Atlan CEO Prukalpa Sankar

Auren Hoffman (00:00.91) Hello fellow data nerds, my guest today is Prukalpa Sankar. Prukalpa is the co-founder and CEO of Atlan, a data ops platform used by teams at NASDAQ, Elastic, and the UN to manage data collaboration at scale. Atlan is raised over $200 million and I'm a small investor, very happy investor, and growing revenue 7x in the last two years. Prukalpa, welcome to World of DaaS.

Prukalpa (00:22.199) Yeah.

Prukalpa (00:29.327) I'm excited to be here and thank you for being a very early supporter.

Auren Hoffman (00:31.31) super excited to be here as well. I just want to question I have is like you've got this transformation in your company, so many other companies are transforming their company where they're making sure that all their employees are becoming AI native using AI tools. What are you doing to get everyone to use those tools?

Prukalpa (00:55.619) Yeah. So we started A Transformation. We put out a mission to the company end of last year, early this year, saying we want to be an AI native company. And I think it's important to first just define what does it mean to be AI native. And the way we defined it was if we were building the company from the ground up today in a world where AI exists, how would we reimagine?

Everything.

Auren Hoffman (01:26.22) Yep. And how would a new company basically kill you, essentially?

Prukalpa (01:29.595) Exactly. I think the, because I think we're still most, where most companies are, is you are trying to augment an existing human process with AI, which I think is being AI curious, you're not actually being AI native. And I think the transformation to being AI native was really important to us. Like we've been like, we've been adopting AI tools for like a while. But

I think the model, somewhere in the last year, I think the models got good enough to say, hey, actually, we can truly reimagine this process and this business process to be AI native. And that kickstarted this transformation journey. It's frankly even harder than I expected it to be. And our takeaway was it needs to, like the transformation needs to happen at multiple levels. So the way we did this was we first set up

an AI task force, which is this whole separate team whose role it was to ship business first use cases. So it operated as its own team. said here, like first we needed to show people that this was real and it could work. So show people within the company that it could work. And so that was the first couple of things that we did. So we shipped out like a couple of real life business use cases with AI.

Auren Hoffman (02:42.13) Show people within the company.

Prukalpa (02:55.279) Then we started.

Auren Hoffman (02:56.436) Sorry, do you mean use cases meaning like like product enhancements or something or.

Prukalpa (03:01.987) And now product and asset, we've been shipping AI stuff into our product for a long time. For us, being AI native was, how do we as a company become AI native? this was, exactly, was customer support. How do we reimagine an existing internal workflow and showcase that this stuff really works? Once we did that, we launched a mission to be AI native.

Auren Hoffman (03:06.68) Yeah.

Auren Hoffman (03:11.32) So it's like an internal process, like recruiting or something. Okay.

Prukalpa (03:31.163) And with that, we launched an AI productivity challenge across the company. So it became really bottom up. That worked beautifully because the energy across the company, people helping each other out, we have this black channel with like 200 people on it and they're constantly submitting stuff as to what...

what they're doing into their day-to-day workflows. We made it like fun. were people like, there were a people who want iPads. Like there was just stuff like that that we did that I think created the bottom up momentum across the entire company. So for example, today, I think I would be very surprised if there's anyone at AppLint who does not work with AI on a daily basis. And now we put on the third level of our execution, which we're calling AI native departments, which is

we're changing our objects. So in our customer experience team, we have basically these agents and like, it's just like multi-agent workflow. All these agents have names. All these agents have like, and so we're basically.

Auren Hoffman (04:34.713) So the agent has a name like, like, like Susan type of thing, like.

Prukalpa (04:38.829) Yeah, like Susan and Lane and like we might make it fun. Like, so we might even name some agents after like people on the team, as, and so that's kind of the third workflow that now we're reimagining the departments with like, essentially a new org chart. So, and what would that org chart look like? And what does the human agent collaboration workflow look like? Which is when I think we truly build an AI native department. That's kind of where we are today.

Auren Hoffman (05:07.302) What are some of the things that you're doing today that you weren't doing like six months ago or a year ago that is improving things?

Prukalpa (05:19.515) A large part of it is the ability to process like what does AI do really well? AI does two things really well. AI has the ability to process vast amounts of information. AI has the ability to generate things in a way like that's the core element of generative AI. So one element that we have gone really deep into is how do we now, we listen to every call like

Before we needed to look at conversion rates and then against conversion rates, we needed to go and say, okay, here's the change that we need to, here's what we need to investigate because this is dropping versus now actually we can do a completely different workflow. We can listen to every call or AI can listen to every call. And so it's translated into things like we have live, we listen to every single call, we score every single call. We take those into signals that we're driving better product, better like.

Auren Hoffman (05:59.067) Yeah.

Prukalpa (06:17.775) customer experience, here's the gaps that this customer might be facing in that implementation. We send that back as coaching to individual team members. then we're able to, so A, think just the ability to process signals has changed dramatically. The other has been a lot of generative use cases, right? So we can try, we can write proposals better than just about anyone. can, we can scale like reporting in a way that like.

One of our best use cases has been, yeah.

Auren Hoffman (06:47.446) When you want to write like an RFP or something like that, are you guys writing that writer or is there just like a random software company that has a good RFP writer that you use or something?

