How work gets done while you sleep | Aneel Bhusri & Joel Hellermark

Added: 2026-06-02

[00:00:08]Hello, folks.

[00:00:09]It's incredible to see you all here in New York.

[00:00:13]As Lauren mentioned, when I started Sana, I was most obsessed with the teams that we built.

[00:00:20]I wanted to build truly a generational team of some of the most incredible talents on the planet. And as I was doing this, I tried to study some of the greats.

[00:00:32]And what I discovered through this study is that most of the greats throughout history viewed the most important product that they would ever build was the company itself. The team, and how they organized.

[00:00:47]Whether it was Woz and Steve Jobs in the original Macintosh team, or the group at Xerox Parc in the late 70s, or the teachers and students at Black Mountain High, they operated in quite different ways.

[00:01:05]They were small, interdisciplinary teams of polymaths.

[00:01:09]They didn't really operate with the traditional command and control model.

[00:01:14]They introduced novel ways of operating, and what I believe is that these models will come back. This is the era of the return of the polymath.

[00:01:26]And as Ethan mentioned, the advancement in agentic AI enables us to fundamentally re-found our companies, re-found our products to serve this era.

[00:01:38]At Sana, we use this quite simple mental model to try to articulate what this will mean for our teams and our products.

[00:01:46]It's the five levels of autonomy.

[00:01:50]The first level is quite simple, really.

[00:01:52]It's the model that you already know—the pyramid.

[00:01:56]Tickets flowing between humans with a lot of clicks across dozens of enterprise systems. Then we move to the era of co-pilots.

[00:02:07]Here, AI models are drafting suggestions for us.

[00:02:11]We have a dozen pages open with different co-pilots and different interfaces that are assisting us in our work.

[00:02:19]At level three, this is where it really starts getting more interesting.

[00:02:23]Agents start actually doing the work and looping in the humans when necessary.

[00:02:29]At level four, we call this policy-bounded autonomy.

[00:02:34]We see the AI systems using our knowledge and our policies to automate work more end-to-end. And then we hit the apex.

[00:02:44]This is when AI systems become autonomous.

[00:02:49]Based on our rules and guardrails, they can self-improve these systems and run work end-to-end. Getting to this apex, we think, will be a long journey.

[00:03:00]And there are three core components that will enable this.

[00:03:05]The first one is a multiplayer AI experience.

[00:03:10]AI systems today are largely single-player.

[00:03:13]You sit there with your AI system, instructed along the way, but you don't really share context.

[00:03:20]You really don't loop in different departments or multiple agents.

[00:03:24]We think this has to change.

[00:03:26]You need a multiplayer AI experience that routes between teams of humans and teams of agents and brings them all together to collaborate.

[00:03:34]The second is living memory.

[00:03:38]Here, we need AI systems that recognize our company's policies, our company's knowledge, and can self-update this.

[00:03:46]The self-driving knowledge base and the self-driving policy graph.

[00:03:51]We move from systems that have amnesia and forget every single time we talk to them to systems with a company memory that self-updates.

[00:04:01]Then we have the final component, which brings together all of this into ambient agents. Today, all of our agents are sleeping until we start talking to them.

[00:04:12]I want agents that never sleep, that run 24/7 in the background, checking compliance, updating the rules, developing new workflows on my behalf.

[00:04:22]This is what ambient agents enable.

[00:04:26]Building on these primitives, today, we're incredibly excited to introduce two new products. The first one is Sana for ITSM. This enables Sana to run IT for you.

[00:04:43]Workday already has all of the rules and policies, covers the full employee journey.

[00:04:48]With Sana for IT, it can solve the tickets too.

[00:04:52]It can handle the full journey for an employee along every single step.

[00:04:57]From onboarding to offboarding to just-in-time requests, Sana solves that for you.

[00:05:03]The second product is the Travel Agent.

[00:05:07]The travel agent makes booking travel a truly delightful experience.

[00:05:15]A conversational experience which automatically files the expenses and updates the budgets. It's truly magical.

[00:05:26]I thought I might want to show you how this all comes together into a single experience for your employees.

[00:05:33]How agents can fundamentally reimagine the employee experience and provide a concierge for every single employee.

[00:05:40]Jenny, do you want to bring up the phone?

