in terms of why did you do that and why did you do that? You know, we talk about in, you know, post-mortems like, you know, five Y's, ask Y once, ask Y again, ask Y again. I think that's like a great way to understand how an engineer thinks and ultimately, you know, what do they prioritize? What do they care about? And, you know, what kind of taste do they have in terms of making the inevitable hard decisions? Engineering is all about trade-offs. So how do they navigate trade-offs?

Mark Vigoroso (28:02.535)
Yeah.

Mark Vigoroso (28:23.046)
Yeah. Yeah. I love that. I love that. Well, I had an interviewer once asked me, he knows who he is. I don't know if he's listening. He had he had me describe in great detail how I make a ham and cheese omelet. And, and, and

Peter Bailis (28:37.734)
Mmm.

Mark Vigoroso (28:40.326)
He asked me about another half dozen wacky questions like that. After the fact, he circled back and he told me exactly why the logic and what he learned from how I described how I made an omelet. And it was mind blowing to me. So be aware if you ever get asked a question like that, they're listening very carefully to the words you say, the words you don't say, things like that. anyway, let's see. Okay, if you could make one technology.

Peter Bailis (29:01.05)
Mm-hmm, exactly.

Mark Vigoroso (29:11.379)
practice mandatory, like whether that's in development or design or testing. If there's one practice that you would want to make mandatory across all of either your team or the industry that you think would move the needle or significantly improve quality or speed.

practice that you think is that.

Peter Bailis (29:40.023)
Hey, Mark, I think your audio kind of, I don't know if your audio switched off, but something.

Now I can hear you, yeah. So I caught you at the one technology piece, but I think I...

Mark Vigoroso (29:50.401)
Okay, I wonder what happened. Let's see. I see what happened. Okay, we'll rewind So I guess so I'll just rewind so could you Recommend or if you if you could choose one technology or a practice that that you'd want to make mandatory across your team or maybe broader than just your team maybe the industry that you think would Move the needle on quality or speed or efficiency

or robustness in terms of the quality of the software? Is there anything that you've learned or you've experienced over time that just can't be missed in terms of the discipline in software development?

Peter Bailis (30:29.138)
One thing I'm really bullish on that he makes life better for everyone using software is embracing some of the new agentic protocols like MCP. And I would love if every SAS vendor had a great MCP coverage for their for their platform. It would make life way better for for end users. And I think I think it's happening. But if I could wave a magic wand and have it happen tomorrow.

Mark Vigoroso (30:42.493)
yes.

Peter Bailis (30:58.588)
That'd be incredible because the coverage is still not nearly where it needs to be for a lot of basic stuff that you'd want to do, crossing your CRM with your email, with your calendar. The models are getting almost good enough to do this type of stuff and it's not there yet.

Mark Vigoroso (31:10.813)
Yeah.

Mark Vigoroso (31:14.13)
Yeah, that's a good one. That's a good one. All right, let's see. When you think about the CFOs of the companies that you're working with and maybe the HROs as well, if we fast forward three or five years or so, do you see these guys or gals, are they gonna evolve to be managing labor?

you think about labor in the most inclusive sense, right? Human labor and digital labor, right? Are they going to require different skills, expanded skills, augmented skills to manage this and sort of orchestrate labor across these sort of human and digital lines where you've got agents executing workflow. It's almost like you're auditing the performance of your

Peter Bailis (31:40.914)
Okay. Okay.

Mark Vigoroso (32:09.807)
human workforce and your digital workforce, just like you would in any kind of workforce, full-time contract, whatever. Do you see that role as a leader, as an executive at these companies? Do you see them evolving at all? Is that a reach or is that really gonna become the remit of these traditional sort of business leaders that they're gonna have digital labor that they're gonna be accountable for?

Peter Bailis (32:37.395)
I think it's very likely you have something like that.

And for no reason other than a lot of systems are set up for human access. You know, I have a, I have a identity and I have permissions attached to that identity and that's tied to my role and it's tied to my org chart and it's tied to a lot of elements of, how people operate today. And actually one of the things we've been working on at Workday is called the agent system of record, which is just a very, you know, nice way to basically take the data model you have for people in Workday.

Mark Vigoroso (32:46.013)
Hmm.

Peter Bailis (33:09.327)
And you can now have an agent applied there as well. So an agent in your org chart, an agent with permissions, an agent with an ID or an ID you bring in, but you can embed there. And I think that's in some ways why agents are so powerful as a concept, because they sort of, you know, personify the AI. You know, they're like, what, what should I do with this chat thing? It's like, how do I conceptualize that? Well, when I make it like a person, suddenly I can like think about how to wrap it in my organization. And the bet with agents is sort of a record is that, Hey, we manage all the people.

Mark Vigoroso (33:15.836)
Yeah.

Mark Vigoroso (33:35.921)
Mmm.