Prukalpa (07:02.779) I think we're using a, we are using a tool for RFPs. And so it's been a mix of a build versus buy strategy, at least internally for us. So I think for RFPs we're using like an external tool. For some of the enrichment stuff, we're using tools like clay. And then there's for a lot of the...

internal agentic staffs who are building things internally. That's the flow.

Auren Hoffman (07:34.546) Because just that idea of, okay, got to go, before I go do it myself, maybe I should go search for the vendor, go look at it, go look at a bunch of different things, evaluate some things, look at the different tools. That's almost like a, different type of mindset, different type of employee. How does one push that down?

Prukalpa (07:57.625) Yeah.

Prukalpa (08:01.819) It's hard. And at least our strategy actually has been a little bit more geared towards build than buy. And a part of this is because we are, I think we're also fortunate to operate in the data and AI space. And so I think the right to win for very small use cases of a business application kind of use cases is low today.

And we care a lot about living in the future for our customers because it helps us build the best product for our customers. A lot of customers use Athlon for their agentic use cases and the few like to power their agentic use cases. And we want to be ahead of that. And so the pattern, we buy a few things at the infrastructure layer. We leverage a bunch of the cloud stuff at the infrastructure layer. And then we build agents on top of that. And it's not that complicated to build an agent. Like the real?

The part of all of this is the business context. Like that's the hardest part. The operating model and the business context is the hardest part of making AI successful in a company.

Auren Hoffman (09:09.13) What do you mean? What do you mean by that?

Prukalpa (09:13.573) The difference between a co-pilot and an agent is that the agent understands your context perfectly. So your agent has to be better than any single human being and ideally better than all the human beings around that topic. That's the power of the agent. That means iteration and flows.

Auren Hoffman (09:27.98) Yeah.

Prukalpa (09:40.559) Because for example, like if you think about just a prompt, like a simple first level prompt is not enough. The data context that needs to go behind it is problematic. The workflow context, business, like that's really hard. And so what typically happens is that people ship something. It has 70 % accuracy. So you can use it as a copilot, but it's not at a hundred percent. It's a lot of work to go from 70 % to a hundred percent and people give up in the process. this is like, I mean.

Auren Hoffman (10:03.641) Yeah.

Auren Hoffman (10:08.365) Yeah.

Prukalpa (10:10.395) Right now, only 25 % of AI experiments make it to production. 46 % of AI experiments are actually abandoned. And that chasm is hard to fill. And that chasm, unfortunately, can't be filled purely through technical talent. It needs this iterative operating model between the business

someone who does the job, who knows the job really well, and like technical team, and that operating model, I think, is the hardest part. Like building the operating model and building the context is really hard.

Auren Hoffman (10:49.227) Interesting. Now what like when you're hiring people, have you kind of changed a little bit how you think about hiring the

maybe the types of traits you're hiring for in the Sanon model.

Prukalpa (11:04.579) Yeah. yeah, we have like one of our pillars is AI native talent density. And we've made being AI native like a core part of hiring in

Auren Hoffman (11:17.096) So you have to already be AI native to get hired at

Prukalpa (11:22.467) Yeah, you should already be an AI native to be. I mean, the question, you should definitely be able to become AI native. And so we've made a few changes. One, some of the most basic changes we've made is, for example, we have something called a real life problem statement. We always had this. It's a very core part of how we hired at Atlan people work. Like we believe in seeing people's work rather than just them talking about their work.

Auren Hoffman (11:30.317) Yeah.

Prukalpa (11:46.847) So one thing we did is like we wrote all those prompts. We wrote all of those challenges. And now we actually explicitly say that we are an AI native company. We expect you to work with AI to solve this. And we get you to show us. like when we're like, okay, show us how you're writing the prompt. Like show us how you're like working with AI to solve this problem. And that was helpful.

Auren Hoffman (11:57.347) Yeah.

Auren Hoffman (12:05.73) Yeah. So is that live or do they record it or send it to you or is it an async thing or how do you give this? It's like a homework thing to...

Prukalpa (12:16.441) The prompt that we give is homework thing. People go work on it async, and then they come back and they present with us. And it's like a real life problem solving exercise together. So the intent is we work through the project. Our entire philosophy around the challenge statement has always been it's a way for you to get to know Athlen really well. So we open up everything to you, like you were like a real life Athlenian in the company.

Auren Hoffman (12:26.51) Yeah.

Prukalpa (12:43.653) So we give you access to data, we give you access to tools, we give you access to people, and you work through the problem together. Now we've just added AI as another Kokora collaborator that we expect you to be natively using as a part of your thought process. That has been phenomenal because the interesting thing about AI is that certain skill sets become very, like much more relevant and certain skill sets go away.

So for example, clean thinking is a very, like just core ability to describe a problem in a structured, clean way is a much more important skill set. Any past context, experience, like any of these things, like they just actually become, like they diminish so much. And so what

Auren Hoffman (13:27.553) Yes.

Prukalpa (13:41.657) It's been really helpful to actually see people like that.

Auren Hoffman (13:43.719) I mean, experience could potentially be correlated with being able to describe the problem or something, right?

