How Databricks Uses Databricks: Data-Driven Culture at Scale

Summary

Bruce Wong, head of data platform at Databricks, explains how the company runs its enterprise lakehouse at scale with every employee having access to the Databricks environment and every team expected to use data for decisions.
Genie democratized data access so non-technical teams like facilities can forecast office capacity using natural language without writing SQL, while culture surveys feed leadership dashboards for real-time employee sentiment analysis.
Databricks serves as customer zero for its own products, now running over 100,000 agents in production and building admin agents to govern the governance layer itself.

How Databricks Uses Databricks: Data-Driven Culture at Scale

Bruce Wong, head of data platform at Databricks, walks through how the company operates its own enterprise lakehouse at massive scale. Every employee has access to Databricks, and every team from facilities to HR to marketing uses data to make decisions.
Discover how Genie democratized SQL so facilities teams could forecast office capacity without knowing SQL. Learn how culture surveys feed into leadership dashboards for real-time employee sentiment analysis. See how Databricks became customer zero for agentic AI, now running 100,000+ agents in production, and how the company built admin agents to govern the governance layer itself.
🤝

Chapters

FAQs

How does Databricks build a data-driven culture across all employees?

Every Databricks employee has access to the company's Databricks environment, and teams across the organization — from facilities to HR to marketing — are expected to use data to make decisions. Bruce Wong explains that employees are given one opportunity to claim unfamiliarity with data tools, but not a second.

How does Databricks use Genie internally?

Databricks uses Genie to enable non-technical teams to query data using natural language without needing SQL expertise. For example, the facilities team uses Genie to forecast office capacity, democratizing data access well beyond traditional data professionals.

What does it mean for Databricks to be customer zero?

Being customer zero means Databricks deploys and tests its own new capabilities internally before customers use them, gaining real operational experience with features like agentic AI, Databricks Apps, and the lakehouse. This surfaces edge cases and use cases that directly inform product development.

How many AI agents does Databricks run in production internally?

Databricks runs over 100,000 agents in production internally across its engineering and data teams. The company has also built admin agents to govern the governance layer itself, applying agentic AI to platform management at scale.

