Customer Zero: How Databricks' Own Finance Team Runs on Databricks

Summary

Databricks' own finance team—acting as customer zero—has unified all transactional data from ERP, CRM, HRIS, and planning systems into a governed lakehouse on the Databricks Data and AI platform, reducing the financial close cycle from 15 days to 7 days with a target of 3 days.
Genie 1 enables finance executives to get answers to business questions in real time—such as customer ACV and product growth analysis—that previously required an analyst to pull, clean, and format data before emailing results, eliminating the 'which number is right?' debate.
Unity Catalog provides the SOX compliance and audit readiness layer with access logging and data lineage, while RowZero offers a next-generation spreadsheet that sits directly on top of Databricks for teams that still need spreadsheet-style workflows.

Customer Zero: How Databricks' Own Finance Team Runs on Databricks

The most credible proof of a platform's power is how the company that built it uses it to run its own business. Databricks' Office of the CFO has gone all-in on the Databricks platform to unify transactional data from ERP, CRM, HRIS, and planning systems into a single, governed source of financial truth, eliminating the 'which number is right?' debate that costs enterprises countless hours in every close cycle. In this fast-paced, demo-forward lightning talk, you'll see how Genie and AI/BI enable natural language financial analytics and forecasting, how Unity Catalog ensures SOX compliance and audit readiness, and how RowZero enables seamless data ingestion from spreadsheets into governed pipelines without disrupting the workflows finance teams already use. If your CFO or finance team is still living in spreadsheets, this talk demonstrates what it looks like when finance becomes a first-class data organization. Expect less theory, more demo.
Talk By: Michael Schaaf, Sr. Director, Finance, Databricks ;

Chapters

FAQs

How does Databricks' own finance team use the Databricks Data and AI platform?

The Databricks finance team has unified all transactional data from ERP, CRM, HRIS, and financial planning systems into a single governed lakehouse. This provides a single source of financial truth, and enables the team to use Genie 1 for real-time self-service analytics that previously required days of analyst work to produce.

What financial close improvement has Databricks achieved using its own platform?

By migrating data into the Databricks Data and AI platform and governing it through Unity Catalog, the finance team reduced the financial close cycle from 15 days down to 7 days, with a target of 3 days. Databricks Apps and AI automation enabled use cases previously handled by ERP systems to be migrated directly into the platform.

How does Unity Catalog support SOX compliance and financial audit readiness?

Unity Catalog maintains access logging and data lineage for all financial data, giving auditors a verifiable record of who accessed what data and when. This audit trail is described as a key differentiator that allows Databricks finance to use the platform as the core of its compliance stack rather than maintaining separate audit systems.

What is RowZero and how does it fit into Databricks finance workflows?

RowZero is a next-generation spreadsheet tool that connects directly on top of Databricks, allowing finance teams to work in a familiar spreadsheet interface while reading from and writing to governed Databricks data. It addresses the reality that productivity tools are not disappearing while keeping the underlying data governed within the platform.

