<span>Atlan AI Search: <span class="text-primary">Data Discovery</span> with AI & Metadata</span>

Prukalpa Sankar

0:00 So if you remember the three-tier stack that I talked about with software, not just the data and the logic are changing, but the experience is changing as well. So it's moving from navigation, where you're moving a mouse, it's moving to conversational. It's moving from clicks to intent, and most importantly, which is what I'm really excited about, it's moving from output to outcomes. So we are reimagining the data discovery as well within Atlan for, for the AI world, and how do we bring Atlan's rich context to power the AI search. Tom, passing over back to you.

Tom

0:37 All righty. Perfect. All right. Finding the right data is just too hard. You don't know which BI tool your company uses. Is it Tableau, Power BI, Looker? You don't know which databases even exist or whether your metrics are defined consistently. And by time, by the time you've asked the question in Slack, checked Confluence, sent a few emails, you've already lost days. Now imagine I'm a brand-new business user. Leadership has asked me to deliver metrics on customer acquisition cost, CAC. I don't even know where to begin. Instead of wasting time che-chasing context, I'll start with Atlan's AI search. Let's hop into Atlan and check it out.

1:27 So I ask, "What metrics are used to evaluate customer acquisition cost?" In seconds, Atlan surfaces glossary terms like CAC itself. Suddenly, I know my company actually measures customer... How my cust— my company actually measures customer acquisition cost. But definitions alone aren't enough. I need to know who owns it. So I ask, "Which teams are responsible for tracking and optimizing CAC?" Just like that, I see Razi owns it, and I know exactly who to reach out to with any questions. No guessing, no endless Slack messages. Last but not least, I want to see this thing in action. So I ask, "Do we have any dashboards available for tracking CAC trends?"

2:12 And instantly, I see the live dashboards my company already uses to track performance. By clicking on a dashboard name, the companion sidebar pops out to give me quick context. Diving into lineage also helps provide more signals as to whether this is even the right dashboard. Finally, maybe I want to preview the dashboard right in Atlan and get confirmation it's exactly what I need. That's the power of Atlan's AI chat. It removes guesswork, the endless hunting across tools, the wasted hour in Slack, turning what used to take days into minutes. And look, here's the thing. We actually built this two years ago before GPT was mainstream. Some of you may remember that.

2:58 We built a lightweight LLM wrapper, but we realized quickly the problem was bigger than just dropping a model on top of metadata. The answers were not accurate enough. We made the tough decision to roll it back. And here's the truth. Anyone who tells you their AI search is one hundred percent accurate is lying. The real challenge is grounding AI in reliable context. That's why we built a Metadata Lakehouse on Iceberg and paired it with semantic search plus Elasticsearch. Together, they reconcile intent with precision. So AI doesn't just sound smart, it's actually useful. It retrieves the right answers consistently at scale. That's what makes AI different.