Prukalpa (13:50.949) Yes, maybe exactly. like it's not, which is why I meant like pre context is not that helpful anymore. Meaning this whole, I have done this before. Used to be a very important part of like a hiring process. And I have done this before actually knowledge is not that important. Like because you can learn knowledge very quickly. The question is like, can you ask the right questions and can you iterate? So seeing people like great with AI has been very helpful.

Auren Hoffman (13:58.968) Yep.

Auren Hoffman (14:03.991) Yes.

Auren Hoffman (14:09.379) Yes.

Prukalpa (14:20.315) We do a lot of questions around like, tell us, show us what you do. Don't like talk to us about it. Cause honestly, like if anyone's interviewing in this market and they say, don't use AI, that's just like stupid. I mean, I don't think anyone's going to do that. Yeah. Yeah. The other thing has been actually, are the, like, we've been thinking a lot about what are the traits, like the cultural traits that make someone successful in this new world. And it's surprising. I actually think it's similar to what

Auren Hoffman (14:31.158) Yeah, they're all going to say it. Yeah. Yeah.

Prukalpa (14:50.777) It used to be in the old world. there's this book called The Growth Mindset. And I actually think a lot of the core, like being able to approach a problem with beginner's mindset, being able to have, it's interesting, but this, like, I believe that extremely secure people do better with AI. So for example, like, if you have any element of insecurity in you, where it leads to a natural

AI will not be able to do this better than me. Versus if you are completely secure in yourself, there's an element of, of course I'm going to work with AI too. So there's all these very core personality traits.

Auren Hoffman (15:30.93) right.

Yeah. Sorry. I most people have some insecurity here and there, right? They've got some sort of thing. So there's a spectrum, I assume you're talking about. Yeah.

Prukalpa (15:42.203) Yeah, there's a spectrum. Obviously, there's a spectrum. Like this is not like, you know, of course, everyone has some insecurity. The question is like, the question is, do you take a bigger, like, do you take a growth mindset to the problem? And do you say, Hey, you know what, like, this is this amazing technology, and I'm going to learn how to do do better with it. versus

Auren Hoffman (16:01.858) Yeah. And you also have to say, can't be worried it's going to take your job. Like, cause if you are worried, it probably will. It's almost like you have to be using it as a tool so that you can get better.

Prukalpa (16:14.991) Yeah, yeah, yeah. I mean, I mean, there's this tweet that's going that's making the rounds. I'd like, yeah, it's unlikely to take your job, but somebody else who uses AI is definitely going to take your job right like, yeah, yeah, it's actually become a retention and talent attraction pool for us at Atlin. Because people who are at Atlin, we are a company that's investing in being AI native. We are helping our people.

Auren Hoffman (16:24.128) Yes, yeah, that's almost certainly true.

Auren Hoffman (16:40.14) Yeah. To your learn.

Prukalpa (16:42.851) Yeah, we're helping our people be the most forward thinking people, hopefully on the planet. And, you know, I truly believe we're in the top 1 % of companies doing that. that, you know, it means that talent that's working elsewhere, but their companies are not giving them the ability to do that. They're coming to have talent that like wants to stay. In fact, after our data update agent, because now like humans don't need to like update data the same way anymore, because like AI does it for them.

People were like, oh my God, I'm not going to work in a company where I have to do work manually again. So we believe that, at least for the next, we don't know what the long-term future is. But I think there'll be companies that don't use AI and die. I think there'll be companies that use AI and become operationally more efficient. But I think there'll be companies that use AI and raise their ambition 10x.

Auren Hoffman (17:16.674) Yeah, totally.

Prukalpa (17:39.483) because that's always the challenge of startups. Like you never have enough resources to do the things that you want to do. AI makes that easier. We want to be the third kind of company. We're going to use this to like become 10x more ambitious. And so we're not seeing this as a future where we're going to like not hire human beings. We just want our human beings to be 10x better and more productive than a human being anywhere else. And we think that's a competitive mode.

Auren Hoffman (18:02.281) But do you see it slowing the rate of hiring? Because hiring has a huge cost because of the communication. Every time you hire someone,

It becomes harder in a company communicate. You become more bureaucratic. Things move slower. It's not even the best run companies, right?

Prukalpa (18:13.893) Yes. Yes. Yes.

Yes. Yes. Yes. So I think we end up not hiring for... So it's kind of that thing, right? If you can make everyone 10x more productive, ideally that human being can do 10x more, which means you can grow... If you just think about your growth rate versus how many people you need to do it, that shrinks. And then you go hire people...

to go do the new thing because you have earned the right to go do the new thing and the new thing. like, I think that's the way we're thinking about it. But absolutely, I definitely think it will slow the pace of hiring. It reduces viscosity, can reduce so much viscosity inside companies, which and we're excited about all of those things.

Auren Hoffman (19:05.587) Now you're, you know, part of what you're doing is just like, is your, you guys are in the data world. You're kind of using data, manipulating data. So many companies are, are not using data in the right way. What, what are the, what are the big kind of big kind of mistake blocks you see companies making and how do you see that moving?

over time.

Prukalpa (19:35.183) on

I think I've got an interesting world. think the best way I'll describe it was a customer, a large bank, who told me this recently. said, our company has gone through three waves of data platform, re-platforming over the last 20 years. We've gone from the Hadoop era to the cloud data warehouse era. Now we're in a cloud data lake era. We've gone through all these eras, but nothing has changed for my end users.