Full transcript

[00:20] All right, welcome back to Summit Live. We are here at the Moscone Center at our wonderful studio and you can see this incredible uh expo hall. Wanted to welcome over 160 countries from around the world. Um 100,000 plus people all
[00:36] told for the conference. Actually, it's now the largest AI conference and also the largest data and AI conference in world history. So, I'm Ari Kaplan, one of the co-hosts, global head of evangelism. We also have I'm Holly Smith and I'm here from Devrell and I'm also one of your
[00:53] co-hosts today. Yeah, it is a fantastic time to be here. You just heard the keynote and now we've got a fantastic lineup of guests for the day and also sessions for everyone back at home. So, should we talk about our first guest? Yes, Bruce Wong. This is one of my favorite segments of all time, which is
[01:10] like how data bricks uses data bricks. Yeah, this is like this is the real detail. This is all about kind of how how we eat our own dog food. And if you've never come across that term before, it's not literally about eating dog food. Uh but please welcome Bruce. How you doing? Hi. Thanks for having me. Great to be
[01:28] here. Thank you. Uh so for the people at home, do you want to tell us what is your role at Data Bricks? Yeah, so I am the head of our data platform organization. Uhhuh. And what our team does is we actually own all of the production data as well as the production data bricks
[01:43] workspaces. And so that means a lot of things. Yes, we have an we're a data company. We use a lot of data everywhere. Every single employee has access to our data bricks environment. So it's fair to say you are the data team for the data and AI company.
[02:00] That's right. Okay. No pressure. How long you've been doing this role? Yeah. So I've been here uh four and a half years. Okay. So got to see the product really evolve tremendously over the past four and a half years. I've got to see uh data in
[02:16] AI summit grow like crazy over the last four and a half years as well. And I think it's fair to say I I think a lot of people when they talk about oh we want to be a data driven company. I think that's the first company I've ever worked at where I'm like no we live and breathe this. We truly are a data driven company. It is something that everyone
[02:33] is encouraged to do. I don't think anyone is ever allowed the excuse of like I don't really know how to like you have that excuse once but you can't have it a second time. Yes. Um I know we're going to talk a lot about products and things like that but like in terms of like a culture in terms of a datadriven culture your team are
[02:49] the ones who enable that to help you know the what how many people are we now 10,000 people 14,000 people way over around 12,000. All right 12,000 people let's say split in the middle. um you know you make all
[03:04] of that happen. What do you think is so unique about the data bricks culture that makes us truly put our money where the mouth is? Yeah, absolutely. You know, I I actually really credit the founders to this. Okay. And you know, the founders like, you know, were PhDs out of Berkeley and
[03:21] they're, you know, they're researchers by trade and and so forth. And so that intellectual honesty has really like been pervasive through the culture even today and that intellectual honesty sort of drives this like curiosity and data data information
[03:37] like make sure you know your data. Make sure that we're actually like you know honest about hey what what are the trade-offs here? Yeah. What are we actually seeing? How do we what is the data that's missing that we need? uh what does the data actually
[03:52] like help us make these decisions and so forth. Um and so just like we've seen that as like such an integral part of our culture and and I think that's sort of the exciting thing that we've seen over the the past year or so with Agentic AI is
[04:08] it's just democratized everything to even the furthest parts of the company because you no longer need to know SQL now. Yeah. Uh so I think for a lot of people at home they're kind of like roughly know what kind of data we might have floating around but could you just break it down for us in terms of like the
[04:23] different teams obviously engineering team they're going to be super data driven some of the sales teams again they're also care a lot about their numbers but it's not just that is it all it's all of the teams isn't it that's right and and that's sort of the explosion that we're seeing internally is that every single team
[04:40] uh has data and use cases that can make them more informed and more productive and and so forth. And so like the the ones I love to talk about are some of the the ones that I I was like, "Wow, this is amazing." So, uh I'll give you two examples. One from HR. Yeah.
[04:57] And one from facilities, right? Awesome. Like this is how data driven our culture is. So, facilities actually uses data. Uh and you know, we're out of COVID. We had to return to office. That was a whole ordeal in of itself. Uh but we
[05:13] actually use you know data from like badges to to actually help inform how we're doing from an uh return to office standpoint. Uh there's all sorts of stuff around like capacity planning for our offices. How do we forecast that?
[05:29] Do we have enough like food food inbound for our employees? Like there's a big deal here. And and I should say uh data bricks is investing heavily in facilities like opened uh long ago the London office
[05:45] which is gorgeous. We're going to get bigger. We have a London office right track of how many offices we open like Costa Rica's huge and I think and I think also it's it's if you ever get the chance to go to Jace Brook's office say yes they are gorgeous. I mean yes they
