Sponsored by: RowZero | AI Spreadsheets and the Modern Data Stack

Summary

RowZero is a cloud-native spreadsheet built on the Databricks Data and AI platform that eliminates the core problem of teams exporting data to CSV and Excel, where all governance controls, row-level security, and compute scalability are immediately lost.
The solution handles 31 million rows from S3 and Databricks in seconds while inheriting Unity Catalog permissions including row-level security, giving business users the flexibility of a spreadsheet without sacrificing enterprise governance.
RowZero adds AI capabilities through natural-language queries powered by Databricks Genie plus write-back to Databricks with permission enforcement, enabling self-service data access without requiring business users to know SQL or bypass organizational controls.

Chapters

FAQs

Why is exporting to CSV and Excel a problem for enterprise data governance?

When employees export data from BI tools or Databricks to a CSV file and open it in Excel, all governance controls — including row-level security and role-based access control — are immediately lost. The data sits on a personal laptop where it can be sent anywhere, and it is limited to the local machine's compute and memory rather than the cloud-scale resources configured in Databricks.

How does RowZero handle large datasets from the Databricks Data and AI platform?

RowZero is designed as a cloud-native spreadsheet that connects live to Databricks tables, running compute in the cloud rather than locally to handle datasets like 31 million rows in seconds. It inherits Unity Catalog permissions so row-level security and access controls are preserved within the spreadsheet interface rather than disappearing on export.

What enterprise organizations use RowZero?

According to this video, RowZero's enterprise customers include Databricks and Amazon, where it serves approximately 58,000 users. These deployments demonstrate adoption at significant scale across major cloud and data organizations.

How does RowZero integrate with Databricks Genie for AI-powered spreadsheet workflows?

RowZero includes an AI agent feature that allows users to ask natural-language questions and have a Genie-powered agent translate those requests directly into spreadsheet actions. Users can also write data back to Databricks from within the spreadsheet interface, with permission enforcement ensuring that only authorized writes are executed.

