Sponsored by: EY | Turning Spreadsheet Chaos into Governed Tax Intelligence

Summary

EY demonstrates how tax functions can replace fragmented, Excel-driven workflows with a unified, governed data foundation built on Databricks Apps, Lakebase, and Unity Catalog, improving accuracy, transparency, and speed while reducing operational risk.
This video shows a natural language interface powered by Genie that allows tax professionals to query KPIs, execute calculations, and understand compliance logic across 10 entities without writing SQL or depending on technical teams.
Regulatory changes such as tariffs and transfer pricing updates are handled as data updates within the platform, enabling finance teams to adjust plans faster and treat tax intelligence as a scalable, governed system rather than a collection of spreadsheets.

Chapters

FAQs

Why are tax functions still so heavily dependent on Excel?

As this video explains, tax functions have relied on Excel because it is flexible, familiar, and capable enough to handle complex calculations—but those same qualities make it easy to build solutions in a tool that was never designed for governed, enterprise-scale data. EY's Derek Huffman illustrates this with an analogy: just as a developer could code a video game in Excel cells, tax teams build systems in spreadsheets that have a better home on a governed data platform.

How does Databricks Genie help tax professionals access data without writing SQL?

Genie provides a natural language interface that allows tax professionals to type queries like 'summarize as if I am a CFO' and receive KPIs, calculations, and compliance summaries without SQL knowledge. This video demos Genie surfacing tax intelligence across 10 entities, enabling non-technical staff to interact directly with governed data on the Databricks Data and AI platform.

What Databricks products does EY use in its tax intelligence solution?

EY's solution is built on Databricks Apps for the user-facing application layer, Lakebase for real-time querying of the unified financial data model, Unity Catalog for governance and standardization of tax data, and Genie for natural language interaction. Together these products replace disconnected spreadsheet workflows with a single governed source of truth.

How does EY's solution handle regulatory changes like tariffs or transfer pricing?

This video explains that regulatory changes are treated as data updates within the platform rather than requiring manual spreadsheet revisions across multiple files. When tariff rates or transfer pricing rules change, the update flows through the governed data model and immediately affects calculations and compliance monitoring across all connected entities.

