I had the opportunity to participate in a recent roundtable organized by DBTA. The topic was about the technologies and strategies to get ready for the AI Era.
I was asked a few questions at the end, and I am sharing my answers here:
What is the top data management challenge organizations are facing right now?
The biggest challenge right now is making data usable—fast.
- Companies are collecting more data than ever, from more sources, in all sorts of formats.
- Turning ingested data into something actionable—especially in real time—is where most of them are stuck.
- They’re still fighting with silos, slow pipelines, and systems that just weren’t built for the speed AI demands.
What new technical capabilities do you view as essential to supporting AI workloads in 2025?
There are a few, but the big ones are real-time analytics, hybrid data support, and AI-native features like vector search. AI models need fresh, diverse data to work well. That means.
- Real-time analytics: AI models and dashboards need to be fed with the same data for consistency.
- Hybrid data support: you need infrastructure that can handle everything—structured, semi-structured, unstructured—at speed, and make it all instantly available.
- Vector search: you need to go beyond traditional queries: think similarity search, embeddings, and context-aware analysis.
The database can’t just be a passive warehouse anymore—it has to participate in the AI pipeline.
How do you build for resilience and adaptability in a data ecosystem that needs to keep up with very fast-moving AI trends?
The key is designing for constant change.
- Assume that your data volumes, formats, and even your use cases will evolve quickly.
- Need a stack that’s fault-tolerant, and smart about indexing and scaling.
- At CrateDB, we really focused on that—building a system that can adapt on the fly, without needing downtime or complex reconfiguration. Because if your data infrastructure slows down every time something changes, you’re going to fall behind fast.
How do you ensure that data captured and stored in new systems is truly usable for AI – not just warehoused? Not just another silo?
Great question—because a lot of systems today still treat data as something to store, not something to use.
- To make it usable for AI, you have to break down silos and make data queryable, flexible, and contextual from day one.
- That means supporting different formats, real-time access, and making sure data carries the right metadata and relationships.
- At CrateDB, we combine full-text, time-series, geospatial, and structured data in one place, all accessible via SQL. So the data doesn’t just sit there—it’s ready to fuel models and decisions immediately.
What is the biggest piece of advice you would give data management professionals that are currently dealing with all of these changes, challenges and opportunities?
I’d say: don’t get stuck optimizing for the past.
- The tools and processes that worked five or ten years ago just won’t cut it today.
- Focus on flexibility, speed, and systems that can evolve with you. And remember, the goal isn’t to just manage data—it’s to do something with it. Whether that’s surfacing real-time insights, powering an AI model, or triggering automated decisions—you need infrastructure that helps you act fast, not just store data neatly.
What trend are you most excited about right now?
I’m really excited about how real-time analytics and AI are coming together. We’re seeing a shift where AI isn’t just a backend process anymore—it’s becoming something that reacts and reasons in the moment.
- That puts huge pressure on data systems to deliver low-latency, high-context information instantly.
- It’s not enough to look at what happened last week—AI needs to know what’s happening right now, and why. That’s where platforms like CrateDB are really stepping up.
Access to the full roundtable recording available here