[06:01] are very much they are great places to be you know we hold a lot of customer events there obviously nice place for employees as well selfishly uh but that comes at a cost you don't want to end up with this office that's one um too big and you're not going to fill it with people but at the same time an office that is too small it is miserable to
[06:19] work in no one can really get their job done it's like what is the point of coming to the office you can't get your job done then also the growth side of it as well that's right like where are we going to hire and um you know if you're about to invest loads of money and decking out an office. You don't want to waste that because you got to move two years time.
[06:34] Absolutely. And I've seen I I think that's been the fun part seeing over the past four and a half years. That's like a small microcosm of how we've grown. Like we used to have issues around like running out of chips. Running out of Coke. Coke Zero is Coke Zero is usually the thing that we run
[06:50] out of at the Mountain View office and stuff like that. Uh but like we've actually seen that actually get smoother and better as like we've gotten more predictive and better analytics with things and it's scaled the way that we can actually have uh facilities. Yeah. And credit to our facilities team
[07:05] like they know what they're coming when it comes to facilities but they're not we didn't hire them for their SQL skills. Let's face it. Right. Right. And that's a huge barrier to entry. But like thanks to data bricks, thanks to Genie. Yeah. It's like we put the data in and like the entire team can just like check
[07:22] in where how are we doing? They know about uh you know the offices that we're responsible for. Um and yeah there's like that that tight quality uh you can feel it globally across across the company. So you also mentioned uh people ops and HR.
[07:37] Yeah. So, you know, one of the really really like we we talked about culture, you know, and and I think culture is such an important aspect to every company, uh, but ours especially and um, so I mentioned a little bit about that, but you know, like like many companies,
[07:53] we do culture surveys. We want to hear from all of our employees. We want to get feedback and and we want to take we we take that very very seriously. And I've always been part of like a a task force after a culture survey like what are the things that we can improve and so forth.
[08:08] Yep. And and one of the things that that we've been able to do this past year is we actually put culture survey into data bricks and uh our executive team actually uses Genie to analyze introspect find find uh learn from all
[08:27] that culture survey data through you know with Genie and and so forth. And so it gives our our leadership team a really really great pulse on how our people are doing. We can slice by geo, we can slice by function like very very easily
[08:43] um at amazing speeds. And I think also something that's uh maybe a bit more nuance. So I think back in the day it used to be like sentiment analysis and I've you know if I've spent time answering like a very like specific gripe that I have about how we can make our company better. It's very annoying
[08:58] to then know that it's going to get turned into like sentiment machine 200 and all of that's going to get flattened out. But having something where people can interact with the like, hey, what are people saying about I don't know the tools they need to do their job and you can have that kind of rabbit hole of like finding out the more nuance of it
[09:15] rather than just getting like a bland flattened. Yeah. Are they happy or sad? Right. Exactly. Binary. And and yeah, two things is also to encourage in the audience. uh look at data bricks what we're doing and you could do the same at your organization that is actually the
[09:30] the uh employee satisfaction survey a use case I hear with a lot of our customers too super easy to set up and uh much more intelligent you could ask questions of how to improve and I know data bricks as a result of being this transparency and really understanding
[09:48] how are people where are they struggling where's the opportunity we're now winning a lot of accolades glass door Newsweek like one of the best companies to work for. Um so wanted to point those two things out. Yeah. And you know I think that that's not by accident like you don't get those
[10:04] accolades by by just luck right by thumbs up thumbs down. Yeah. Yeah. And like are people happy or not? Right. you know, the the the level of intelligence that we're getting from our our own feedback surveys. You know, we can use Genie to actually re make
[10:20] recommendations on like what are what are we seeing and what are the what are the recommendations that our own employees have for improving our culture. Uh and and we actually like decide like hey this is this is great. We're we need to invest in mentorship a little bit more right and like um we need to invest
[10:37] in our tool chain and like unlocking more a AI tools and stuff like that. So, and and so we see that feedback loop in in the data we and it's able to surface actually into action and productivity so quickly now. So, so I'm going to ask you a question. Uh
[10:53] it's a bit of a curve. I am sorry. So, I know in the past there have been times where our data team wants to do a thing and they can't because the platform doesn't do it. Can you share any instances where that has happened where you've been working on the platform and it's gone back to the engineering team to say actually
[11:10] team come on get it together? Oh my gosh. Let's see. So, first off, my or is in engineering. So, uh and and that's on purpose. Yeah. Right. So, we're not we're not like uh another siloed or outside. We we are in
