\n In the opening keynote, VP of Azure Cosmos DB Kirill Gavrylyuk described three key shifts driving this transformation, and we saw them play out across every customer story at the event.\n<\/p>\n
The three AI shifts reshaping application architecture with Azure Cosmos DB<\/p>\n
AI is making flexible, semi-structured data foundational<\/p>\n
\n AI applications don\u2019t operate on rigid schemas. They operate on prompts, memory, and context, all of which are inherently semi-structured and evolving over time.\n<\/p>\n
\n This fundamentally changes how databases must behave.\n<\/p>\n
\n Data platforms are no longer just systems of record\u2014they are becoming systems of reasoning, where flexibility is critical to how applications learn, adapt, and generate outcomes.\n<\/p>\n
AI is dramatically accelerating the pace of development<\/p>\n
\n AI, and especially coding agents, are changing how software is built.\n<\/p>\n
\n Developers are:\n<\/p>\n
Iterating faster <\/p>\n
Shipping more frequently <\/p>\n
Scaling from zero to massive usage instantly <\/p>\n
As Kirill highlighted, developers can no longer be constrained by strict schemas. Flexibility isn\u2019t just a convenience\u2014it\u2019s what enables teams to move at AI speed. Databases need to meet the demand with serverless form factor<\/a>, instant and limitless scalability, advanced integrated caching, and provide agent-friendly interfaces. <\/p>\n Semantic search is becoming a first-class query operator<\/p>\n \n The third shift is just as important:\n<\/p>\n \n AI applications require:\n<\/p>\n Vector search <\/p>\n Full-text search <\/p>\n Hybrid search <\/p>\n Semantic ranking <\/p>\n These are no longer \u201cadd-ons.\u201d They are core to how modern applications function. <\/p>\n \n Across Cosmos DB Conf, we saw a clear pattern: teams are building applications where retrieval, reasoning, and real-time context are tightly integrated.\n<\/p>\n OpenAI: Flexibility at planet scale<\/p>\n \n These shifts are most visible in what organizations like OpenAI are building.\n<\/p>\n \n Speaking at Cosmos Conf, Jon Lee of OpenAI addressed how they are operating at massive scale\u2014processing trillions of transactions and petabytes of data\u2014reinforcing that what matters most is not just scale, but the ability to evolve quickly.\n<\/p>\n \n As Jon shared, modern systems must be able to:\n<\/p>\n Scale instantly from zero to massive usage. <\/p>\n Support schema-less design for rapid onboarding. <\/p>\n Enable thousands of developers to iterate simultaneously. <\/p>\n \u201cThe most important thing\u2026 is being able to scale from zero to millions of QPS, being able to scale from zero bytes to petabytes,\u201d explained Jon, adding that speed and flexibility go together. <\/p>\n We have thousands of developers that are actively building products\u2026 it\u2019s really important to make it easy to onboard to databases really fast.<\/p>\n This is exactly the world Kirill described: AI systems demand flexible data models that evolve as fast as the applications themselves.<\/p>\n This highlights how Azure Cosmos DB supports dynamically evolving, large-scale AI workloads<\/a>.<\/p>\n Vercel: The rise of serverless, AI-native applications<\/p>\n \n If OpenAI shows what\u2019s possible at scale, Vercel shows how the shape of applications is changing.\n<\/p>\n \n As Guillermo Rauch, CEO of Vercel, explained, AI is dramatically expanding who can build software\u2014from millions of developers to potentially billions of creators, many of whom are using agents to generate applications on demand. Kirill underscored this point in his keynote when he stated that more than half of Azure Cosmos DB customers are already using coding agents in their development workflows.\n<\/p>\n \n According to Guillermo, this is driving a structural shift toward:\n<\/p>\n Serverless architectures <\/p>\n Ephemeral applications <\/p>\n Instant scaling from zero to viral <\/p>\n \n Data platforms must keep up. To support this pace, platforms need to provide:\n<\/p>\n Built-in best practices (data modeling, partitioning, and optimization). <\/p>\n Intelligent guidance (agent skills and automation). <\/p>\n Real-time feedback on performance and cost.