Auren Hoffman (20:01.506) Yep.

Prukalpa (20:07.321) My users still can't find the data, they still can't trust it, they still don't use it. Nothing I said.

Auren Hoffman (20:10.232) Yep. Yep.

Prukalpa (20:16.827) I'm actually much more optimistic about this than I ever was in the past. I think we have solved a lot of the core foundational architectural problems that data was running into over the last couple of decades. I think we're now at the last mile problem of data. That last mile problem is a heart problem because that last mile problem is not just a technology problem. That last mile problem is a human plus technology problem.

And that last night problem is even more valuable.

Auren Hoffman (20:48.729) And there's a UI issue there too.

Prukalpa (20:53.987) I think there's a UI issue there too. There's a diversity issue. I think the thing about data that most people don't realize, I call it the two fundamental truths of data. The first is that data is the most diverse ecosystem, team, landscape that can exist. And the reason for that is because if you have to make a data project successful, now a data and AI project successful.

But this used to be called like, mean, was, there used to be data science projects in the old world or whatever. Like you need multiple different skill sets to come together. You need analyst, engineer. Now, like, you know, now you need AI engineers. You have, you know, business. Business is this one word that we use, but actually it means lots of different things. like finance business is very different than marketing business is very different than sales business.

Auren Hoffman (21:44.206) So.

Prukalpa (21:45.467) They're like, if you break this down, you're talking about 12 to 15 different personas in a single project. And the thing about these people is that they're fundamentally different. An engineer is a fundamentally different person than an analyst. They have completely different tooling preferences. They have completely different skill sets. They have completely different DNA. Which unlike other teams, like if you think about a sales team, it's like a relatively homogeneous team.

A data organization is fundamentally heterogeneous. This leads to in some ways the tooling proliferation. That's why you'll see 87 % of companies use two or more BI tools. You know, like the number of people who use like number of.

Auren Hoffman (22:23.666) Yeah. I mean, I'm surprised it's not, it's not like, you know, 100 % of companies use five or more tools or something.

Prukalpa (22:31.291) Yeah. It's insane, right? Like the number of like, you know, you would think that these problems are, you know, good turn homogeneous, the number of the number of customers we have that use snowflake and database both is like insane. And the reason for that, I think, is because of this fundamental heterogeneity in the ecosystem, which leads to different tool preferences, which leads to these complex heterogeneous data stacks. And the second problem

Auren Hoffman (22:46.019) Yeah.

Prukalpa (23:01.115) actually in this ecosystem more than anything else has changed. It changes faster than ever before. Two years ago, vector databases were the rage. Three years ago, no one talked about it. Today, you can argue we have stopped talking about it. It's not a thing anymore. And that chain that is the only set in reality, it's just very fast evolving ecosystem. And that core problem...

Auren Hoffman (23:14.811) Yeah.

Prukalpa (23:28.387) is what leads to in some ways. So when you say something as simple as, my business doesn't trust this number, it sounds like a really simple question, but it's a very hard question to answer because you're talking about connecting a data landscape of, you know, hundred different tools. You're talking about bringing together context. Like the reason the number could be off is because finance changed the calculation of the revenue metric. Could also be wrong because the data engineer changed the pipeline. Could also be wrong because the quality check failed.

There could be so many different people involved in this problem. And that I think leads to this complexity of the last mile not being solved.

Auren Hoffman (23:58.478) What, what, what, so we just saw Salesforce bought Informatica for $8 billion, ServiceNow bought data.world, like walk us through like how the market is thinking about all these different kinds of metadata systems.

Prukalpa (24:18.937) Yeah. So I think that brings us to today where this problem was always a big problem. But now it has kind of like, it's almost become 10x more important because of AI. Because AI can't do anything without context. Like context is everywhere, right? Like it's the model context protocol, is, you know, context windows. Like everything about AI is can you teach it context? In fact, my co-founder, Varun, had this

know, very profound statement he made recently, which I think we'll write a blog about where he was like in 1996, Bill Gates wrote like internet data which was happening. Bill Gates wrote this blog post and he said content is king in this era. And I think in the 20, in 2025, you can actually say in this new era, like context is king. And that's what people are realizing, right? So if you think about any of these application layer companies, the application layer company, ServiceNow and so on.

Auren Hoffman (25:05.609) Yeah.

Prukalpa (25:17.753) They have always held business data and business context has stayed in these tools. You could argue that becoming a little less like, you worldwide humans don't create business context. These tools become more irrelevant. Like they've all been built on the concept of humans will create business context. Yeah, they're right. Exactly. So now they have to win the agent again. So they have to become the people that replaces the CRM with the agents.

Auren Hoffman (25:31.948) Yep. Yeah, humans come in and they type in and they kind of like add stuff. Yeah.