Full transcript

[00:07] Hey, good morning everyone. I'm Derek Huffman. I'm EY's Global Databricks Solutions Leader. So, I come from a background software engineering, data engineering, data science for the last 25 years. So, this job's really cool. I get to work with my teams, some are in the
[00:19] crowd today. Just to get to build cool things on Databricks and take to our clients. So, a couple weeks ago I was flipping through the channels and I saw this show come up and it was a video game coding competition. I was like, "That's pretty cool." So, I watched it for a little
[00:31] bit. There's teams, individuals, they're all competing to build these video games. Get some parameters, they get time, they get judged on it and they have to build it. I'm like, "That's pretty cool. I could I could probably still do that, right?" So, what do I do? I go and vibe code
[00:43] something, take a couple hours and build it out. So, what did I get? Couple hours, I got that built. Anybody recognize that? Maybe a little bit difficult back there. Anybody recognize it? Right?
[00:56] Pretty cool. I thought it was pretty cool. My wife didn't think it was that so much. So, the channel I was watching though was actually ESPN The Ocho. How many know that? Right? For those of you who don't know The Ocho, it's like a humorous
[01:09] ESPN channel. They play some just, you know, ridiculous sports. They play kickball, cornhole, dodgeball. Well, that particular show that I was watching that day was the Excel World Championships. Has anybody watched that? Just me? Okay.
[01:23] All right, a couple of you got it. Couple of you got it. So, not only were they coding a video game, they were coding a video game in Excel. Right? What I didn't say, I also have an accounting background. So, I could have done this in Python, React, you name it.
[01:35] I said, "Hey, let's get an extra challenge. Let's do this. Let's actually build this in Excel." So, I had in a couple hours a vibe coded game in Excel in cells. Right? See, it's pretty ridiculous, doesn't it? Like, why why would I do
[01:48] that? What benefit would that have? Why wouldn't I just do that in a native language? Well, we're doing that every day. All of our companies do that. EY's done it. Every one of you Every one of us has seen this situation, right?
[02:01] We're coding things in Excel and building solutions that we really shouldn't be. So, can we do it in spreadsheets? Sure, but why would we? Just like we're coding that video game in Excel, we're building all these systems in Excel that have a better
[02:13] home. Why are we doing this, right? I think we've all seen this Excel kind of spider web and and self-reinforcing loop that happens on the left, right?
[02:25] So, what what is this what is causing this? Well, people don't start off saying, "I'm going to build a system of record in Excel, right?" It starts off with, "Well, the data that I need doesn't exist in the system, the business logic isn't there. There's some
[02:38] technical limitations. I have to call IT and wait a couple months and build something and test. I just don't have time or patience for that. Or hey, the CEO's calling and they need a quick reconciliation of these numbers, so I need to very quickly spin something up
[02:50] to to answer that question. I can do it faster. I can get the answer I want with my custom business logic, but ultimately, if I'm building it myself and and I'm guilty of this over the years, if I built it myself in
[03:02] Excel, even across 100 different Excel tabs and formulas, I know what I did and I can fix it, right? So, I have that control and that trust. So, we end up with is this huge self-reinforcing loop, right? How many of us have seen that?
[03:14] We've all seen this, right? Thousands and thousands of these exist everywhere. This has caused us chaos. So, again, in this day and age, why are we doing this? I mean, the real impact of this Obviously, the the fragmented data,
[03:27] right? So, now we end up with multiple versions of the truth. We end up with not only data in different spots, manipulated data, transformed data that should be in data pipelines and govern. We end up with
[03:39] business logic that only lives in Derek's Excel workbook that is on his desktop and I hope he doesn't leave because nobody's going to be able to unwind that. Funny story, I had an Access database with an Excel workbook that I built as
[03:51] an intern over 25 years ago. 15 years later somebody called me asking if I could still support that. I had no idea how to do it 15 years later. What else does this do? Manual
[04:03] workflows, okay? So now I have all of these solutions. I get new data, I got to copy it out of this system, paste in this Excel workbook. Maybe somebody more advanced has figured out an automated refresh, but then I got to make sure these tabs link, trace it. We all know
[04:15] the drill, right? And then we have kind of an inefficiency. This is all leading to this inefficiency of our skilled talent. So all of our people that should be making better decisions are all wrangling data in Excel. So at
[04:27] the end of the day, that's really driving to impact of 50% or more in productivity and cycle times for decisions are 50 to 70%. Things that should be decided in in days or hours are taking weeks or even a month or
[04:39] more. So we look at If we look at all of our functions and I started looking a few months ago at our business and what we do for our clients, what is