Full transcript

[00:07] Thank you so much for being here everyone. My name's Brek. I'm the founder and CEO of Rose Zero. And today we're going to talk about spreadsheets, which are an uncomfortable topic for many data teams.
[00:19] I know this cuz I worked in data for a while. I've started Rose Zero 5 years ago and prior to that I was a principal engineer for AWS S3. So I've been in the data space
[00:31] for about 12 years. And one of the things I've learned during that time is that data professionals hate Excel. This is like a universal thing.
[00:44] They see it as chaos. They see it as antagonistic to everything they're trying to achieve as BI leaders. And yet it is ever-present in just about every
[00:58] organization we work with. So this picture describes the state of the world at basically every company in the world. We have a data warehouse on the left-hand side
[01:10] where we've worked super hard to set up great data governance. We have our medallion architecture with bronze, silver, and gold data. And that is exposed to end users in a governed secure way via BI tools like
[01:24] Power BI and Tableau. And yet the number one thing people do with these tools is export to CSV and open in Excel. And data teams get really
[01:36] get really upset about this. They feel offended that their dashboards weren't good enough. Um But it's not really a fair fight because the spreadsheet is just this super flexible interface. And the spreadsheet
[01:50] is self-service. So if you're trying to deliver self-service without the spreadsheet, you just can't do it. So, what is the problem with this picture?
[02:07] It's really that the spreadsheet It's really three problems with Excel. Excel is insecure, it's ungoverned, and it's slow. And these all follow from its local architecture, right? Excel originated in the '80s, when it was, you
[02:22] know, it was a different time. It was pre-cloud. Uh everything was on the PC. Uh and all the data and the compute happens locally. And so, as soon as you export a CSV file from a Tableau or
[02:34] Power BI dashboard, or a Databricks AI BI dashboard, that thing is is ungoverned. You may have set up row-level security in Databricks or role-based access control, and as soon as that data is on
[02:47] the employee's laptop, it's gone. It can be sent anywhere. Um the other problem with the laptop is it's limited to the compute memory resources of the laptop. We have all this abundant compute in the cloud, and
[03:00] yet, when the data is local, you can't access any of it. But an important thing to notice is none of those are criticisms of the
[03:12] spreadsheet interface. The spreadsheet interface is great. It's flexible, it lets us do formatting and custom formulas, and everybody knows how to use it, and it's fast and intuitive, and it has familiar
[03:24] keyboard shortcuts. Why do we have to give up all of that to solve the security, data governance, and performance issues? And that is what we uh
[03:37] aim to solve at Rose Rocket, which I'll talk a little bit more about. Um So, spreadsheets remain an essential data tool throughout the industry. Um and you don't have to take my word for it. Uh one of our customers is actually
[03:50] Databricks. Uh so, we announced last week that Databricks is a RowZero customer and here's a quote from someone in their FP&A team. Um, Databricks FP&A uses spreadsheets because they're the most flexible, familiar way to build complex
[04:03] models with auditable inputs and outputs. So, it's really these two features of the spreadsheet that are important to Databricks. Flexibility, which is unmatched by any other data interface we've ever created,
[04:17] and auditability. You can't have an AI black box inside of your audit envelope. And spreadsheets give you a great way for a human or an AI to show
[04:29] their work such that you can see what it's actually doing and see uh do the inputs and outputs make sense for what I'm trying to do. Another one of our customers is AWS.
[04:42] We have 58,000 users at Amazon. Amazon is a very spreadsheet heavy culture. And here's a quote from one of their data leaders. Dashboards answer the questions data teams anticipated, but spreadsheets answer the questions that
[04:54] change every day. And Databricks and Amazon are two very sophisticated companies. These aren't like some backwards, old industry companies who just haven't like gotten with the cloud. These guys are shipping the frontier
[05:07] that we're all using and yet spreadsheets are essential there. And so, I submit to you that the problem is not the spreadsheet interface. The spreadsheet interface is great. It's flexible and it's auditable.
[05:20] The problem is Excel's local architecture. So, how do we keep all the good parts about the spreadsheet interface without all of the baggage?
[05:32] And that's what we solve at RowZero. So, the that first picture I showed you at the start of the talk is the one up top. We can see the problems really are these right two boxes.
[05:44] CSV export is a security nightmare. And once you export into Excel, you're limited to the compute and memory resources of your laptop. Excel has a 1 million row limit, which hasn't changed since 2007.
[05:57] It's 2026, so we're 19 years down the road. Data sets have gotten a lot bigger since 2007, and they've also moved to the cloud. Uh Google Sheets, if you're using that, is even more limited than desktop Excel. It's about 10 times less
[06:10] capable. It has a 10 million cell limit, but in practice, you'll find it falls over around 10 tens of thousands of rows. Desktop Excel can go uh beyond that. So, we replace this old picture with
[06:24] this much simpler new picture. Row Zero, which I'll demo shortly, is a secure cloud-based spreadsheet uh that runs on AWS or Azure. It sits right next
[06:36] to your Databricks warehouse, and it scales just like Databricks does. Row Zero workbooks scale to 2 billion rows, and by running in the cloud, we keep all of the sensitive data off of employee laptops.
[06:49] We should never be shipping data to a laptop. That is the fundamental problem, and that's what the cloud architecture lets you solve. Um so, rather than talk about it some more, uh I will jump into a demo here
