Spreadsheets in the modern data stack: security, governance, AI, and self-serve analytics

Summary

• Spreadsheets persist: Despite cloud data warehousing and BI tools, spreadsheets remain the most ubiquitous analysis tool for business teams across all departments.
• Self-serve analytics gap: BI tools promise self-service analytics but fail to replace spreadsheets, forcing teams to export data from dashboards for analysis.
• Modern spreadsheet tools: New AI-powered spreadsheet platforms bridge governance, security, and scale, enabling safe data analysis without BI tool complexity.

Spreadsheets in the modern data stack: security, governance, AI, and self-serve analytics

Talk by Breck Fresen, CEO, Row Zero (Sponsor Speaker)
Despite the proliferation of cloud data warehousing, BI tools, and AI, spreadsheets are still the most ubiquitous data tool. Business teams in finance, operations, sales, and marketing often need to analyze data in the cloud data warehouse but don't know SQL and don't want to learn BI tools. AI tools offer a new paradigm but still haven't broadly replaced the spreadsheet. With new AI tools and legacy BI tools providing business teams access to data inside Databricks, security and governance are put at risk. In this video, Row Zero CEO, Breck Fresen, will share examples and strategies data teams are using to support secure spreadsheet analysis at Fortune 500 companies and the future of spreadsheets in the world of AI. Breck is a former Principal Engineer from AWS S3 and was part of the team that wrote the S3 file system. He is an expert in storage, data infrastructure, cloud computing, and spreadsheets.
Intelligent analytics for everyone: https://www.databricks.com/product/business-intelligence
BI meets AI: https://www.databricks.com/resources/ebook/business-intelligence-meets-ai
Explore the only unified platform for agent systems: https://www.databricks.com/product/artificial-intelligence
See all the product announcements from Data + AI Summit: https://www.databricks.com/events/dataaisummit-2025-announcements

FAQs

Why do business teams keep exporting data to spreadsheets?

Despite BI tools claiming to enable self-service analytics for 20 years, most business users still prefer spreadsheets for analysis. BI tools often require SQL knowledge or are too rigid for ad-hoc exploration, making CSV exports the fastest path to insight.

What are the security risks of spreadsheet analysis on cloud data?

Uncontrolled spreadsheet exports lose data governance and audit trails, expose sensitive information to unsecured tools, and create compliance gaps. Without governance, teams may lose data lineage and violate regulatory requirements on sensitive datasets.

Which BI platforms offer AI-powered self-service analytics?

Modern BI tools are adopting AI to bridge the gap between data and business users. The video discusses how semantic layers and AI-assisted query generation enable business teams to analyze data without learning SQL, finally delivering on BI's decades-old promise of self-service analytics.

How can modern data stacks address the spreadsheet problem?

Integrating spreadsheets directly with cloud data platforms through connectors like Databricks Genie allows business teams to work in familiar tools while maintaining governance and data security. This approach honors how business users actually work rather than forcing them into rigid BI interfaces.

What is the best AI for business teams to analyze data?

Genie and similar AI assistants embedded in spreadsheets and BI tools enable business users to query data through natural language. These tools eliminate the SQL barrier while maintaining governance through the Databricks Data and AI platform.