Prukalpa (25:47.803) which is what Salesforce is trying to do with agent force and so on. They've all announced these great AI dreams over the last 12 to 18 months and they've realized as they take these things to production that there's no way they can do it without context. And the context layer is in the metadata layer. And so one of the things that many of these companies are trying to do is that they're trying to buy these capabilities. They say, hey, I need to create the metadata layer inside of my business applications.

so that I can go win the agenting game. And so that's one kind of like change that's happening in the business application where a lot of the business data lives. On the other hand, you're seeing almost this creation of the data and AI clouds, which is kind of where like Snowflake and Databricks traditionally played. And these companies actually are in an interesting space where

They have been built on the premise that you will move your data from the business application layer to this data layer or the central data layer. And so now the business application companies make a play to say, no, you're not going to move your data out of us. Like you're going to keep your agents on us. These companies are now trying to find capabilities to say, hey, know what, like we need to be the place that we will

build and maintain all your agents of the future. And so what capabilities do I need to make that happen? I need the capabilities to be able to look at context across your ecosystem. want to be.

Auren Hoffman (27:17.63) Yeah. And so you think these business application companies are going to be more likely be the losers or what houses net out?

Prukalpa (27:35.511) You know, time will tell. My sense is that we will have a few key changes that will happen. I think the business applications will open up their data through open table formats. So you're seeing this already happen. SAP recently launched something called the SAP Business Cloud, where they're opening up their data through an iceberg native function and open table formats. What that means is...

Auren Hoffman (27:57.99) Yeah.

Prukalpa (28:03.897) that you don't need to move data anymore. You can keep your data in SAP and you can bring compute to SAP. And so I think that will minimize the amount of need or storage movement where we just keep migrating data from home and, it's insane. Like the last 20 years, the ecosystem just migrated data. Like the amount of money that's been spent on migrations is insane. Like we just keep migrating stuff. And so think migrations go away. I think data becomes open.

I think compute layers will stay. So, fundamentally, so Iceberg is a new, it's been catching a lot of attention in the ecosystem. A lot of customers, like about 600 of Snowflake's customers today are using Iceberg. So, that's like 10, 12 % of large enterprise customers. What you're seeing is it's an open, it's fundamentally an open table format, which says that I...

Auren Hoffman (28:33.333) What do you by open?

Prukalpa (29:02.403) In the old world, you used to take data to compute. So you used to move data to a compute computation layer like a data warehouse or a data lake house. With an open table format, you can bring compute to your data. And you can make any compute to your data.

Auren Hoffman (29:09.815) Yeah.

Auren Hoffman (29:18.062) Why is that better? Is it just more efficient because you're not transferring terabytes of data back and forth or?

Prukalpa (29:26.605) It's definitely more efficient first because you know, actually 40 % of someone was telling me this tag recently, 40 % of cost of compute is actually insert operations, which is literally like migrating, like migrating data from one place to another. So A like adjust far more efficient. Second, it keeps it open because now there's all the interesting innovations and compute layers that are happening. So I have customers telling me, Hey, for this operational workload, I'm using this new

performance based engine in the ecosystem. And so what you can do is you have an open ecosystem. Like you can have customers that use Snowflake for their analytics workload. You can have customers that use Databricks for something else. You can use the new startup for the third thing and the fourth, right? And so that.

Auren Hoffman (30:00.206) Thank

Got it. So basically you've got like a data store, data store for whatever you're sitting, just sitting there. It doesn't, and then you can just bring the tool to the data rather than bringing the data each time. Every time before you'd had a load into every tool and it took a lot of load and all this other thing. And okay. And then of course now you, and then, and then the data would

Prukalpa (30:27.477) Exactly. Exactly. Exactly. Exactly.

Auren Hoffman (30:34.868) You'd have a problem before syncing the data because it it would go out of date or whatever. And so now you can always have the most synced data, the most up-to-date data in one place.

Prukalpa (30:44.571) Exactly, exactly. And so what I think that creates is you have a bunch of your, wherever your source data is getting created, that data will start becoming open through OpenTable formats. We will start bringing compute to that. And there will obviously be more compute with the AI agent.

Auren Hoffman (31:05.624) And do you think like, like applications that are having, they're creating proprietary data within the application, let's say, let's say, Salesforce or something, they are gonna, are, are, are they going to move to this kind of more open data thing? Or do you think they'll try to be, they'll, they'll try to put roadblocks to make it harder to move that data.

Prukalpa (31:27.801) I think it depends on the strategy of companies. I think that there are two kinds of companies that will get created. In fact, like last week, there was all this noise in the ecosystem about Salesforce shutting off the Slack data API access. On the other hand, like I think there are companies like SAP that were traditionally slightly more closed ecosystems actually that are saying, actually, we're going to open up our data through a business cloud.

My sense is that what customers will want is open and interoperable. Because in the world where innovation is rapid and in a world where we just have to accept that that's the world that we're going to live in in the next 10 to 15 years, there's going to be a lot of change. Customers will want to innovate rapidly.

Auren Hoffman (32:11.265) Yeah.

Prukalpa (32:18.031) And so customers will want to have the choice. So my sense is if you believe that ecosystems will be built based on customer choice, my sense is the world, like going after open table or open interoperable formats is the right thing for the customer. If you believe that the right thing for the customer is what the industry will go to, I believe that that is what will happen.

Auren Hoffman (32:35.168) Yeah. I mean, today, like after I saw that, that Salesforce announcement, like today, I wouldn't, I would never implement Salesforce if I could go, if I could avoid it. Now, if you're already, if you're already part of that ecosystem, it's not that easy to get off of it.

