Transcript for "Agentic Analytics Summit Keynote":
Welcome, everyone. Really glad you are here. This is our second agentic analytics summit. Thank you to everyone who joined us last year. A lot has happened since then. The industry has been moving toward agentic analytics faster than I expected. And, honestly, it's exciting to watch. Today, we have a lot to share. I want to start with a question that we have been asked a lot over the past year. What actually is agentic analytics? Is it just a chatbot on top of your dashboard? Is it AI writing SQL? Here's how I think about it. Every time there has been a fundamental shift in technology, not an incremental improvement, a real platform shift. Software has been rewritten from scratch. The tools that won the previous age rarely win the next one. Business intelligence, and end of the day, is just a software. So when the underlying technology shifts, business intelligence shifts with it. We have already seen this happen at least three times. The first generation of BI business objects, Cognos, Hyperion was built for largely pre Internet world. IT runs slow to operate and slow to change. Then Internet happened. New tools arrived like Tableau and Qlik that enabled people to move through analytics faster. And then finally, we started to see the rise of warehouses such as Snowflake, BigQuery, Databricks, and Redshift. They paved a way for a new set of tools like Looker, Sigma, Metabase. These tools were built for direct queries. No extracts enabled by cloud data warehouses architecture. Now we're in the fourth shift, AI native. That's what we're building with Qiskube. There is a narrative going around right now that AI is killing software. Everyone can write code. Everyone can build tools. So why pay for software? Why buy BI platform if you can wipe code your dashboard in an afternoon? I think that's a wrong way to look at it. Yes. The simple internal tools used to build on platforms like retool, you can build themselves using wipe code. That's real. But at the same time, software engineering teams, they can now move at a way faster pace and build more complex and scalable solutions. So it means that bar for professional software is going up again. More features, more complexity. So there are two things that happening at once at the same time. So the floor is rising, but the ceiling is rising as well. And the gap is actually doesn't close. It just moves up. And the next generation of BI is going to be exponentially more complex than what we see now. And look. I want to be fair here. There are good use cases for wipe coating a dashboard. If you're one person company or just starting a new product and prototyping, you don't need a full blown BI platform. That's going to be overkill. You just should wipe code your dashboard and use it to track analytics. But the moment your company starts to grow, the moment you have multiple teams looking at the data, then when you need to start thinking about the data platform. The dashboard is not enough anymore. You need a complex system that has governance, security, control, and all other enterprise features. And nobody wants to wipe code, row level security, or maintain multi tenant permission model and deal with complexity of semantic modeling and make sure that the metrics stay consistent. For a while, a whole conversation around AI and analytics was dominated by one idea, text to SQL. Ask a question in plain English, get a SQL query back. But text to SQL was never the real problem. The hard part was never the syntax. The real question, can people do things now that they couldn't do before? Are they getting answers faster? In the scope of what is possible actually expanding, That's bar for meaningful innovation, and that's what agentic analytics has to be. So what is agentic analytics actually? For us, it's two things. First, it lets people do things they couldn't do before. A business user getting a report, a dashboard without filling a ticket with engineering. An analyst exploring data without writing simple from scratch. Second, it makes the people who are already good at this faster and less frustrated. Data engineers can move faster and build semantic model faster. But neither of those things works if the pieces aren't connected. A great AI chart on top of a separate dashboarding tool, on top of a separate exploration layer, that's just more duct tape. You need one platform where all of it shares the same understanding of what the data means. That's the context layer. And the backbone of the context layer is the semantic model. Self serve is getting better. More people are building more dashboards. That's great. But what we are seeing is that as the volume of analytics works goes up, more dashboards, more builders, more questions, the inconsistency compounds. Three teams, three definitions of active user. The AI answers fast, but AI may produce a lot of wrong answers at a pretty fast scale. Adjenic analytics needs something solid underneath it, a context layer with a real semantic model at the core. Without it, we're just creating more wrong analytics faster. So the semantic layer is the foundation. We just established that. But here are the things. Semantic layer isn't a new idea. It's been around for a long time, and it's always come with a problem. There is a tension that existed in BI forever. On one side, governance, consistent definitions, one version of truth. On the other, flexibility, business users exploring freely, analysts creating their own metrics on the fly. Every BI tool picks a site or tries to find a middle ground and usually does neither well. Lock it down and nobody uses it because it's too slow. Open it up, and six months is later, you have 13 definitions of churn. The industry has been going back and forth on this for a long time. This is the problem we set out to solve. We built our semantic layer and the way to query it specifically to address this tension, governance and flexibility at the same time, not one or the other. We build it SQL first. That means it is extensible at query time. You're not locked into what the data team predefined. And AI is pretty good at SQL. So AI can construct ad hoc calculations on top of the governed metrics that