Full transcript

Uh thank you everyone for coming. Uh. I'm one of the founders of Row Zero. Uh so Row Zero is a spreadsheet for the modern data stack and I'll get into the details of the product uh later on in the presentation. Um yeah, glad to see some faces here. Spreadsheets are not the sexiest topic in the era of AI, but there are those of us that care a lot about spreadsheets. So, I suspect you're here because you either love spreadsheets or you're on the data team and spreadsheets are a pain in the neck.
Uh, so we'll address some of those issues. Uh, so if you take a look at this diagram, I I suspect that it is a familiar workflow uh for many of you. Uh, you have a data warehouse. You guys are using data bricks. You put BI tools on top of your data warehouse. So Tableau, PowerBI, uh Looker, Quicksite, your team will build dashboards that your business partners consume. And uh the most common action on those dashboards, I suspect is export to CSV so the data can be opened in a spreadsheet.
Um and we hear this from tons of our customers and every company we talk to. Uh for 20 years, BI tools have been selling the promise of self-s serve analytics and it still has not materialized. Spreadsheets are the most ubiquitous data analysis tool in the world. Supposedly, Google Sheets and Excel have billions of users. Uh and so spreadsheets aren't going away. Uh there are many problems with this workflow. Uh which I will start to dive into.
Uh so you might be wondering why why are people exporting data from BI tools? Every BI tool claims it's self-s serve. You can get your business people to work with data, but but that's just not the reality. Uh we hear this from every company. If you talk to consultants, it's a a running joke in the data world that every BI every BI dashboard is just a very fancy CSV export. Now the reason BI tools evolved, this happened 20 years ago, is data sets started to outgrow the capabilities of Excel.
And uh you know when Excel was invented 40 years ago and even for the first 20 years, a 1 million row data set was not uh not very common. But today, every business user regularly comes in contact with data sets that either exceed the Excel row limit or cause Excel to crash. Um, and for those of you who are not aware, Excel has a 1,48,000 row limit. Uh, it it hits performance constraints long before you actually get to that point.
Google Sheets is about 10 times uh less powerful. It has a 10 million cell limit, so it will crash much much faster than Excel. Uh, BI tools offer power. So, you can put them on top of giant data sets, create these dashboards. Doesn't matter if your data set is a billion rows. Uh, your BI tool can handle it. they connect to the cloud data warehouse. When Excel uh was first designed, cloud computing wasn't a thing. Uh the same goes for Google Sheets.
Google Sheets launched in uh 2007, which I think is the same year that S3 launched. And so cloud computing had not been invented. Um and then the nice thing about BI tools is they offer better security. Uh you don't have data floating around on laptops. You can enforce uh security from your data warehouse, arbback, and RLS. The problem that dashboards and BI tools have created is every time a business business partner comes to a data team and asks uh for access to a data set or a different slice of data, the data team immediately creates another dashboard and the first thing that business person does is export to Excel.
And so what ends up happening is data teams will create thousands of dashboards. Uh, our largest customer has a data team of 200 people and they've created over 3,000 uh, Tableau and Quicksite dashboards. I thought that was impressive and I was talking to a guy earlier today who said they did an audit last year. This is a telecom company and uh, they had created over 9,000 dashboards in the past five years. Uh, and so not only is that a waste of time if people are exporting to Excel, uh, but it's hard to maintain.
Like that's a a burden on your team. Uh the challenge with dashboards and BI tools is they're rigid. They're only as good as the use cases and analyses that are pre-anticipated by the creator. Uh so you have to anticipate the filters or the group buys or the aggregations that your business teams want and it's impossible to anticipate all the different configurations which is why business teams come back to the data team again and again for new slices of data, new dashboards and you just end up in dashboard overload.
Um, and then lastly, despite what BI companies say about self-s serve analytics, business people don't create dashboards. They will consume them, export from them, but they don't want to learn Tableau. They don't know SQL. Uh, they just want to get their data in a tool they know how to use. Spreadsheets aren't going away. I assume you've guessed that based on the subject of this presentation. Uh they're ubiquitous for the main reason is because of the UI.
Everybody knows how to use a spreadsheet. You're introduced to them very early on in your career, oftentimes in high school, college. Uh it's just it lets you see your data. It lets you scroll through your data, interact with it on a cellbyell basis. And it's the most intuitive UI we have for analyzing data. your business partners in finance, operations, sales, and marketing uh only want spreadsheets. Despite uh the proliferation of BI tools, uh they're still going back to the spreadsheet time and time again.
But spreadsheets have problems. So, like I've said, I love Excel and Google Sheets. Uh but they they come with some some real heartache. So, the issues with Excel and Google Sheets are not the UI. The spreadsheet UI is great, but spreadsheets have this stigma because of their power, their connectivity, and their security. So, as as I've expounded upon, uh, Excel and Google Sheets have these small row limits, but outside of power, they're not connected.
Uh, so Excel obviously can connect to data bricks, but you're using an OBBC driver, which is tied to a laptop, which means you cannot share that file with other people. Uh, it's also really slow and fragile. So with modern data sets will take a ton of time to load data uh in an ODBC driver. Uh so they're not connected. Google Sheets also has similar connectivity challenges. Um there have been some advancements but generally they're really slow connections and brittle.
Uh and then the last is security. And I think this one probably um gives data teams the most constrnation. Uh, anytime somebody is downloading a CSV or opening an Excel file on their laptop, that's now locally stored on someone's computer. It could be emailed to the wrong person. The laptop could be lost. Uh, it's now totally untraceable. So, you've you've lost that data. And the security risk with spreadsheets is this latent risk that many many big companies have just accepted.