[07:01] and just show you how it works.
[07:16] Great. So, this is Row Zero. Everybody here already knows how to use it cuz it works just like Excel or Google Sheets. You can do data entry, uh and you have access to the full Excel formula language. In Row Zero, formulas
[07:29] are syntactically identical to Excel. Uh you can even import existing XLS workbooks, and they will just come in. Uh Row Zero supports the more than 500 built-in Excel formulas.
[07:41] Um and the keyboard shortcuts are the same. So, if you have like a finance team who's a power user that's using like the alt core keyboard shortcuts, we support those as well. So, any
[07:53] Excel user from kind of casual and novice to, you know, died in the wool can come in and start using Row Zero. I'm going to switch to full screen here. But the difference is we run in the
[08:05] cloud. So, I'm going to show two things. I'm going to do two Let's see. We'll start with an actually a file import. So, I could upload a file from my
[08:17] computer or in this case I'm going to connect directly to an S3 bucket. Row Zero also supports Azure blob containers. And here I've got a 1 gig CSV of all domestic flights in the US from 2018. We also support parquet files
[08:31] and JSON lines and whatever data you've got. As soon as I hit import, the data comes in. If I control arrow to the bottom, we can see we're on row 7.2 million. So, you
[08:44] wouldn't be able to open this in Excel. And from here I have access to the full Excel formula language. So, I could do like a quick count if and find out how many flights departed Seattle's SeaTac Airport that year. And when I hit enter,
[08:57] I get an instant answer. So, it's crazy fast over these large data sets because the compute's in the cloud. And we right-size this to your workload. We can scale this up and down as necessary.
[09:09] For enterprise customers, they're generally using us as their front end to their data warehouse. And so, today we'll talk about Databricks. So, I can insert what we call a connected table,
[09:22] which is like a live connection or pipe that connects Row Zero directly to an underlying database. If you know SQL, you can write that directly from the tool.
[09:34] You can use AI to talk to Genie to have it uh pull in data sets from your um from your metrics views. Uh or you can just drag and drop using uh the query builder here.
[09:46] Um for this demo, I'm going to use this demo Snowflake table, which has a row or excuse me, Databricks table, uh which has a row for every New York State hospital admission going back to 2009.
[09:58] This is 31 million rows, and you'll see we just pulled that in in 5 seconds. The data comes into this data table or data frame that's floating above the sheet, and this looks like a preview, but it's not. All the data's here. Like
[10:10] if you see this scrollbar over here, this is like a little window into your Databricks table. And if I start scrolling, this will scroll forever cuz all the data's here. Um to sort of prove that, I can right-click this guy and explode the data into the underlying table.
[10:24] And if I control arrow to the bottom, you can see we're on row 31 million, and you know, all of the Excel functions are still there. So, how many emergency room visits have there been since 2009? And I got uh an instant answer with this
[10:37] big data set. Now, the benefit to keeping the data in this connected table is that it's still connected to Databricks. So, I can go over to this calendar icon over here and set up a scheduled refresh for this
[10:50] table. So, this is now going to update every night at let's say 2:00 a.m. And now I can build my model or financial forecast or whatever I'm doing around this live updating data set. So,
[11:03] a really simple example is a pivot table. So, I'll put it right below this guy. Uh and for this data set, it's interesting to look at healthcare spend by admission type by year. So, I'll just
[11:16] drag and drop just like I would in Excel. Uh and Roseau supports all the formatting and data visualization features that you'd expect. So, we'll throw some currency formatting on this guy, uh
[11:28] adjust the column width so we can see what's going on. Uh and then we'll create uh, maybe a stacked bar chart to visualize what's going on here.
[11:42] Okay. And so now we can see this like clear dip in healthcare spend during the pandemic and the recovery afterwards. And this is a live updating model and forecast. So every night at 2:00 a.m. this connected table's going to update,
[11:55] the pivot table's going to update, and then the graph will redraw. This is a very simple example. It's super common for our customers to build very complex workbooks uh, in Roam Zero that may have like 50
[12:07] sheets. This is exactly what they do in Excel. Uh, in Excel the thing that's different is you typically see a raw data tab on the first like one or two sheets, which was a CSV export from somewhere that
[12:19] you've copied and pasted. And so in Roam Zero everything is the same except you replace that raw data tab with a Roam Zero connected table, which gives you that live pipe to Databricks.
[12:32] Roam Zero sits on top of the Databricks and Unity Catalog permissions model. So we inherit all role-based access control and row-level security from the underlying data warehouse. Uh, so the way most enterprise customers
[12:45] set this up is they'll have a Databricks user per employee. Uh, you can also do service accounts if you need to if you have a smaller number of roles. Um, and then when I load this workbook, it
[12:57] executes with only what I'm allowed to see. Uh, and if I share this workbook with someone else and they open it, it keeps the skeleton of the workbook the same, but it swaps out the contents of
[13:10] this connected table with the rows that that user has access to. So it sort of rehydrates with just their view of the underlying data. So super common use cases for this are like you're a sales
