Count Agent Demo

Hi. I'm Josh. And today, I'll be showing you Count, the collaborative agentic analytics platform your whole team can work with. I'm going to reveal how quickly you can go from question to real usable insight and report. Not a silo dashboard you'll forget about, not a black boxed text only chatbot response, an actual analysis generated using an intelligent compute layer that keeps costs down that you and your team can review, refine, and act on. In count, you can connect just about any data source. It could be your governed warehouse, your semantic layer, a dbt model, a Google sheet, a CSV, really whatever you're working with. In an empty canvas, you can prompt, connect, or begin to build out your analysis however you like. I'll just add a CSV and place a sticky down to get started. You can also access the agent from the toolbar at any time. Click anywhere in the canvas to place an agent and ask a question. Behind the scenes, the agent is writing code, pulling data, and building out the analysis, but you get to see all of it. As it reasons, thoughts and visualizations update in real time. Count shows its work, so you can have something auditable, editable, and interactive. Now let's say my team and I want to iterate on the result. We can use Count's suite of Canvas tools to advance the analysis and ideate using code cells, visuals and tables, data controls, stickies, text, shapes, rich media, and comments. I can also just copy and paste or option click and drag any part of the analysis to explore findings further. Anything the agent produces, you can take and make your own. Within the visualization editor, I'll adjust to investigate how location relates to churn risk. I have another file that contains customer location data. Let's start by adding that to the canvas. Then let's join the two datasets. From there, I'll select a new chart type, adjust my levels and weights, and there we go. At any point, a teammate can jump in, add context, leave a note, and build on what the agent and I are creating. Oh, there's my colleague Amy. Hi Amy. While she's working on that, let's send a follow-up to the agent. I can also add additional context from anywhere on the canvas. I can grab a chart, a sticky, a table, nearly anything. And now they're part of the conversation. This kind of iterative back and forth is where real insight happens. And because Count's compute layer distributes queries intelligently across your warehouse, our servers, and your browser, the agent can run hundreds of queries in seconds without ballooning your data costs and maintaining depth and accuracy at scale. When I've got something worth sharing, I can create a presentation using slides and templates. Let's borrow one of Count's ready made templates. Let's drag this chart and this table. Click present, and there we go. Using any cell on the canvas, you can easily build out presentations, dashboards, and reports, and present them all without leaving count. Everything here is governed. As mentioned, count connects to your existing metric definitions, whether that's count metrics or dbt. You can even version control through GitHub, so changes are tracked and nothing breaks downstream. And our governance layering gives you the permissions and security of an enterprise BI tool. Teams from Bumble, Substack, Intruder, and more are already using count to answer the questions that matter most. Diagnose what's driving metric changes. Run cohort analyses to understand user behavior. Flag anomalies across geographies and products. Generate executive summaries that surface only what's new and significant. Even onboard to unfamiliar datasets by having the agent surface structure, key metrics, and quality data issues. So that's Count. One canvas where agents and people explore data collaboratively and transparently. A compute layer built for the speed and scale agents need, and the enterprise grade governance, security, and control your business requires. It's the fastest way to go from question to insight to decision, and we can't wait to see what you build.