There is a question circulating in boardrooms and data leadership meetings right now that goes something like this: "We already have BI. We already have a database solution. So why do we need something different?"

It is a fair question. And it’s one that Ari Kaplan has fielded across five continents. As Databricks' Global Head of Evangelism, Ari has spent his career at the intersection of data, technology, and business transformation, including a tenure as President of the Independent Oracle Users Group, an organization representing more than 22,000 database professionals. He has seen firsthand what works, what ages poorly, and what the next decade demands.

In this conversation, we get into conversational analytics, the architecture behind Genie and Lakebase, what it actually takes to earn executive trust in AI-driven insights, and why the companies failing to act may find themselves further behind than they think.

Insight Without Action Is Just Trivia
Catherine Brown: Ari, executives love the idea of talking to their data, but where does the breakdown occur with conversational analytics?

Ari Kaplan: Absolutely! The bottom line is that companies want to talk to their data, governed within the context of their own business. But here is where it falls apart: if all you are getting is facts — trivia, essentially — and there is no path to action, then what is the point? The value of any insight is only as real as what happens because of it.

That is the difference Databricks is trying to make. It is not just conversation, but it leads somewhere. I was talking with Etihad Airlines, one of our major Genie users. Their finance team was asking questions like: "If oil prices go up next quarter, how do we adjust our pricing so we still make money and customers still want to fly?" That kind of analysis used to take months, but with Genie and predictive analytics, they are running those scenarios in real time and actually making the call to change pricing and routing almost immediately. That's a great example of a business being run on its data, not just informed by it.

Catherine: So, when does a natural language answer need to become something more, like a workflow, a transaction, or an actual decision?
Ari: It should happen immediately if you want it to matter. A report that sits there and nobody acts on is just noise. The goal is to get from insight to action without creating something brittle.

You don't want a system where someone just takes an AI answer and runs with it blindly. You still need a human in the loop. What you do want is agility: data that keeps refreshing, insights that stay current, and people in the organization who have the oversight to make sure it all stays connected to how the business actually runs. The humans are still the interface. That doesn't go away.

What Genie and Lakebase Actually Solve

Catherine: What does Genie solve for the business user that dashboards and BI copilots still struggle with?
Ari: Okay, Genie goes beyond traditional business intelligence tools by letting you simply talk with your own data in the context of your business. It lets users self-serve their own AI-driven insights.

Think about what a dashboard actually is. It is a fixed question, asked over and over again, just with new data plugged in. Companies have been doing that for nearly 30 years. It works. But it does have a ceiling. And that ceiling is: the question never changes.

Genie blows past that ceiling because now a non-technical person, an executive, a sales leader, a finance manager, can just ask what they actually want to know in plain language and get an answer. There's no ticket to the BI team and no waiting a week for a dashboard that is almost right but still needs tweaking. It's worth noting that the BI team does not disappear. They get to focus on the genuinely complex stuff instead of fielding requests for one-off reports. In the end, it is a better situation for everyone.

And “Genie Code” is, for more technical users, enabling data scientists, data engineers and others to be so much more effective and efficient in their jobs and the tasks they need to accomplish.

Catherine: How do Genie and Lakebase fit together? Because from the outside, they can sound like two separate products that may or may not interact.

Ari: Genie and Lakebase really are two halves of the same idea. Lakebase is one key area where your transactional data can live. It is the modern database built to handle billions or even trillions of records. Lakebase is just one of the many sources that Genie connects with. Lakehouse (DBSQL) is another, as is anything federated through Unity Catalog: slack, SAP, Google Drive, Sharepoint, etc.

Genie is how you talk to all your data, from Lakebase to your warehouse data in Lakehouse and beyond. So one stores it, that's Lakebase. And one surfaces it, that's Genie. And the whole thing runs through Unity Catalog to manage and maintain governance controls. So the right people only see what they are supposed to see, and everyone is working from the same definitions. Across every team and every business user, the definitions are the same: customer means customer and profit means profit. That shared language sounds simple, but it is actually what makes the whole thing trustworthy.

The Case for Moving Off Legacy Databases

Catherine: Here's a fun question: A CIO says I already have a database. Make the case for why that is not enough anymore.

Ari: I love this one because I lived on the other side of it. I can tell you that the foundational architecture behind traditional databases genuinely has not changed in decades. Those foundations were built for a different era.

Here is what has changed. With Lakebase, you can provision a new database instantly — not in days or weeks. And most databases created today are provisioned by AI.
The second thing is cost. Traditionally, if you needed a production environment, a test environment, and a QA environment, you were making three copies of your data. Triple the storage, triple the cost. With Lakebase, you can fork off a hundred environments without making a single copy of the underlying data. I was just in Brazil, and companies were telling me they cut their total cost of ownership by 40% after the move. The highest I have heard is 98%. Not everyone gets there, but the savings are real wherever I go.

And then scale. Arctic Wolf, the largest network operations center in the world, manages over one trillion records on Databricks every single day. The old assumptions just do not hold anymore.

Catherine: What is the real difference between asking a question of your data and running your business on it?
Ari: The easiest way I can put it: one tells you what happened, the other helps you decide what to do next.

