Each year, the Azure Cosmos DB Conference provides a real-world glimpse into how applications are built at global scale—not just in theory, but in production. This year's event made one thing unmistakably clear: AI is no longer just another workload. It's fundamentally reshaping the way applications and the data platforms that power them are designed.
In the opening keynote, Kirill Gavrylyuk, VP of Azure Cosmos DB, outlined three major shifts driving this transformation. These themes echoed through every customer story and technical session. Let's explore what they mean for developers and organizations building AI-powered apps.
The Three AI Shifts Reshaping Application Architecture
1. Flexible, Semi-Structured Data Becomes Foundational
AI applications don't thrive on rigid schemas. They depend on prompts, memory, and context—all inherently semi-structured and constantly evolving. This demands a fundamental shift in how databases behave. Data platforms are no longer just systems of record; they are becoming systems of reasoning. Flexibility is no longer a nice-to-have—it's essential for applications to learn, adapt, and generate meaningful outcomes.
2. AI Accelerates Development Pace Dramatically
AI—especially coding agents—is changing how software is built. Developers are now:
- Iterating faster than ever before
- Shipping more frequently, with confidence
- Scaling from zero to massive usage in moments
As Kirill emphasized, developers can no longer be constrained by strict schemas. Flexibility enables teams to move at AI speed. Databases must meet this demand with serverless form factors, instant limitless scalability, advanced integrated caching, and agent-friendly interfaces.
3. Semantic Search as a First-Class Query Operator
The third shift is equally critical. Modern AI applications require:
- Vector search for similarity queries
- Full-text search for traditional keyword retrieval
- Hybrid search combining both approaches
- Semantic ranking to deliver the most relevant results
These capabilities are no longer optional add-ons—they are core to how applications function. Across Cosmos DB Conf, a clear pattern emerged: teams are building applications where retrieval, reasoning, and real-time context are tightly integrated.
OpenAI: Flexibility at Planet Scale
The practical impact of these shifts is best illustrated by organizations like OpenAI. Speaking at the conference, Jon Lee of OpenAI shared how they operate at massive scale—processing trillions of transactions and petabytes of data. The lesson? What matters most isn't just scale, but the ability to evolve quickly.
Jon highlighted that modern systems must be able to:
- Scale instantly from zero to massive usage
- Support schema-less design for rapid onboarding
- Enable thousands of developers to iterate simultaneously
He noted, “The most important thing… is being able to scale from zero to millions of queries per second, being able to scale from zero bytes to petabytes.” This philosophy aligns perfectly with the three shifts discussed earlier.
Learn how flexible data design enables rapid onboarding | Explore semantic search integration
What This Means for Developers
Cosmos Conf 2026 made it clear: building AI apps requires a data platform that embraces flexibility, speed, and intelligence. Whether you're a startup launching a new AI feature or an enterprise scaling to billions of requests, the principles remain the same. Azure Cosmos DB is evolving to support these shifts with serverless models, integrated caching, vector capabilities, and more. The future of application development is AI-driven—and the database needs to keep pace.