[00:05:51]So here you can see Sana.

[00:05:53]Sana brings all of our agents at Workday together into a unified application.

[00:05:58]So whether it's the Recruiting Agent, the Travel Agent, or Sana for ITSM, or any custom agents that you built, as an employee I get one unified interface to experience all of this.

[00:06:10]So I thought we might start with booking my trip home.

[00:06:14]Can you book a flight home for Sunday?

[00:06:20]Brilliant. Sana gets going.

[00:06:22]Here Sana is looking at my individual memory.

[00:06:24]It's looking at the company memory and the company policies, looking at my calendar and setting up this travel.

[00:06:30]But this is going to take a couple of seconds, so let's have Sana run this in the background.

[00:06:36]Then I can jump back to my task list.

[00:06:38]Here I can see all of the stuff that Sana is working on in the background.

[00:06:42]We have hiring five AEs for Singapore, Salesforce access to Kai, and of course the travel trip.

[00:06:49]So let's start by looking at some of the work that Sana is doing here in recruiting. I can see that Olivia is building out her team.

[00:06:56]She wants to add another five AEs to Singapore, and Sana is orchestrating this process.

[00:07:01]I can see the candidate pipeline, but Sana also leaves traces along the way.

[00:07:05]So I can see every single step what Sana has done.

[00:07:09]From creating five job requisitions, sending five requisitions for review, publishing the job posts, scheduling 38 first-round interviews, running the hiring loops, and then sending it to Olivia for approval.

[00:07:23]So that all looks good, but then Sana is going to elevate some of these for me for approval. So let's see here.

[00:07:30]Let's go through some of the human-in-the-loop steps and approve them.

[00:07:34]So the first one here is an offer letter for Aisha.

[00:07:37]What I can see is that it all looks within bounds, so I think this one is pretty good to approve.

[00:07:42]I'll send that out.

[00:07:45]But what's interesting here with the pattern that we're developing with the living memory is that Sana learns from every single interaction.

[00:07:52]Here it notices a pattern that I've repeated and asks whether it wants Sana to run that for me the next time.

[00:07:59]So here I can just say "yes, update the policy," and Sana writes that back into the company memory.

[00:08:06]So here's an example of in-band.

[00:08:09]But what Sana can also do is, of course, flag the out-of-band ones.

[00:08:13]So here we have Priya.

[00:08:14]Priya is actually getting a very good salary.

[00:08:18]I hope the Sanians in the team doesn't see this one, but I'm actually going to send this out as well.

[00:08:26]Congratulations, Priya.

[00:08:29]Better do a good job.

[00:08:33]But let's look at Sana running IT for me as well.

[00:08:37]So as mentioned, Workday runs the entire employee lifecycle from hire to retire.

[00:08:43]And along the way, you will have access to different systems.

[00:08:45]You'll need to resolve IT requests and so on.

[00:08:48]What you can see here is that Sana routes these tasks to me only when needed, auto-resolving the tasks that I don't need to do.

[00:08:55]So here we have Kai.

[00:08:56]Kai is our new sales lead for APAC. This all looks reasonable to me, so I'm going to approve this request as well and make sure that Kai gets access to Salesforce before she joins her role.

[00:09:09]But here Sana notices a pattern again.

[00:09:12]So it sees that the last couple of sales roles that we've had, I've always approved them.

[00:09:17]Do I want to give that autonomy to Sana the next time?

[00:09:20]I say "yes, always approve."

[00:09:22]Sana adds that back to the company memory and can then resolve that for me next time. Brilliant.

[00:09:31]So that is Sana for IT, running the IT workflows in the background.

[00:09:36]But if you all remember, I also had to get back to Stockholm at some point, so we started this with a trip back to Stockholm.

[00:09:42]Here we can see that Sana has provided a suggestion to me, checked my calendar so I get back right before the product review, which is good.

[00:09:52]I really want to stay on top of everything that we're building with this amount of exciting things going on, so I'll book that.

[00:09:59]But Sana doesn't stop there.

[00:10:01]Sana also goes through, splits the expenses, files the expenses, updates the calendar for me, and does all of that work in the background.

[00:10:18]Our hope is that you've filed your final expense at this point.