An international airline I work with had 80 vendors — catering, security, electronics, all of it. They asked Genie: Which of our suppliers is overcharging us? Rank them. Something like that would have taken a team of analysts months. Genie answered it. Turned out their orange juice supplier was the biggest overcharge. Now they know exactly where to go to renegotiate. That is running a business on data.

Supercell — they make games with hundreds of millions of monthly players — uses Lakebase for real-time matchmaking, toxicity controls and in-game purchase decisions. All of that is happening live, at scale, because the data infrastructure can keep up.

And iFood in Brazil, which handles more than 90% of the country's food delivery, uses Lakebase to route motorcyclists through dense cities in real time. These are not companies that are thinking about using their data better someday. They are already doing it.

Governance Makes Analytics Trustworthy

Catherine: Why do semantics and business definitions matter for executives to trust conversational analytics?

Ari: More than most people expect, honestly. Here is a simple example. You ask three executives at the same company to define profit. You will get three different answers. Same with customer churn, same with customer satisfaction. If your analytics system does not know whose definition to use, you are going to get answers that cause arguments instead of alignment.

Databricks gives executives the ability to define those terms themselves without any software development required. An executive can say: to me, a sale does not count until the 30-day return window has closed. That definition is now baked in for everyone using that Genie space. You can even upload your HR manual or operations manual, and Unity Catalog will use that to understand your company's language better.

FordDirect did something like this across their global dealership network . In thousands of locations, very non-technical users, car salespeople, received daily operational reports through Genie. Reports that included which customers are coming in, which vehicles are arriving, and which cars are being recalled. They ran a satisfaction survey, and it had a 95% approval. For a non-technical audience at that scale, that number is almost unheard of.

Catherine: What goes wrong when conversational analytics is not connected to governed operational systems?
Ari: You get drift, and you lose trust, and once you lose trust, it is very hard to get back. We surveyed more than 20,000 customers, and the single biggest barrier to AI and data adoption was a lack of trust. Not cost, not complexity — trust. And that distrust comes from systems that are not governed, where different teams are defining things differently, and the answers start to diverge from reality. Hallucinations are the dramatic version of this, but the quieter version — slightly outdated data, slightly inconsistent definitions — is actually more dangerous because it is harder to catch.

Catherine: For leaders who say natural language analytics is too risky because terminology is always changing, what needs to be true before they should move forward?
Ari: They are not wrong, the risk is real. Without guardrails, general-purpose chatbots are genuinely risky because they do not know if the data they are drawing from is from last week or five years ago. And, they do not understand your business context. The answer is not to wait, though. The answer is to be deliberate about how you roll it out.

Fox Sports is the example I keep coming back to. Their reputation is entirely built on accuracy. People are making decisions, sometimes financial ones involving wagers and bets, based on their content. They implemented a Databricks-powered public chatbot. But they were careful about deploying it. They pre-defined what could and could not be asked. They had humans shape how answers were framed. And they built in controls to prevent hallucination. Today, 25% of all questions on Fox Sports run through that environment. All because they got the foundation right first.

The Competitive Gap Is Already Opening

Catherine: For the leader who thinks Genie is just text-to-SQL with a better story, what are they missing?
Ari: A lot, to be honest. The fundamental thing they are missing is that Genie is built on their data, with their terminology, and governed by their business rules. A generic chatbot has been trained on the internet — Taylor Swift lyrics, historical trivia, whatever. Genie knows what your company means when it uses the terms customer and it knows your fiscal year. It knows your product catalog. That granular specificity is important. And Genie goes way beyond question and answer. Genie does open-ended research. It makes predictions. It connects to your adjacent systems like Salesforce, Workday, SAP, Slack, so you can ask questions that span your entire business. I have seen executives use it to ask genuinely open questions like, "What should I be focusing on today?" And those executives walk away with two ideas they had never considered. That is not text-to-SQL. That is something very different.

Catherine: Last question — 18 months, what separates the companies that are operationalizing a modern data stack from the ones still running demos?

Ari: I was actually part of the backstory behind the movie “Moneyball”, which centered around how one industry (Major League Baseball) had to change. The teams that did not adapt got left behind dramatically and quickly. The phrase from that story was "adapt or die." It sounds extreme, but I keep coming back to it.

I have been to Brazil, Australia, India, across Europe and Asia in the last few months alone. Everywhere I go, companies are moving fast. The ones that are serious about this right now are pulling ahead. And the gap is not going to stay small — it is going to compound. The value is real. Companies are getting better insights, yes, but they are also automating work that used to eat up entire teams, transforming supply chains and accelerating software development. All of it is available right now. The executives who treat this as something to evaluate later are going to find that later arrived sooner than they expected.

The Data Stack Has Changed. The Question Is Whether You Have.

Here is the thing about Ari's argument: it is not really about products. It is about a shift in what is possible and how fast it is becoming the baseline expectation.
Genie and Lakebase are not a BI tool and a database that happen to live on the same platform. They are a single answer to a question every data leader is being asked right now: how do we make our data actually work for the people running this business? One stores and governs it at a scale legacy systems were never designed for. The other puts it in plain language in front of the people who need it most, with the context and guardrails to make it trustworthy.

The organizations getting this right are not just running better reports. They are making faster decisions, catching problems earlier, and finding opportunities that their competitors are still building spreadsheets to find.

Explore what Genie and Lakebase can do for your organization.