already exist. The data team's definitions stay intact, and the user gets the specific answers they need without waiting on anyone. We released our GenTIC analytics product in GA on October. Since then, we have been busy improving what we had and building new features on top of it. I'm excited to show you the progress. Since October 29, AI chat usage on Qube has grown significantly. This chart shows that growth. But what's more interesting than the number itself is what we are seeing in how people use it. Power users moving through analytics faster. Business users getting reports and dashboards they couldn't get before. The patterns we talked about earlier, that's what this chart reflects. We now have over 400 customers building on QoNTube across industries, company sizes, and use cases, household brands, growth unicorns, and enterprise leaders. And a lot of what we have been building in the last six months has has been shaped by what these customers are asking for. Today, I'm excited to share two major launches, embedded agentic analytics and cube into flow. First launch, embedded agentic analytics. Embedded analytics is a well established category, but what we are introducing isn't just embedded analytics with AI on top of it. It's built AI first from ground up. Embedded analytics isn't new for us. Since our earliest days as a headless BI platform, Kyiv has been deeply committed to this use case. Many of our customers build sophisticated custom embedded analytics experiences directly on top of our core data APIs. Full control, full flexibility, their brand, their user experience, that commitment has never wavered. What's changed is the world around us. Analytics is becoming agentic, and our customers who build embedded analytics on Qube are now asking the same question. How do we make our embedded analytics experiences AI first? We heard that and we double downed. Embedded agentic analytics is our answer. Here's what we are seeing in the market. Software companies that invested in embedded analytics a few years ago are now facing a new expectation from their customers. They want AI first experiences and analytics layers that understand their data and can answer real questions. The demand is real, and it's accelerating. We built embedded agentic analytics for this. Whether you are upgrading an existing embedded experience or building from scratch, we have a pass for you. We build a complete infrastructure for embedded agentic analytics, and what makes it different is that customers are never locked into one approach. You choose the level of customization that makes sense for your product and your time line. And as your product grows, you can move between these options without starting over. Four options. First, the analytics chat API, a streaming API to build a fully custom analytics chat experience. High control, high customization, and it works in agent to agent communication So your eye AI can talk to hours. Second, analytics chart and dashboard iframes, the fastest path to embed and crop in a chat interface or a dashboard in your product with minimal engineering. Third, creator mode. This is the most powerful options. Embed full workbook and dashboard creation directly inside your application. Your customers can build and share their own dashboards inside your product, an entire agentic BI experience in your application. And finally, our core data APIs for teams that want maximum control at their data layer. Everything we have always offered is still there. The best way to understand what embedded agentic analytics looks like in production is Brex. They will be speaking later today, so you will hear it directly from them. But here's the story. Brex built spaces, an AI operated financial reporting workspace embedded inside their product. A finance team asks, why is marketing 23% over budget? And gets an answer in seconds. Before writing a line of code, they have evaluated CUBE, DBT, semantic layer, and LookML to check what put the best solution to be AI ready. They chose CUBE for many reasons, including our ability to scale to many customers they have and provide a production scale caching capabilities. So that's embedded agentic analytics from the iframes and APIs all the way to full creator mode experience with Brax as a real customer example of what it looks in a production. Now let's talk about the second launch, cube in the flow. You shouldn't have to leave where you're working to get an analytics answer. Right now, most tools make you do exactly that. Open a new tab, find the dashboard, remember where things live. Your data should be accessible where you already are. Two integrations launching today. First, Slack. You can now connect to cube directly from Slack and ask analytics questions just like you would in cube chat. The answers come back in Slack with your permissions and data access fully synchronized. The same governed experience you would get inside the product now straight in Slack. Second, our remote MCP server. MCP is a protocol that lets AI agents connect to external tools and data sources. We've built a CUBE MCP server so you can connect Claude, Charge GPT, or any MCP compatible AI agent directly to CUBE. Authorize once and ask your data questions from inside whatever AI tool you're already using. Your AI assistant talks to Qube, gets a governed answer back, and then surface data back however you want. Slack and MCP are the first two, and we have more integrations coming. The goal is simple. Cube where you already work. So that's where we are. A semantic layer that works for AI, not just for humans, embedded agentic analytics for software companies who want to give their customers something better and keeping the flow showing up in Slack, in your AI tools, wherever you actually work. None of this is just a road map. It's already in production. We are a small team, and we have been moving fast. And, honestly, I'm pretty proud of what come out in this last six months. We think we're building the right thing, and I'm glad you're here building with us. That's a keynote. We have a full day of sessions ahead and a hands on workshop later today. I hope you stick around for all of it. Thanks for being here. Enjoy the rest of the summit.