And so I think each of these companies are going to have some sort of complicated matrix as they're making a decision there. But it definitely makes me more reluctant. And even if I was a company like on the margin, should I invest more in Salesforce or should I invest more in these other tools? I'm probably going to invest more in these other tools today after that announcement. But clearly Salesforce is smart. Like they're making this bet for a reason. They're making a closed bet for a reason. what, how do you steal man there?

strategy.

Prukalpa (33:28.443) I mean, it's a good moot. If Salesforce wants its agents, if you, if if Salesforce believes that agents is its future and today the only thing that it has, and if you can, if you think about the world where the CRM goes away and agency plays the CRM, then how do you retain the right over the next two to three years to become the company that owns the agents? Like what's your moot? And if your moot is not shipping fast, shipping the best product, the fastest then.

Auren Hoffman (33:32.738) Yeah, for sure. Yeah.

Auren Hoffman (33:57.09) Yep, just certainly not them, yeah.

Prukalpa (33:59.767) I think it's a good note and I think companies that have business context today definitely have an advantage because that's the core of what helps make agents work. And I will tell I think on which strategy will play out.

At least what I tell customers is I tell customers, you should build your stack for a few layers to be open, a few layers to be interoperable no matter what. And a few layers you should expect to use multiple tools. So the way I think about it is like you should expect to be multi-cloud platform. You should expect to be multiple business applications. You should expect to be multiple data platforms.

Auren Hoffman (34:48.336) Yeah.

Prukalpa (34:56.791) You should have a centralized context layer. You should have a centralized governance layer. Both of these should also be open and interoperable so that when you are experimenting with multiple models and foundational models, you're experimenting with multiple clouds. You're experimenting with all of these. You have some layers that you keep centralized so that your context doesn't get logged into a single application. Your governance doesn't get logged into a single place.

Auren Hoffman (35:14.486) Yeah.

Prukalpa (35:25.733) But it has to be open and interoperable for you as a customer because Mr. Customer.

Auren Hoffman (35:28.494) Got it. So you get your context in one place. In some ways, you almost have your data in one place. But then you could be very promiscuous about using GCP or AWS. You could be very promiscuous about using Anthropic or OpenAI. You try all of them out all the time, constantly just trying to learn about what's the best for you. And sometimes it's a cost thing, and sometimes it's a performance thing. So depending, you may have different things you would be optimizing for.

Prukalpa (35:43.065) Yeah, yeah, yes, yes, yes, yes, absolutely. honestly,

Exactly. Exactly. And sometimes it's a use case thing. Like there's some use cases that Anthropic is better for. There's some use cases that OpenAI is better for.

Auren Hoffman (36:04.174) Interesting. Now you are running a, even though you're a startup, it's complex because you've got people all over the world through, you know, every time, you know, all the different time zones and stuff like that. How are you, how do you think about managing it? Just somehow you have to like async it out or something because it's going to, it's hard to have synchronous collaboration when you have people that are, you know, 12 hour differences and stuff.

Prukalpa (36:29.273) Yeah. Yeah, so we decided to be a completely global remote first company. That means that we have team members in over 20 countries. we are global and remote. And our strategy on this, mostly working from home, we have a global WeWork membership. So, you know, our colleagues can go into like their nearest WeWork.

Auren Hoffman (36:46.87) mostly working from home type of thing.

Auren Hoffman (36:53.506) Yeah. Yeah.

Prukalpa (36:58.469) for sync things that they want to do, but mostly working from home. Our high level strategy on this has been to say, what do you optimize in-person time for and what do you optimize remote time for? So the way we think about it is, it's remote for productivity and execution. It's in-person for problem solving and human connection. And so we have very, we've tried to very.

Auren Hoffman (37:24.934) And what about like, what about non okay, but you also have like random meetings that happen throughout the week or something. Are you just trying to like reduce those as much as possible? So you don't have as many video meetings and stuff or.

Prukalpa (37:39.653) So we have something called Atlan Standard Time, which we've created, which is 7am to 11am Pacific, which is the time that we expect every Athleanian to be online for collaboration hours.

Auren Hoffman (37:53.578) So 7 a.m. to 11 a.m. Pacific, if you take a job at the company, you've got to be on those four hours. OK, so that's how you get everyone to synchronize.

Prukalpa (38:02.254) Yeah.

Prukalpa (38:06.499) Yeah, that's the synchronous time. That's for the collaboration hours. That's for the like, you know, that's that's the huddle

Auren Hoffman (38:11.31) Okay, that's what I was asking. That makes perfect sense. So there's some period of time where no matter where you are in the world, you got to be, and for some people, obviously if you live in San Francisco, that means you're getting up a little bit earlier than you would normally do. For other places, it means you're changing your habits a bit. Yeah, exactly.

Prukalpa (38:28.346) Yeah.

staying up later. Yeah, yeah, yeah, yeah, yeah. And our only thing is we typically tell teams not to burn both sides of the candle. So don't wake up super early and stay up too late. So our teams typically, like we have teams in Asia and India, they typically start their day much later. And they go on so that they can stay up till the suspect wakes up. And our

Auren Hoffman (38:53.602) Got it. And so about like, during the, during the recruiting process, they know, okay, this is the hour. they're, they're opting into this.

Prukalpa (39:01.465) Yeah, yeah.