[04:51] more Excel intensive than tax, right? Not many people can think of anything more than that, right? So I said, "Let's take this this theory. Like can we do something better in tax? Let's get rid of Excel in tax." That's
[05:03] almost sacrilegious in the tax world, right, to say that. So can we get better tax planning? Can we get lower costs for our customers and IT spend in managing this? Can we get improved security and reduce cycle
[05:15] times? So we set out with this vision. Can we automate the traceability, get our tax professionals, the CPAs, the tax attorneys, trust in that calculation all
[05:28] the way back to source so they know exactly what happened? Can we get the governance of the life cycle and and speed this up? So in today's day and age, we cannot do this anymore. Like we've all seen what's happened over the last couple years with
[05:40] with tariffs, with economic impacts, suppliers going out of business. We need to quickly model something. We We just can't keep operating like we have for 20, 30 years anymore, right? So, we need a better option.
[05:55] So, what did we do? Obviously, we did on Databricks, right? I said, let's try to do this as much on Databricks as we can. Now, I got to go back and revisit this this week after uh Monday's keynote, right? With all the things that were announced, so we'll update this a bit, but I'm excited for that. Shiva, I know is probably uh worried
[06:08] about all the work that we're going to have uh come Monday to rebuild a lot of this. So, we built a solution, and ultimately tax takes data out of ERPs and other source systems there at the bottom. So, you know, let's connect it with SAP BDC. Let's pull trial balances.
[06:22] Let's throw all of those those records that tax keeps in Excel spreadsheets and business logic and context and lake in lakehouse and lake base. Let's build AI agents to automate a lot of what tax teams are doing, not only to do the
[06:34] calculations, but the checks and the reviews and the recommendations. Well, you know, most tax people aren't SQL experts, they're not Python engineers, so naturally we need some kind of a way for them to interact
[06:46] naturally. So, perfect time to introduce Genie. Let's wrap all of this with uh with apps. We'll build some custom user-friendly apps that are very lightweight that allow these teams to operate seamlessly and very um
[06:59] effectively in in what we used to be their Excel world. So, we uh first demoed this to the team, and our our goal with our tax teams that are serving our customers as well as our our clients' tax teams,
[07:14] we want to give them a natural language way. They want to speak in English, they want to or whatever their native language is. They want to speak in tax, you know, jargon and interact with it. They don't want to have to convey requirements and work with a team to
[07:28] build a report or revise a report or implement this new policy. As tax law changes around the world, what other whatever type of tax you're looking at, something's always changing, whether it's state and local, European, federal,
[07:40] international, transfer pricing. There's so much going on that we need to continuously adapt and we just don't have that time to keep engaging. So, how can we empower our tax teams to do that on their own and still keep that trust and speed?
[07:58] So, what we did This is all sample data. Don't worry, nothing confidential is getting out here. It's all mock data. Some of it may not make sense if anybody's a tax professional, attorney, or CPA in the in the audience here. But, we we set this out like what are those
[08:10] common scenarios that they would do? Not only building the tax calculations and the filings, but what are those questions that need to be answered and who are they communicating to? How can we streamline that communication?
[08:23] So, you see what we did. Um again, all the the infrastructure, the the engagement that we had to get the data, maintain tax policy documents, read it, store it. We won't show all that today, right? That's that's not the fun stuff.
[08:35] But, when we get Genie on top of this, you can see what we did. So, on the left here, you'll see that there's a financial briefing. Now, what do our CFOs want to see? Do they want to see all detailed records and the walks of the calculations and all of that? No,
[08:48] they don't they don't want to see that, right? They want to see what are the key metrics I care about? Numbers. What are the highlights of those metrics? What do I need to consider as a CFO or a
[09:01] C-suite executive, right? They want a quick snapshot. So, our solution we built with this prompt. We said, "Give me a uh briefing of our performance or our tax situation um as
[09:14] if I'm a CFO." That's all we told it. That's all we told it. So, in this case, we had several entities. I think we had 10 or eight entities that we had in total. So, you can see on the left what it did.
[09:26] Oh, it had 10. So, it says, "We have 10 entities." Genie was more right than I was. So, you have 10 entities. Here's your pre-tax profit, your income tax liability federally, the breakdown of the consolidated snapshot.
[09:38] But, what's interesting, if you can see on the bottom here, note on fines and penalties. I didn't tell it about this. Our tax Our tax folks didn't didn't code this in. It knew that context, and it looked at
[09:51] those those amounts, that $386,000 of fines and penalties, that's a non-deductible expense. It had that context. It was aware, and it knew a CFO may actually care about that.