[13:22] manager and you have sales people are a bunch of different territories. They're only supposed to have access to the data from their territory. The sales manager can build a template that like shows how they want to slice
[13:35] and dice the data, publishes that template, and just shares it with their organization. And now everybody is looking at the same view, but only for the roles or only for the rows that they have access to.
[13:48] Um a few other features I want to touch on quickly. Rose Zero also supports write back to Databricks. It's super easy. All you do is just select the data that you want to write back. So, maybe I want to put this table back in
[14:01] Databricks. And you just I'll just right-click and say export to. And I'll pick my connection. Um and then I'll pick my, you know, catalog, schema, table, and I can just
[14:14] push it back. I can only write back to what I've been given permission to do. So, you have to give that user permission to do write back. Um it is very common maybe to only allow users to write back to a bronze schema,
[14:27] and that's where an ETL pipeline will pick up and publish the data through to gold tables where it can be consumed by other BI tools. Um Rose Zero supports multiple databases
[14:39] and connection types. So, you can be talking to Snowflake and Databricks in the same workbook or SQL on-prem SQL Server and Oracle and combine that data with XLOOKUPs or VLOOKUPs
[14:51] um just like a user would do in Excel. Um we also support real-time collaboration. So, you can see this share icon over here. It works just like Google Docs. So, you can have multiple users co-editing a workbook at the same
[15:04] time if you want. Um we do SSO and SCIM integration. So, you integrate your Entra or your Okta uh and you sync your groups with SCIM, and then you can share your workbooks
[15:17] uh you know, with the groups that you've defined in your uh in your identity provider. Um one last thing I want to touch on is AI. So, I'm going to make a new workbook.
[15:29] And so, I built that previous workbook by hand. And so, I can actually just open this AI agent and do that on full auto. So, I'll say, "Show me healthcare spend
[15:42] by admission type by year." This will take about 30 seconds. Um, and what's happening is the Rose Zero agent is going to figure out what connections it has
[15:54] access to, and it's going to use Genie to generate text to SQL to for what, uh, what query it should execute. That query will execute as me
[16:06] uh, with my row-level security permissions or our back flowing through from Databricks. It's going to put that in this connected table. You can see the rows ticking in right there. And then the Rose Zero
[16:18] agent is going to pick up the rest. Um, and let's see what it decides to do. So, it's looking at the schema and there it's created a pivot table.
[16:32] And I suspect we're going to get And so, there we get the same graph that I did by hand. Um, but this was just full auto. Um, the really neat thing about this and and the spreadsheet as a canvas for AI
[16:45] generally is the spreadsheet forces an AI to show its work. If you think about other AI tools like a chat box, under the hood the way they're arriving at their answers is by writing complex Python or SQL code to, you know,
[16:59] produce an artifact. Uh, but the business users don't know Python and they don't know SQL. So, they can't actually audit it. So, fundamentally chat box interface is a black box for
[17:12] how we're getting answers. And so, if it's correct, great, but if it's a mission-critical financial forecast or $100 million budget, you probably don't want to depend on that. With uh the Rose Zero agent, I can now just
[17:25] like double-click this pivot table, and I can see what it's pivoting, and I can modify it if I want. The spreadsheet is this live canvas as opposed to just this static chatbot output. Um
[17:38] one final I guess two final features that I want to cover. We've only got 2 minutes left. Um first is a bunch of the admin permissions that you get. So, just as you the data professional are an
[17:50] admin for your Tableau or Power BI environment, you're also the admin for your Rose Zero environment. So, uh as the org admin, you get access to a bunch of controls like which domains are
[18:02] in your organization, um and you can configure your AI provider. So, that Rose Zero agent that I mentioned earlier, it's all a bring-your-own-key model, so you can pick you know, AWS
[18:15] Bedrock and which models you want to make available to your org. You can go directly to Anthropic, OpenAI, or Gemini if you want. Rose Zero also supports workspaces. Um so, large organizations like Amazon
[18:28] will decompose their company into uh you know, an a workspace per business unit or per job family. So, Amazon has a workspace for Amazon tax, for Kindle, for AWS sales and marketing. Um and then
[18:41] you can appoint a workspace admin uh for each one to define who's a member, as well as export controls. So, if I scroll down here, I can do things like set export limits to disable download from
[18:54] Rose Zero, so data cannot leave if I don't want it to. Um this also turns off copy to clipboard, so you really cannot get data out. Um you can also set up life cycle policies to automatically clean up old workbooks um
[19:07] if you need to, or in the case of finance, workspaces is user typically admins want to retain that stuff for at least 10 years. So, just all the security and data data governance controls that you would expect, uh,
[19:19] but brought to spreadsheet workflows. Um, so I'll pause there. Uh, I think the last thing that I want to end with, uh, is just a quote from Databricks.
[19:35] So, Databricks says, "We chose Row Zero because it integrates seamlessly with our governed single source of truth in Databricks and handles our data volume with lightning-fast performance." So, uh, yeah, I think if Databricks is picking Row Zero as their front end, uh,
[19:47] it's something you guys should seriously consider. Um, thank you so much for being here. I'll stick around for questions. Thank you. Yeah. >>

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.