[00:10:25]And this is a fundamentally new pattern of how to use enterprise software, moving from co-pilots to thousands of agents running in the background.

[00:10:36]Here I can check in on all of the tasks that Sana has completed for me, from the Salesforce access to the Northwind renewal, rolling out the new EU expense policy, and a lot of other aspects.

[00:10:47]We see how proactive agents, ambient agents constantly running in the background, living memory, and the human plus agent collaboration is coming together into a truly magical experience.

[00:11:01]Thanks Jenny. And building this, I think we truly have one of the most talent-dense teams on the planet.

[00:11:15]We've gathered some of the best designers, researchers, engineers, and product leaders to reimagine how you use enterprise software.

[00:11:24]We have incredible product velocity and we're only just getting started.

[00:11:30]So as we make this transition, I think today, hopefully, we can officially announce the death of the co-pilots.

[00:11:38]We're moving from employees with dozens of open interfaces with different co-pilots to the agent-operator model.

[00:11:48]We're moving from 10 co-pilots per employee to thousands of agents that they operate. We're moving towards proactive intelligence, always-on intelligence.

[00:12:01]We're moving from systems and agents that constantly sleep to always awake agents.

[00:12:07]We're moving from agents with amnesia to agents with perfect memory about our organization and our individual preferences.

[00:12:15]This is truly a re-founding moment for us at Workday.

[00:12:20]But it's not the first one.

[00:12:21]I have the pleasure of working with one of the greats of our industry who's navigated this again, and again, and again.

[00:12:30]And I'm really excited to say that he's back.

[00:12:34]I want to welcome up my good friend and mentor, Aneel Bhusri.

[00:12:39]So what do you think?

[00:12:53]I'm glad there's not an agent for CEO so far.

[00:12:58]It's just really exciting.

[00:13:00]And I visited you in Stockholm last summer and a lot of your team, and I feel like at that moment I saw a vision of the future and bringing the two companies together has come into fruition.

[00:13:14]It's really pretty amazing.

[00:13:16]And one of the first gifts that you gave me when I joined was the SaaSpocalypse.

[00:13:26]Actually, I think there's some of the investors in the audience who might have given you that gift.

[00:13:33]I remember joining and I was quite interested.

[00:13:39]It was sort of a year after the DeepSeek moment where these simple narratives tend to get out of hand.

[00:13:48]But I also think there was some truth to that.

[00:13:51]So I'd love to have you unpack that, Aneel.

[00:13:53]Where do you think some of these investors are right and where are they wrong?

[00:13:56]So when I came back three months ago, I'll be honest, I was nervous.

[00:14:02]You know—"is there some truth to it?"

[00:14:04]And I've been through so many deep technical meetings and understanding how AI works. And I've been on the AI journey for a decade when it started out just being called "big data."

[00:14:15]And this last chapter has happened a lot faster than I thought it would.

[00:14:18]But I think we all expected this was going to happen at some point.

[00:14:22]And it just dawned on me.

[00:14:25]The two technologies are really incredibly complementary done the right way.

[00:14:29]AI is an incredible reasoning engine and effectively is trying to replace the human brain. It's like no other technology shift we've ever seen.

[00:14:36]And I'm old enough that I've been through quite a few of them.

[00:14:39]Usually at the starting level, I was one of the first couple hundred employees at PeopleSoft.

[00:14:44]I was the co-founder with Dave at Workday and always on the product and technology side. I've never seen anything like this before, and I think none of us have.

[00:14:53]At the same point, reasoning engines aren't really good at deterministic work.

[00:14:58]Running a payroll, closing the books.

[00:15:01]These are things that have a start to a process and an end to a process.

[00:15:05]They have to get done right.

[00:15:07]You can't do payroll right 90% of the time and not know which 90% of the people got paid and which ones didn't and then where the errors were.

[00:15:16]It has to be right 100% of the time, every time.

[00:15:18]That's the same way with closing the books.

[00:15:21]I think the real power is bringing the two technologies together.

[00:15:25]I do think all bets are off as it relates to what's happening in the industry, given what we can do together.

[00:15:34]Orchestrating workflows, static workflows, is now something that technologies like Sana and an agentic platform can do.