Peter Bailis (33:39.238)
you can actually manage agents with the same data model. And I think there's a high probability that concepts from there become really important in the future because I don't want to have an agent execute with full system access. I want to have it execute with specific access. And the best way to govern access today is largely through human-based identities or some evolution of a human-based identity over time.

Mark Vigoroso (34:03.921)
Got it, got it. All right, two more, Peter, we'll let you go. But one is curious on this one. Workday, speaking of what you're great at and what you've done and a strong foundation around managing people and money, right? Is there a day in the future where that expands, right? If you think about adjacencies, natural adjacencies to that, right? Whether that's additional functional areas within

your typical corporation that would make sense. I don't know, I'll throw one out, like maybe supply chain as an example. Is that even on the radar when you start to think about long range planning and growth? Or do you think this is pretty much, you you guys have established what your identity is, basically looking at, you know, HR and finance workflows.

and automation pretty much in perpetuity. Any thoughts there?

Peter Bailis (35:09.486)
I think we're just scratching the surface on people and money to be candid in terms of even market share and, you know, the, upside opportunity in these domains. So we'll never forget that core, but look, customers really

Mark Vigoroso (35:13.382)
Yeah.

Peter Bailis (35:24.665)
Want a handful of solutions and a lot of people come to work day and run both their people data and their money data and their planning, you know, and then a range of other stuff. Like now their contract management solutions, like they run, you know, work day offers it. You have a great suite of products. People tend to tend to adopt it. There's great attach across our customer base. So I would not be surprised if we do expand in the future, but again, it's a balance. Don't forget the core and.

Mark Vigoroso (35:44.615)
Yeah, yeah.

Mark Vigoroso (35:50.183)
Yep. Yep.

Peter Bailis (35:52.665)
We're ambitious, right? We want to grow. think this is a great technology base to go and build off of. And frankly, I think that, you know, there's not a lot of other people at our scale who I think have the, you know, one scale to team three ambition to go, to go as big as we are, especially when think about those back office functions.

Mark Vigoroso (35:54.567)
Yeah.

Mark Vigoroso (36:12.997)
Yeah, for sure. No, got it, got it, got it. Last one, then we'll part for the day. But I think this one's interesting because you think about workday and the role that you're in, everybody likes to make comparisons. I think the human brain likes to compare. Everything is like something else, right? There's nothing that's really unique. And I'm curious when people compare workday

to other providers in the market, whether they're SAP or Oracle or UKG or other platforms that are tackling HR and finance workflows. From your spot, are they asking the right questions when somebody in sort of a buying group, whether it's a technical or a business buyer, they're trying to figure out how Workday

solves their problems distinctly different and valuable than maybe some alternatives. Are they asking the right questions? Are there questions that they're missing? Are there questions that you would advise them to ask to get to the heart of the true identity of Workday both now and going forward if you follow the question.

Peter Bailis (37:37.701)
We have a great comparison set and I some great competitors. The reality is when you ask the question, where's the focus? The best solution for people and money really falls out. We really do have, when you look at our comparison set, the focus at scale, that's pretty unmatched in these areas. And I think that comes out in the product and comes out in the platform and comes out in a lot of the big bet replacing when you look at our AI strategy, when you look at what doing for.

Mark Vigoroso (37:41.115)
Yeah.

Peter Bailis (38:06.564)
frontline when we look at we're doing for, you international and for medium enterprise, like it's kind of on all cylinders right now. And I think that that focus and you would get a, you know, significant market cap, large scale revenue machine, like workday, you know, moving and humming. we just have focus that our top competitors don't have on people and money. They're running, I would call it hyperscalers, but they're running some form of a scaled.

you know, compute cloud, they're selling a lot of other parts of the business. It's like, and I think focus matters a lot. And at end of the day, even at this, at this scale, it all comes down to execution and building great products. And I think that like, you know, there's no substitute for focus and actually, you know, building a quality product. And that's why I believe in my met with our customers and, and prospects, know, like I spend my fair bit of time and, know, uh, and deals myself, right. That focus shows.

Mark Vigoroso (38:44.743)
Yeah, that's right.

Mark Vigoroso (39:01.169)
Yeah. No, that's great. That's phenomenal. That's a great way to land the planes. So much. Thank you so much, Peter. I enjoyed it. I would love to park a time in 2026 and we'll come back around for chapter two on this one, but I really do.

I appreciate your time with us today and your willingness to share your views and perspectives and it is quite a ride. Congratulations on the success and there will be more I'm sure we will be hearing in 2026. But thank you so much, Peter. Thank you to everybody who's joined us this year on this road. We will be signing off for 2025.

Peter Bailis (39:23.984)
Okay. Okay.

Mark Vigoroso (39:40.506)
Peter is our anchor leg. And I think we did ourselves proud. So thank you again. And we wish everybody a happy holiday season, healthy holiday season. And we'll catch you in 2026 on the next episode of The Enterprise Edge. Thanks all.

​