[10:09] It also notes at the end, this should be reviewed for compliance purposes. It's really interesting point, because you can make a mistake in a lot of regulatory, you know, agencies, you know, reporting.
[10:21] If you mess up in tax, that is millions and millions of dollars of fines, penalties, interest, you know, potentially public news that you don't want to get out. So,
[10:33] so tax is one of those areas we need to make sure that that data's right, and we need to highlight to the humans in the loop here what they need to care about, and where do they should they dive further if there is a risk. So, this is
[10:45] one of those extra um steps or checks that that Genie provided us. Point out to the CFO, "Hey, maybe talk to the controller. Maybe talk to some tax professionals. Make sure that that you're doing a double click on
[10:57] this for compliance purposes." On the right over here, again, going it understands all of those key content the key contexts. What level does a CFO want to set? You know, I'm just looking at my entities, my pre-tax
[11:11] profit. What is the highest to lowest ranking? What are my key book to tax adjustments, right? So, so as you don't know, there's the books that you report in your financial statements, and then tax looks at it completely differently. Cuz why not, right? We have to adjust
[11:24] everything, which makes it even more complicated, which is why we use Excel, and we use all of the other systems. So, you can see here are the big deal. So, if I'm I'm looking at a CFO and I'm looking at at my, you know, 10K, my 10Q and I say,
[11:37] "Okay, here's my numbers. The tax is telling me something different. Why is this number different than this?" Remember those reconciliations that we had in Excel on that first that that web? Right here. Here's my key adjustments, my key figures, why this is happening to
[11:50] somebody that's not a tax professional. So, it summarizes all of that, but if again, we look at the bottom,
[12:02] it's actually highlighting that I have a deferred tax asset. If I'm not aware of that, that's a huge benefit that I have that I can take advantage of for the coming years. So, again, just one of those checks that I don't have to have a tax accountant
[12:14] detailing every line item, every detail up to a CFO. And if hey, if something slips through the cracks, it's right there. Right? It's very, very straightforward.
[12:27] So, we continue that. Intercompany eliminations is probably one of the most complicated things that happens in a company. And we're we're looking at that here and it's explaining intercompany, which is huge area of Excel, huge area of of
[12:39] manual labor. So, I'm not going to go into intercompany eliminations. I'm not going to bore you on, you know, Thursday at 9:00 a.m. on the last day of a conference. But I'm going to highlight down here. Look at the here number five. Top three items for tax planning.
[12:52] Okay, we didn't we didn't code that. We didn't code a rule. Historically, our tax teams would have had to analyze that, summarize that, figure out what are the the top three things that are most important. That's hours, weeks of work.
[13:05] We get that summarized here. Genie just told us. Like, what do we think is most relevant for what it knows of the tax situation I was looking at. Investigate entity 5102's pre-tax loss.
[13:17] See if that's able to be utilized, if I can carry across all of my other entities and get a benefit out of that. There's something here on my transfer pricing service eliminations. It knows
[13:30] because of this intercompany elimination that's happening, I need to make sure I have arms length transactions for all of those and so that I can take advantage of those eliminations. Very important compliance step to make sure that I'm getting the right documentation, the
[13:42] right reporting, the right transactional structures to be able to take advantage of those. And then it's pointing out some anomalies that just look odd in the data. It's not saying it's wrong, but hey, there's something weird here. You may want to go look at this.
[13:54] And then the five key takeaways and action items. So again, summarizing those key bullets, what do I care about? What do I need to take take forward? It's not a big PowerPoint that I had to make. It's not a big Excel sheet. I don't need to go back and forth with my
[14:06] tax team to understand and get inundated because this is one sliver as a CFO that I need to care about in a wide range of things.
[14:18] This one was fun. So one of the things I I talked about to a tax team is they said, well, we we have to build these reports. We have to build pivot tables, right? Excel. We build these pivot tables, Excel row and they I want my system to kind of put that out
[14:31] for me automatically, but if I need to change something, I need to add something, I need to change a calculation, I need to transpose the rows and columns. I need to call my IT team and they need to build that report and then I need to test it. And hey, when I do transpose the rows and
[14:44] columns, I change this business logic and I broke it. So I need to go back and retest everything all the way through. So I asked the engine here, say, okay, the standard report has rows
[14:56] that are entities, columns that are my line items, my adjustments. Well, that's not how anybody looks at it in tax world. So believe me, this was the like drop the mic moment when I showed this to tax. I just said, show me that table,