[00:15:43]Why shouldn't we orchestrate those workflows, especially the employee-centric ones, which are so many of those workflows?

[00:15:51]I do think technologies that are more around orchestration across applications, I think they're at risk in the SaaSpocalypse.

[00:15:59]Deep domain-specific applications, maybe they are down the road with the models, but we're using the models and we're using the tools.

[00:16:08]We're getting massive productivity gains with Claude Code and Cursor.

[00:16:13]I think this is a chance for us to be a disruptor, not a disruptee.

[00:16:18]I think what you're doing is letting us be a disruptor.

[00:16:21]I like being a disruptor.

[00:16:22]Some people say I'm too disruptive, but I like that.

[00:16:26]I think it's a chance for us to really spread our wings.

[00:16:30]Speaking of which, we just introduced these two new products today with Sana for ITSM as well as Travel.

[00:16:38]This is fundamental category expansion for us.

[00:16:43]Can you unpack what your thesis is with Sana for ITSM? Why are you so excited about it?

[00:16:50]Do I think we're going to be the leader in ITSM overnight?

[00:16:53]No, it's going to take time.

[00:16:56]People all think of Workday as a Fortune 500 solution.

[00:16:59]We have over 300 of the Fortune 500 companies running Workday, but we started out in medium enterprise and worked our way up.

[00:17:06]I think what you're going to see with ITSM, based on the great products you built, we'll start up there adding more and more ticketing capabilities, more and more workflows, and we'll keep moving up market.

[00:17:17]Much more functional, much more organic, and agentic, and I think it's going to be at quite a lower cost.

[00:17:25]That's the way you start every new product.

[00:17:26]You start in the lower end of the market and you work your way up as you add capabilities. I think that's going to be the case with Travel as well.

[00:17:34]When I think about Workday, I think about great technology companies.

[00:17:40]I look at the Apples, and Microsofts, and Googles of the world that have transcended technology generations.

[00:17:46]They were successful in all generations.

[00:17:48]For Workday to be considered a great company, a great technology company that has great values and takes care of its customers, we can't just be a winner in cloud.

[00:17:56]We've got to do a great job in AI, and that's kind of the reason I came back.

[00:18:00]It was, honestly, to work with you.

[00:18:03]I'm wearing this in your honor— for people that aren't tennis fans, this was one of the original Björn Borg jackets from Sweden.

[00:18:11]It is.

[00:18:12]Very subtle reference, right?

[00:18:16]And I'm a huge believer in what you're doing, and excited about your new role, and I love working with Gerrit.

[00:18:24]I mean, you know, the Europeans are kind of taking over Workday.

[00:18:30]Yeah, so we haven't officially announced the new role yet.

[00:18:34]Maybe we should.

[00:18:35]Well, we should.

[00:18:37]So, yeah, we're doing earnings in about an hour, and that's where it gets official for the world, but you'll be the first to know, Joel is being named our new Chief AI Officer for the company.

[00:18:53]And I will honestly say one of the reasons I came back was to work with you, and to work with Gerrit.

[00:19:01]It really is a re-founding moment.

[00:19:04]If it was the 30th version of cloud, there are better people to run the company.

[00:19:08]That's not my gift.

[00:19:09]I try to be an innovator and a pioneer.

[00:19:12]This time around, you know, I'm 60.

[00:19:15]I'm not the 39-year-old who started Workday with Dave.

[00:19:17]I feel like I'm more that agentic orchestration agent just trying to orchestrate all the work with all the experts like you guys that are really reinventing the world of work, and I think it's really a reinvention.

[00:19:31]It's not just adding AI. It's really reinventing it, and that's what you showed.

[00:19:35]So now that you're the big boss on AI, what do you think you're going to do?

[00:19:42]Where are we going?

[00:19:44]One of the first things Aneel was very nice to say was, "Don't screw it up."

[00:19:50]It feels like a pretty important thing to fix at this moment, so I'm sure I'll be very, very busy.

[00:19:57]I mean, I came into this role, and it was also a big reason why we partnered with Workday and became part of Workday was, this wasn't about playing safe.

[00:20:08]This is a fundamentally new era for software.

[00:20:12]We're not really selling software anymore.

[00:20:14]We're selling work itself.

[00:20:16]And this is a massive TAM expansion.