Auren Hoffman (39:04.334) Okay, interesting. What do you, when you think about, I mean, doing this remote first has both its advantages and disadvantages. How else have you made it work? What else have you done to make it work well?

Prukalpa (39:19.331) Yeah. So few things have been A, we have really believed in, like we really believe that only two ways you can run a company. Either you're all in office and the whole team isn't in office. Like at least like the team that's working together is in office or you are like fully remote. Like we don't believe in the middle thing.

Auren Hoffman (39:32.686) Mm-hmm.

Auren Hoffman (39:36.866) Yeah, yeah.

Auren Hoffman (39:40.994) Yeah, and by the way, I think I think the whole team has to be not in the office, but they have to be on the same floor like you said. They can't even be on different floors and stuff that you're basically remote at that point. Yeah.

Prukalpa (39:45.955) Exactly. Exactly. Exactly. Exactly. And so one has been just like making sure that that is important. The second thing has been being really intentional about human death. And like that has meant, for example, when we when we meet in person, we have this ritual, we call it, well, now it's quite like an atlant, an atlant special meal. And when we do a dinner,

We have this specific question that we ask that gets people to open up and talk about. So we have just these rituals for her.

Auren Hoffman (40:19.028) what is what's what's the questions?

Prukalpa (40:22.425) I mean, so it could be like, for example, there's like this one question that if an alien showed up on your doorstep, where would you take it for a meal and why? Or like, if you're gonna go to Mars, what is the one item, personal item you would take with you? Or if there's one person in the world, you could call and thank that you haven't thanked, like who's that person and why? And so it's just things that you...

Auren Hoffman (40:32.75) Okay, I love that.

Auren Hoffman (40:47.959) Yeah.

Prukalpa (40:50.691) It's the right thing. You could be on the same floor as a human being for an entire year and you might still not know them. And if you go to one of these things and they're actually well run, you could walk away after four hours and you could really know the human being. And so it's been architecting the in-person time and being really intentional about that. It's been creating a bunch of rituals in the way we run the team. So for example, we have

a very specific thing called unblock meetings. And an unblock meeting is specifically meant to get cross-functional collaborators to say, hey, here's the thing we're going to unblock. And there's a format that I'm in. So you write up a problem-supposed solution document. So it's been creating these rituals in the company of how you work and making it very specific to how you work. There have been things like we have a very big culture of gratitude and feedback.

And so every time we do things, we have something called I like, I wish, And people do these for each other. We do this for the company. You have an Atlan version of I like, I wish, I buddy for Atlan as a company. So just creating more of those, think. It's been trying to standardize culture. Culture is one of the hardest things, I think, because if you're in office.

and more than 50 % of the people like normed the culture, you just see it and it becomes you. It's just harder remotely. So we've written like a 75 page leadership handbook. And so it's really like, here's the, yeah, and it's like, here's dilemmas that you're going to deal with as a leader at Atlan. It's not about like being a leader, it's just about being an Atlan leader. what is, you know, like this is what it means to be a leader at Atlan. And how do we, how do we operationalize those things? So it's very-

Auren Hoffman (42:19.667) Yeah, yeah, yeah, it's easy. Yeah.

Auren Hoffman (42:26.318) I'm happy.

Prukalpa (42:45.691) I'd say like my previous company used to be a fully in office. Like we were in office, we had bum beds in office. were like, it was like a truly in office. Like it was like everything, like everyone lived within like, you know, one mile radius of the office. Like it was a very in office culture. Uh, and as long as it's fully remote culture, I think it needs, you have to be really intentional about being a remote company and, uh, you can't just slap on what works in person remotely. think you have to be really intentional about building a remote company.

Auren Hoffman (43:16.02) Yeah, interesting. Now you're a couple of personal questions. You're the only Prukulpa I know.

Prukalpa (43:25.615) Yeah.

Auren Hoffman (43:26.735) And you had mentioned before, it's kind of like a unique name. Is there like a story behind that?

Prukalpa (43:33.763) I am the only Prukalpa in the world. But if you Google me, it's like me. So that's my defense. People always ask me like, how do you know you're the only Prukalpa in the world? My thing is like, if you don't have like an online profile right now in the world, like, you know, there could be a Prukalpa somewhere in the world that's like not connected to the internet. But if you're, you know. Yeah. Yeah, it's just me. So my mom,

Auren Hoffman (43:48.45) Yeah, yeah, but I'm just, doing it right now. I'm Googling you, I'm open-eyeing you, and I can't find another one. Like, it's just you, you're the only one that shows up.

Prukalpa (44:02.811) My standing joke is my parents thought about SEO before Google happened.

Auren Hoffman (44:07.362) Totally, yeah. it good to have a name or bad? Do you think that's a good thing?

Prukalpa (44:14.651) I love it. I think it's great. I love having, I mean, have like, like I have all the like, it's so easy for me to get all the domains right, because it's just me. So I love it. I always have the app, I don't have the complete, but I don't even, I don't even need to like use my last name for anything. It's like pretty cool. There are some downsides, like for example, if you, you know, like anything I know on the internet, like, know, there's, you know, when I was a kid and I was on Orkut.

Auren Hoffman (44:24.694) Yeah. Do you always have the at Procopa handle?

All right, that's amazing, okay.