<\/p>\n \n Speaking on why he turned to Azure Cosmos DB, Guillermo said, \u201cI wanted a system that gave me an economical thinking where the developer writes a query and they understand its cost.\u201d\n<\/p>\n Developers need immediate feedback on the cost of their decisions, making efficiency a built-in design principle, not an afterthought.<\/p>\n This reflects a broader shift toward AI-native apps built on globally distributed, serverless data platforms like Azure Cosmos DB<\/a>.<\/p>\n Walmart: Reliability and performance at scale<\/p>\n \n While AI is transforming how applications are built, one thing hasn\u2019t changed: Performance and reliability remain mission-critical.\n<\/p>\n As Kirill emphasized, AI does not remove the need for reliability, security, and performance. <\/p>\n \n In fact, it raises the bar. This was reinforced in sessions like Walmart\u2019s, where Technical Fellow Sid Anand explained that large-scale applications must:\n<\/p>\n Deliver low-latency experiences globally. <\/p>\n Remain available through regional failures. <\/p>\n Maintain consistent performance at massive scale. <\/p>\n \n \u201cWe want people to be able to add to their cart and view cart no matter what is happening in a given cloud region\u2026and we need all of these interactions to be low latency because any type of latency friction will cause a drop-off,\u201d said Sid.\n<\/p>\n From gigabytes to petabytes, from hundreds to trillions of transactions, modern systems must operate seamlessly under unpredictable demand.<\/p>\n These requirements align with how Azure Cosmos DB is designed for global distribution and low latency at scale<\/a>.<\/p>\n Cost efficiency becomes a core design principle<\/p>\n \n A final takeaway from Cosmos Conf: as systems grow more complex, cost becomes just as important as scale.\n<\/p>\n \n Across the keynote and sessions, we saw a clear shift:\n<\/p>\n Developers need cost visibility in real time. <\/p>\n Architects need to design for efficiency upfront. <\/p>\n Teams want to consolidate platforms and reduce complexity. <\/p>\n This is where innovations like Azure DocumentDB<\/a> come into focus. <\/p>\n As highlighted in the keynote, Azure DocumentDB offers over 40% lower cost vs. alternatives<\/a>, and enables high performance with simplified architecture. It also supports open-source, multi-cloud portability scenarios. The result is a broader choice for builders: <\/p>\n Design and architecture examples that developers can start building now<\/p>\n \n Beyond the keynote, there were a number of demo-driven sessions at Cosmos Conf across app architectures, repeatable patterns, and best practices for building and scaling AI-enabled solutions.\n<\/p>\n For example, Farah Abdou, a lead machine engineer at startup SmartServe, shared how her team rebuilt their architecture<\/a> using Azure Cosmos DB as a unified \u201cagent memory fabric.\u201d By combining vector search for semantic caching, change feed for event-driven coordination, and optimistic concurrency for conflict prevention, they were able to reduce costs, enable sub-100ms agent handoffs, and eliminate state conflicts.<\/p>\n Another topic we get asked about a lot is how users protect and govern their AI applications. Pamela Fox, a Microsoft Principal Cloud Advocate, walked through how to build secure, multi-user AI systems<\/a> using the Model Context Protocol (MCP). By authenticating users with Entra ID and storing per-user data in Azure Cosmos DB, she enabled role-based access with Microsoft Graph<\/a>, and practical development workflows using tools like VS Code and GitHub Copilot.<\/p>\n From these hands-on patterns to large-scale production systems, the lesson was consistent: teams are designing for scale, efficiency, and real-world usage from day one.<\/p>\n Key takeaways\u00a0<\/p>\n AI applications require flexible, schema-agnostic data models.\u00a0<\/p>\n Serverless and instant scalability are becoming default expectations.\u00a0<\/p>\n
![\"Azure-DevTools-Dark-2\"](\"https:\/\/www.europesays.com\/ai\/wp-content\/uploads\/2026\/05\/Azure-DevTools-Dark-2-1024x576.jpg\")
\t\t\t<\/p>\n