[00:20:19]If you just think about the market for software, significantly smaller than the market for labor that runs on top of this software.

[00:20:26]And building agents that can start handling more of that work, I think is incredibly inspiring.

[00:20:32]What we spoke about here previously, we're sort of moving through these levels of autonomy.

[00:20:37]And I think it will be a bit analogous to self-driving cars.

[00:20:40]I think people overstated how quickly that would be.

[00:20:43]You have to run through the edge cases for quite a long time.

[00:20:47]And during that era, you need a lot of humans in the loop.

[00:20:50]That's why you had a lot of humans sit alongside the self-driving car systems.

[00:20:55]And if you think about what's the best platform to derive the policies, the knowledge that these systems run on, that is truly Workday.

[00:21:04]So building the policy graph, building what Gerrit likes to say as the sort of "world model of work" that I think is incredibly, incredibly powerful.

[00:21:12]Building the knowledge graph that these agents run on, I think that's incredibly inspiring.

[00:21:17]So that's what we'll be focused on.

[00:21:18]Well, it's exciting. We are, as you and I have talked multiple times, we're trying to return to our startup roots.

[00:21:25]So having someone that is a recent startup founder driving, I think is absolutely the right thing for the company.

[00:21:32]We have clear lines of ownership now on AI vision, on our applications, and our APIs.

[00:21:41]And I'm really excited about where we're headed.

[00:21:43]So now that Sana is part of Workday, what new things do you think you can do?

[00:21:47]You show those applications, but you now can touch 75, 80 million employees.

[00:21:53]What new things do you think about?

[00:21:55]I think with these primitives that we're building out, it's interesting to see the sort of boundaries of software blur.

[00:22:05]What really is ITSM—what are the boundaries of these different systems?

[00:22:10]If you look at the core of it, it's that knowledge graph, it's that connectivity, it's those agents, it's that multiplayer AI experience of humans and agents collaborating.

[00:22:20]And so as we build those out, I think we're going into a lot of those adjacent domains such as ITSM. But when you think about a domain like ITSM, the sort of core component of that is the employee lifecycle.

[00:22:35]It's when the employee gets onboarded, when they get offboarded and the journey along the way. So it makes a lot of sense for a product like Workday to own those workflows. And we're seeing this in a lot of other domains too.

[00:22:49]So it feels like we truly have this sort of gold that we can build with.

[00:22:56]We can expand into new categories, we can build the policy-bounded agents, we can increase the level of autonomy of these agents over time.

[00:23:06]And we do that with some of the deepest relationships with the most important companies on the planet.

[00:23:12]So as someone who really loves building products, we can bring this to scale at a whole other level. One of the things that I'm also very interested in, coming from our European and Scandinavian roots, is that over the past couple of years, people were complaining quite a bit about the user experience of Workday.

[00:23:34]And so my internal goal is to make Workday the Hermès of enterprise software.

[00:23:47]I will own the UI comment, but it also looks fabulous now with Sana as the kind of the gateway to every employee's use of Workday.

[00:24:00]Yeah, and the user experience is shifting, right?

[00:24:04]We've spoken about that of AI being the new UI. As you could see in the demo today, I wasn't really clicking through dozens of different enterprise applications.

[00:24:13]I had it all orchestrated in a single interface for me, where I could oversee thousands of agents doing work, escalating to me when necessary as well.

[00:24:23]So I think this is a very good moment in time to re-imagine the user experience.

[00:24:28]Yep. Well, I couldn't be more excited about what Sana, your team, what you're gonna drive going forward.

[00:24:38]You know, this is a story of Workday and Sana coming together when one plus one is way more than three or four or five.

[00:24:45]I mean, I think we're redefining how business and work gets done.

[00:24:49]And you know, you're gonna lead us into that future.

[00:24:52]As Dave said, I remember when Dave handed over the reins to me in 2015, we shook hands, we hugged on stage, he walked off and turned around and said, "Don't screw it up."

[00:25:03]I'm gonna say the same to you.

[00:25:05]Maybe you should walk off right now.

[00:25:08]You won't, and you won't.

[00:25:10]Thank you very much, Aneel.

[00:25:12]And thank you, team.

[00:25:13]Thank you.