[11:28] engineering. We work very closely with engineers. Yes. Uh we work with the highest levels of our product management team as well as our tech leadership on a very regular basis. And you know this is actually one of I I actually feel like this is one of the really cool things about my team and
[11:46] just what we do at data bicks is that we are c we call ourselves customer zero for on purpose. And more often than not we'll we'll start using the product and and start using preview versions of things that are just not ready to go yet. Yeah. and we'll help find that product
[12:02] market fit. we'll help give feedback early on uh ahead of our customers uh to help understand what is this thing about like is there something here uh and we've seen the earlier versions of like you it's amazing to see the the stuff that we're announcing today and
[12:18] you know my team got to play a really cool role in that cuz we're like oh the early version was like way off right and but it it was through feedback through partnership and through trying to use like the latest and greatest parts of data bricks internally that we actually
[12:33] arrived at something like really fantastic now. Yeah. And I know we're going to talk about AIBI later. I remember there was at one moment I think we were playing for a third party data visualization tool and it came from on high of like hey hang on a minute. Don't we have one of these things? Uh why can't we use
[12:49] this for everything? Why does everyone not want to use our tools? And I remember I mean you probably were in the thick of it way more than I was. just kind of like this barrage of requests going in being like, I've been told that I have to use it and it doesn't have this, this, this, this, this, and this. You know, the the funny thing is I would
[13:06] say in the past 2 years out of the four and a half I've been here, y in the past 2 and 1/2 years, the the tide is so turned into like everyone wants the new stuff. And I'm like, we're we're almost there. We're getting ready. We're rolling it
[13:22] out. And I can't I you know I actually have to staff people on the team to like okay what does this mean? How do we roll this out into our environment? You know we obviously have our demo environments but everyone everyone wants it against real data right and that's sort of the the the big
[13:38] unlock. And so you know we actually have to work as fast as we can to keep up with product innovation. And like the number of requests that I get of like hey you know I want a lake base I want an app. I want uh agent bricks, right?
[13:54] Like, hey, can you enable this so so the whole environment has access to uh different things. Yeah, that's actually uh I would say that's actually probably like 30% of like what my team has to do right now. It's just to keep up with how fast things are moving and also the internal demand. The internal demand is
[14:11] just off the charts. Oh my god, the the things I sit in marketing but you know also adjacent to all the other teams and we have so many use cases but uh Genie and that whole umbrella just makes it super easy. I have on my iPhone or you can do on your desktop uh you know Genie
[14:28] one so I can look at the marketing related questions. We have a, you know, CMO dashboard. Our, uh, Rick Schultz, our CMO, uses it. Every one of our executives, we had Mate coming on, uh, recording of that of how he and his team, you know, get information of
[14:44] what's working, what's not, were there bottlenecks, were there opportunities, but I, it looks like we're only a few minutes left, but I wanted, we were talking beforehand about some super fascinating humans and agents. What do we have more of?
[14:59] Yeah. So we really saw the rise of Aentic AI in the last six months and so you know we've always had a lot of active users uh but our act like our active user/ agent count has gone through the roof and so you know we have
[15:16] uh we have over 10,000 employees I think it was you said 12,000 something in that range but you know at this rate we actually have over a 100,000 agents interacting with data bricks in the system now and and So, we've actually seen a huge
[15:32] acceleration. There's always someone's agent running something at every given point in time, right? And the other thing that's really been interesting for my team, we actually uh were struggling to keep up with that demand. How do you govern all of these agents? And so, we
[15:49] actually built our own agents to govern agents. We're like, there's no way we can do this like hand, you know, human versus agent. We're like, you know, we sort of asked the question, what if they were on our side, right? And and so we actually, my team has actually
[16:05] built an army of our own admin agents to help with that governance of all the agentic use of data bricks inside uh as well. Well, if you ever write a blog on that, I'll be happy to read it and I really want to hear more detail on that. Absolutely. We have we have eight
[16:21] different talks at the summit uh today and I you know the funny thing is like you know one of them's on cost one of them's on admin one of them's on data governance but like actually most of the talks are about how we you trained an agent to do each of those things in each
[16:36] of those areas and stuff. Okay. So, wonderful. Absolutely fascinating. So, for everyone at home, unfortunately, uh you won't be able to see it today. Uh but all of the videos do go online afterwards on YouTube, so you can um do that. Thank you so much for joining us.

Learn more about the Databricks Data and AI platform.

The information provided herein is for general informational purposes only and may not reflect the most current product capabilities or configurations.