[15:09] but transpose the rows and columns. I thought this was going to be super simple and not impressive. Like everybody got hung up on this. And it did it in a second. It just I'll put that report. Everybody was wild with this because they would have had to call that
[15:21] IT team to do that before, and they would have to do all that testing in the update cycle. Oops. On the right, so the other the other thing that we we mentioned earlier, we need to capture data. We need to capture
[15:33] information. Doesn't exist in source systems. We What are those things that require us to do those adjustments into the tax view of the world? Well, got they capture in Excel, right? And a whole bunch of Excel spreadsheets
[15:45] across dozens of people that are individually all tracking this. And then one poor tax leader tax department or VP of tax has to make sure that that whole team is QA'd all of it. What if we just build an app? Well, they like Excel, so we just made a
[15:57] table. So, let's build a table and give them the ability go in, edit all of that, add rows, edit fields, change whatever you want. They're like, "Oh, we can't do that. It's not audited." Well, yeah, it is. We we've got unity, we've got lineage, we
[16:09] have the whole audit trail now that we're saving behind the scenes, and we're locking it down so only people that are able to uh do this by the company's permissions are able to do it. So, we can go in and allow them now to edit their own data as if it was a source of uh
[16:22] record a system a source system. So, all of that's automatically integrated all these calculations. So, imagine we're integrating with Unity Catalog. So, again, we want to go back that trust aspect, that last bullet that
[16:35] I mentioned that overarching is all of this is really cool, but if they can't see how did I get to this number and every step that happened along the way, it's all worthless. So, you see here, we've given a display of all of the rules, all of the
[16:47] permissions, all of the the data quality of the checks that have happened. Here's everything that happened, and I can turn those on or off. If I want to add a new rule and turn it off and test something, run tax uh scenarios, I can do that now, and I can see what are the the rules
[17:00] that were applied in this calculation, and what was that net effect? Yeah, we don't have to code in all of it. We don't have to ask somebody what that net effect was. We can see our lineage all the way from our calculation at the bottom all the
[17:13] way up to our source of truth that came from the the ultimate systems at every step of the way. So now I'm a tax reviewer. I've got all of these calculations. I have, you know, have to get the the books closed. I have to get my filings done.
[17:25] Did I miss something? Right there, our tax reviewers. If you only can focus on five things, look at these five things and make sure you QA these numbers. It's finding data anomalies. It found out that there's some null values coming in. Is that
[17:38] accurate? Did we get bad data? We don't have any GL account backup for something. Right, I again, we didn't code any of this. Genie knew this. It it it correlated this and merged this context with our data.
[17:50] No transformation rules, current currency exchanges. All of these things were kind of factored in of high-risk areas to go double-check to make sure we don't make any errors in our tax filings and we don't get hit with those, you know, massive fines and then we don't
[18:02] have a, you know, a a black eye here as our tax team of making a mistake. So this quote, I said this is like sacrilegious to say to a to a tax team to say we're getting out of Excel.
[18:14] This is what our tax leaders, the second we demoed this, that was exactly the quote. I made sure I had that verbatim and I'm capturing it because, yeah, I'm I'm I'm memorializing this when a tax team is saying this that we can get out of Excel. This is this is a, you know,
[18:27] pretty senior tax leader that said this. So as you think about wrapping this up, right? None of us love Excel spreadsheets. We know that our companies are dependent on it.
[18:39] We can reduce this. So, not only with what was there last week in Databricks, but what was there this week and announced, we can even take this you know, a step further. So let's get rid of our um reliance on spreadsheets. Let's get that data into lake base.
[18:51] Let's get it into lakehouse architectures. Let's get all of our rules and lineage. We can lower our operational risk and get higher transparency. All of the information's there. With Genie, we've got the ability to just make it user-friendly for somebody to get to. We
[19:03] don't have to call somebody to get those data traces anymore. We don't have to have them dig into 20-year-old SQL code that nobody remembers how it was written, right? And we can most importantly refocus our talent. So, we have a lot of bright
[19:16] people, not only in tax, but, you know, finance, supply chain, operations, sales. Why are they in Excel? We need them out doing the work that they need to do, not managing and QA'ing and fixing Excel workbooks.
[19:30] So, hopefully all of you uh have a takeaway out of this and happy to follow up with anybody offline on this and some other areas. So, thank you, everyone.

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.