Uh we've talked to a number of companies that have said every time they do a security review, a data security review, the top risk is spreadsheets stored on people's laptops. Uh we talked to a Fortune 500 company recently whose CEO's laptop was stolen in a major airport and their their biggest concern about that theft was the spreadsheets stored on that laptop. Um one of our customers is again a big company. uh they have had CSV exports from their BI tools walk away and go to competitors and they want to eliminate the CSV risk and the Excel risk from their company.
Uh but the UI is not the problem and so there should not be this stigma around spreadsheets. There should be a stigma around locally hosted uh insecure unconnected disconnected and uh unpowerful spreadsheets. And so if you could imagine a spreadsheet in the modern data stack, it would be powerful enough to handle modern data set sizes. Uh so it would open billions of rows. You would have the same formulas as you get in Excel and Google Sheets.
You could pivot, graph, filter, do everything people are used to doing. So there'd be no learning curve. It would be built in the cloud, so it could connect natively to data bricks and other cloud data warehouses. Uh you can read from it. You could also write back to it. So complete interactivity with your data store. Uh and then you would have enterprisegrade security. So you'd enforce arbback, rls, inherit those from uh your data warehouse.
You could restrict data export. Uh maybe you could configure it so data can't leave the spreadsheet, but business teams still get the spreadsheet they want. Uh and you'd allow sharing, collaboration, but with full traceability. And that is what we've built at row zero. Uh the diagram on top is the one I started with and I've sort of covered all the risks with that. You've got way too many dashboards. CSV exports are a security nightmare.
Uh and the legacy spreadsheets have performant issues. You can't do modern data analysis in legacy spreadsheets. Now with row zero, we sit on top of data bricks. You can import data from data bricks, write it back to uh your cloud data warehouse and we live alongside BI tools. Uh when you think about ways to handle spreadsheet proliferation, data teams essentially have two options. You can either acknowledge that it exists and let people download uh from your BI tools or you try to clamp down on it and then you end up creating thousands of dashboards.
And in a row zero world, you maintain your executive scorecards, your highle dashboards. uh those are still great for reporting and we're not a replacement for those. But all the other ad hoc analysis that your business teams are doing can fall into row zero. You can share connected data sources with them where the data is just dropped into a spreadsheet and they can perform all the analysis they need without the help and support of a data team.
Now you might be wondering in this era of AI, do we actually need spreadsheets? Uh valid question. Uh we believe yes. Yes, you do need spreadsheets in the era of AI. Um for two reasons. Uh so LLMs are great, but there are two inherent challenges. Uh so first is LLMs are not yet good enough to just give us an answer that we inherently believe. Uh if you use a natural language prompt, get an answer from one of the many AI analytics tools.
Let's say the answer is 42, you don't know if it's accurate, right? AI often often makes mistakes and so you need some way to audit that response. And the majority of AI analytics tools today are performing analysis in Python or SQL. They're like writing code to interrogate and analyze data and then spitting out a response. Sometimes some visuals, graphics, pivot tables, uh subsets of data. The problem is the business people we've been talking about do not know Python or SQL.
So they have no way to look at this code and assess whether the result is accurate. Uh and so so like what's the solution there? Uh the spreadsheet is the UI that business people can audit and can augment. And so if an AI agent produced a spreadsheet with pivot tables, cell references, formulas, graphs, uh any average business person can take a look at that data, take a look at the graphs and assess whether they think the analysis is valid.
If not, you can try again. The other issue is performance. Uh so context windows for LLMs are not big enough to feed in entire data sets. Uh, and even if you could, it's not clear that feeding an entire data set to an agent would make the analysis any better. Really, what you need is to provide LLM with tools to interrogate and analyze data. And currently, as I mentioned, they're using code to do that. But another option you could imagine is a spreadsheet API where an LLM can now create graphs, can create pivot tables, can write formulas, um you know build an analysis and probe data with spreadsheet functions and spreadsheet formulas and that's where we are headed.
Um so even in the age of AI, I think spreadsheets will continue to exist. At row zero, we've built a spreadsheet that runs in the cloud. Uh so it connects directly to data bricks. It can open billion row data sets and we provide enterprisegrade security. You can limit data export. You can limit sharing. Uh you have full traceability and versioning. Data teams love us for the built-in connectors. Uh like there's no easier way to connect your business partners to your cloud data warehouse. uh you have shared data sources where you can share a query with folks who don't know SQL and allow them to run it using their own credentials in data bricks.
This will enforce RLS and arbback uh and so you don't have to worry about security issues and people exporting uh data to their local machines. Business teams love it because it works just like Excel. There's no learning curve. Uh every BI tool talks about self-s serve analytics but for a spreadsheet there is truly no learning curve. All the formulas are exactly the same. It's a thousand times more powerful. So, give us your biggest data sets.
Uh, and it's connected and dynamic. No more building one-off analyses that you need to refresh every week or every month. When you build an analysis off of a connected data set, every time that data updates in data bricks, your spreadsheet refreshes, so your analysis is live uh and current. Uh, AI is on the road map. Uh, so you can expect some exciting things from us over the next few months. If anybody would like to talk about spreadsheets, we're at booth 308 and I'll be available uh after this talk.

Learn more about the Databricks Data and AI platform.