Full transcript

[00:07] So, I'm here to talk to you about how we use Databricks internally in the office of the CFO. So, 10 years ago, you had all these different disparate sources of truth and systems. You didn't know if the data was
[00:20] right. You had to map it in manually in Excel. With Databricks, when I started here 7 years ago, I saw the potential for it to be the single source of truth for all of our data. Back then, it was mainly a
[00:34] tool for data scientists and data engineers. You had to understand how to code, tune clusters. Fast forward to now, it we're at this inflection point with Genie 1 where Databricks is usable by
[00:47] pretty much everybody in an organization. So, when we think about finance as customer zero, so we own all of the transactional data that we use to run our business. And so,
[01:00] our strategy first is to unify all of the data in a Databricks, um govern that data, and then with Genie 1, we're able to answer questions in minutes that used to
[01:12] take us days. So, with Genie 1, you can ask a question, whereas before you had to go to an analyst, they'd pull the data, they'd clean it, they'd have to put it into a slide template, email it off. So, now we're able to get answers to
[01:25] questions about our business in real time. So, um also, there's no more what's the right number because once we migrated the data into Databricks, and then we govern it, then everybody in the organization
[01:37] understands where to go find it. And we're able to close the books faster because with Databricks apps and um AI, we're able to migrate use cases out of
[01:49] our ERP or third-party solutions. So, I'll give you a couple of examples of that, but that's allowed us to bring our close from 15 to 7, and then we're targeting 3 days. And then also, Unity Catalog is super important when we think about governance
[02:03] and being able to make sure that we're our auditors are comfortable. We have like the access, the logging, the lineage around the the data. So, Unity Catalog is a major differentiator for us and really what enables us to be able to use Databricks as the core of our stack
[02:16] in finance. So, as I mentioned before, the strategy is unify every transaction in Databricks. And then, productivity software is not going anywhere. You're still going to
[02:28] have sheets, slides, docs. In fact, we we just closed a deal or signed a deal with a partner of ours, Rose Zero. Very exciting next-generation spreadsheet that sits directly on top of Databricks. I would go check them out. They're here
[02:41] at the conference. Um but a lot of what we were doing in spreadsheets or disparate systems is now being done in Databricks. So, when I done I'll show you the demo and you'll be able to see how this works in real time.
[02:54] So, when we think about the the app side of things, um as I mentioned before, we migrated two use cases in the order to cash workflow. Um our revenue sub-ledger, so we do all of our revenue recognition in Databricks.
[03:07] All of the data that we needed for revenue was already in there instead of doing it in our ERP and SAP or doing it in a third-party like uh Zuora RevPro. We built the revenue recognition logic in Databricks so that we have our
[03:21] building blocks of revenue for management reporting and our our uh GAAP-compliant revenue, ASC 606, in one place. So, you're easily allowed or you're easily able then to do analytics on both the building blocks
[03:34] and the actual revenue at the account level with Genie One. So, Genie One is what mostly what I'm going to talk about today because that with Genie code, I think, is this major inflection point for finance.
[03:48] Uh I'm going to tell you a a brief anecdote. I came from a place They have a big tower across the street and at the end of my tenure there, I was there for 7 years, we tried to build essentially what Genie wanted is now and we had 100
[04:01] people in the room and each of them owned a piece of the data estate and we had some app developers in the room and we said, "Okay, we're going to go build this app for one of the executives so that he can ask questions about customer ACV or
[04:16] ARR." And for 6 months, we probably spent millions of dollars and we failed because we would have to batch upload the data, the data wasn't governed, you had to make sure that each person was able to go and
[04:29] like actually deliver the data. You know, there's the LLMs were not a thing then, so like the the search functionality or what the output was that was giving to the executive was not very good. With Genie One,
[04:41] everybody has access to what we tried to build for that executive and it's not expensive anymore. So, it's a very exciting time with Genie One because it's sort of addictive. Like if you use it on your phone and you start
[04:53] asking questions about it, you don't even have to go to your your computer anymore. So, I'm going to stop there and I'm going to start showing you uh a demo. So, I teed this up in the spirit of
[05:07] the World Cup going on, France is the favorite right now. I'm not a big soccer guy, but um I've heard they're pretty good. So, what I'm going to ask Genie One is to analyze what the top 10 customers in France are, like I want to
[05:20] understand their product growth and their 3-year CAGR. So, I'm going to let that run and then while that's going on, what I wanted to show you was how we actually um
[05:33] how we how we unify all of our data into a place that's easy for folks in the finance team to navigate. So, for those of you familiar with Databricks, the catalog can be kind of um intimidating
[05:46] and even with Genie code which allows you to sort of like uh find things in the catalog more easily, it's still not something that I think is is super user-friendly at this point. So, our finance data team has done a great job of cataloging all of the