Auren Hoffman (44:35.704) Yeah.

Prukalpa (44:44.203) And you know, have these like really, you know, everyone has these like really funny, like, weird posts that they make when they're a kid, like, you you write like, WRT instead of it, you know, like, you you can actually find that stuff about me. No, if you research me enough, there's some embarrassing stuff that I can never get away from. But, but I love it. And yeah, my mom, mean, prokapa means brave and courageous. It's a Sanskrit word, which so so I'm named after it's actually the

Auren Hoffman (44:50.987) Yeah, yeah, yeah.

Auren Hoffman (44:55.614) Hahaha.

Prukalpa (45:13.195) one of the thousand names of awesome artists in India. And so yeah, it also means something cool, which...

Auren Hoffman (45:26.048) I love that. That's amazing. A couple of personal questions. What is a conspiracy theory you believe?

Prukalpa (45:33.787) I am not much of a conspiracy theory person, but I have a standing joke with my co-founder on this, which is that I think I genuinely think flight upgrades have nothing to do with points and have to do with some version of karma or something like that. Because my co-founder and I, I travel more than him. And every time we travel together, he gets the upgrade and I don't get an upgrade and I don't. And it makes no sense to me as to why.

Auren Hoffman (46:03.066) Wait, sorry, I'm not following. Walk me through this again?

Prukalpa (46:06.891) I, so like, you know, flight updates, like if you're like traveling in a flight, you know, I, you would think that upgrades have to do with points and you know, how much you travel and things like that. And I will never get an upgrade and my co-founder and I, when we traveled together, he travels much faster than me, but he always gets an update and it makes no sense. So, so I keep telling him, like, I'm like, this is like karma or like, it's like, there's something else in the world.

Auren Hoffman (46:12.963) Yeah.

Auren Hoffman (46:17.196) Yeah.

Auren Hoffman (46:33.774) Do you have some theory as to why he gets it and you don't?

Prukalpa (46:37.315) Yeah. I mean, I don't. Yeah. Yeah. Yeah. I feel like there's some, you know, like, he's done something in his life where someone has mocked him in some database or something like that. And he just like, it makes no sense to me as to why he gets upgrades and I don't.

Auren Hoffman (46:40.556) Maybe he's like super nice to the check-in person or something, or he slips him a 20.

Auren Hoffman (46:58.764) Interesting. I had a friend who always used to get upgrades and it was never clear why, but he, do you know the singer Moby?

Prukalpa (47:09.188) No, what is it?

Auren Hoffman (47:10.03) Okay. Anyway, there's a singer Moby who's kind of famous person and my friend looks just like him. And, and so the theory, our theory was that like everyone was mistaking him for like, for a famous person and they would constantly get like upgrading good. So maybe your co-founder like looks like Brad Pitt or something. Yeah.

Prukalpa (47:13.434) Okay.

Prukalpa (47:17.403) my gosh.

Prukalpa (47:22.177) Every my gosh.

Prukalpa (47:28.815) Maybe he looks like someone. That could be a reason. That could only be the reason.

Auren Hoffman (47:33.902) Cool. All Last question we ask all of our guests. What conventional wisdom or advice do you think is generally bad advice?

Prukalpa (47:43.981) I'm with you.

I have a few on leadership. I think the conventional advice on hiring leaders is typically this hire really smart people and let them go figure it out. there's some element of, actually, is, I recently wrote this post about micromanagement. And I personally think that the best way to set up a leader

for failure is to just hire them and then they like go figure this out. And the reason for that is I think they don't have context. There's clearly a theme here for me on context, but I think that context is really important. Like if you think about the job of a leader, it's about making great decisions. To make great decisions, you need context.

Auren Hoffman (48:20.038) Yeah.

Prukalpa (48:43.483) When you join a new company, you won't get context. There is you won't have like the years of context. And so I think it is your job to make sure that this person gets the context that they need before they can go be awesome and make great decisions. Otherwise they make suboptimal decisions. I think the worst decision is not a it's not a good decision or a bad decision. It's a suboptimal decision. And so something I believe is, you know, how do you

Auren Hoffman (49:08.271) Yep.

Prukalpa (49:12.355) Like for example, we have something called a decision review. So every time I on-board a leader, like their first 20 decisions, they review with me. And we go through them together. so I match 20 decisions.

Auren Hoffman (49:22.795) The first what? Just the decision, like they actually just walk through live with you.

Prukalpa (49:28.281) Yeah, they walk like, this is the decision I'm making. This is how I'm thinking about it. This is why I'm thinking about it. And so there's just, and so I personally believe that the way to solve problems is not just to hire smart people and let them go do it. I don't think you have to hire smart people. And I think you have to hire great people. But I think that there is a much higher responsibility to be extremely involved for the first three to six months of their journey in a new company.

Auren Hoffman (49:33.546) that's an interesting idea.

Auren Hoffman (49:44.14) Yeah.

Auren Hoffman (49:55.81) Yeah, yeah, I love that. That's really great. Well, thank you, Procopa, for joining us at World of DaaS. I follow you at Prukalpa because you have, of course, that handle on X. I encourage all of our listeners to engage you there. This has been super interesting and a ton of fun.

Prukalpa (50:13.083) Thanks for having me, Auren. This is awesome.