[05:59] different tables that we need use to run our business. And so, that's not just from where they live, but also how to get access to them. Because with Unity Catalog, if you want access to say commissions data, which is
[06:11] in here, that goes through uh approval process where because that includes compensation data. So, um that's super important so that we make sure only the right people have the right access to the data.
[06:23] And I also try to be transparent with folks. We are not done unifying our data. This is the road map right now. We're So, this has been um a 3-year journey for us. We started this 2 years ago, and we'll have all of this
[06:35] done by the end of this year. But, what I what I try to, you know, tell people is like Genie was amazing as long as your data is in Databricks and governed, but this is not easy to do. It requires really tight partnership between CIO,
[06:48] CTO, and CFO to do this. So, if I go back to Genie 1, uh it's still thinking, but you can see here that what it's doing is with business
[07:00] semantics and ontology, it's looking at it's answering my question by looking at the catalog. So, it knows what I mean when I say dollar DBU, it knows what I mean when when I ask it for 3-year CAGR.
[07:12] And then what it's doing is it's looking at all the other assets that have been created in Databricks to be able to get that context to give me the right answer. So, um this is may take a little bit
[07:24] more time to run. As you can see, it's not the fastest thing in the world, especially as it thinks. Um so, we'll come back to this. I thought this might happen. Um so, one other interface, this is all
[07:38] This is not real data, by the way. >>
[07:50] revenue, it's all of our sales capacity, we're looking at productivity. And as you can see on the right, I there's a spike in the trailing 7-day average around July 1st, 2023. I was able to ask Genie in the actual
[08:03] dashboard what was causing that. It's synthetic account, I think 2469. So, this is a great way like if you're in your interacting with data via the dashboard to get insights. Now, the problem with this is that it unlike
[08:15] Genie 1, it doesn't it doesn't have that full ontology. And like it's it could understands the business semantics, but it's not looking at the entire catalog. So, Genie 1 really is the future of how we're going to be doing analytics, and
[08:28] we'll still have dashboards, but I I view Genie 1 as as the future. So, um anyone here a Knicks fan? Knicks? Okay. Okay, awesome. All right.
[08:41] Well, I I brought a prop with me because um as anyone who has been in New York knows like I'm a Laker fan, but like you have to wear this in order to like stay safe. Um these Knicks fans are pretty passionate.
[08:54] So, um I thought it'd be to do something fun because I know it's the end of like a long 3 days of conference is I want to show folks like how I use Genie code to create some third-party data. So, what I asked it to do over
[09:08] here is I said, all right, create a data table that has all of the NBA teams and their conference. And then I thought I said, okay, well, I also want you to add their win-loss uh like wins and losses in here. And I
[09:22] was able to do all of this with natural language, and so it writes all the Python for me, it writes the table to the catalog. I mean, this is something that is like for someone that's non-technical like me is magic. I used to have to have people do this for me,
[09:34] and now I can just do it myself. Um, so then what I did is I don't know how many folks are um familiar with Lake Flow Designer, but I find this to be
[09:48] the missing piece for looking at data in a no-code low-code way. I mean, I I think folks are familiar with Alteryx, um but what I did here is I was able to just pivot that data. I was able to like filter on the Knicks and the Lakers and
[10:01] see what their win-loss is by year. So, it's a great way to be able to explore data and to be able to combine data sets. And then I asked uh Genie code or I'm sorry, Genie 1 to analyze
[10:15] if there's any kind of correlation between wins and losses and uh NBA team spend on Databricks. I can't show you how much each NBA team is uh spending, but I can I can tell you that
[10:28] my favorite team is spending a lot, which I like. Um very happy to hear that because I'd like them to win more games. Um so, you can see here that the key correlation was that yeah, there is there is some correlation is if you're
[10:41] in the top spend tier, you actually win 65% of the time versus 55 versus 30. So, this is just an example of how you can go through that workflow of like creating third-party data, taking the data that you already have in your
[10:53] lakehouse, and then using Lake Flow Designer to explore it, and then using Genie 1 to analyze it. So, one last thing as I go back to the initial Genie query. Okay, great. So, it
[11:06] ran. So, one thing I wanted to show you was it's giving me these insights. It's telling me who the top customers are, but when we think about ontology, so it tells me which sources it's coming from.
[11:19] So, I know this table because that table has essentially all of the customer consumption. So, I'm able to say, "Okay, I trust that where the data's coming from." And then with ontology, you're actually able to then see
[11:33] the context and where it's getting the context from. So, you can see here that it's defining annualized revenue run rate as most recent week's dollar revenue times
[11:45] 52. And that was created by someone, I don't know who this person is, but their name's Yiming Yiming Li. So, this is something that is fairly new in the platform and it's really important because LLMs without context
[11:58] don't give you the right answer. And so, this is something that I think is is really going to be important as you think about how you set your lakehouse up to be able to have all of your data and then also to make AI actually work for you.
[12:10] All right. Well, thanks everyone for coming. >>

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.