In February, we launched the Partner Well-Architected Framework (PWAF), moving our partner guidance from static PDFs into AI-ready guidance. For the first time, it spans all three of our core partner architectures: Built-On, Connected, and Data Collaboration. In my PWAF launch session at PKO, I talked about the innovation window: the moment when the pace of change at Databricks and across the market makes a whole new class of products possible. Every capability we ship is an opportunity to build a new, differentiated product, often with a new revenue stream attached.

As you can see from the announcements we're making this week for the Databricks platform and our partner program, this window only keeps getting wider. We built PWAF as AI-enabled guidance using AI tooling. Our goal is to keep pace with what Databricks ships, moving at the speed of the product and the speed of the AI market. So, here's a refresher on what PWAF is, a look at what we've added since February, and a preview of where we're taking it next.

Anchored in Architecture

PWAF starts with its architecture center. It builds on the well-architected principles you already know (the cloud Well-Architected Frameworks and the Databricks Lakehouse Architecture) and focuses them on the patterns our partners actually build with: building your product on Databricks, connecting to your customers' Databricks to run jobs on their behalf, or sharing your data products through the Databricks marketplace. As partners increasingly build data and AI applications, agents, and AI-powered experiences on Databricks, PWAF provides a common set of patterns and standards that help accelerate development while aligning with platform best practices.

We built guidance for all three of these partner architectures: Built-On, Connected, and Data Collaboration. For Built-On partners, it's anchored by Firefly Analytics, our reference implementation. The whole architecture center is AI-ready by design, so you can point your coding agent at it and start building.

Beyond the patterns, the architecture center spells out the technical standards every integration has to meet, set to the same bar our partner engineering team validates against. It also shows you how to instrument your solution so your adoption and DBU impact are measurable, providing the data that can move you up the partner tiers and grow your GTM benefits.

Brick by Brick: What's New Since February

As Stephen put it, we've added a lot of bricks to the wall.

• The Databricks AI Partner Dev Kit. We've packaged 15+ AI-developed skills your coding agent can use, covering integration patterns, telemetry instrumentation, and even prepping for your partner validation call. Every skill is fully tested and ships with its own test suite. Instead of reading a pattern and re-implementing it by hand, you hand the skill to the coding agent of your choice and let it build against vetted, well-architected standards. Back in February, I demoed building a JDBC integration with PWAF; it took around 20 prompts of back-and-forth with a coding tool, using my own Databricks expertise. With the Dev Kit, that same integration came together in a single shot, and partners are telling us they're seeing the same on their own builds. You spend your time on what makes your product different, not on how to install a connector or implement a user-agent.
• New and expanded pattern guidance. Since February, we've published net-new coverage for Clean Rooms, software-defined storage, and Marketplace apps, and refreshed our guidance and standards on the capabilities moving fastest: Genie, Lakebase, and MCP server onboarding. As Databricks launches new capabilities and new patterns and standards emerge we’ll keep adding more guidance to PWAF.
• Firefly is now open source. Firefly Analytics, the reference implementation we built for Built-On partners, is live as a Databricks Labs repo you can clone today. It ships working examples of the hard parts of building an app on Databricks: auth, IAM, and SSO/SPN flows; enterprise-grade security and scale; embedded apps; and AI. Take it as a starting point, extend it, make it your own, and if you find a pattern we're missing, tell us and we'll add it.

Partner Engineering in the AI Era

The way partners and Databricks build together has changed, and it will keep evolving with the AI era. What used to pass between our engineers and yours can now run agent to agent, freeing both teams for the complex architecture problems we can only solve together.

Our partner engineers are builders, and this is how we build alongside you: not with a single skill, but a full suite. Skills handle the routine integration work, architecture guidance tackles the harder design decisions, and reference implementations like Firefly give your agent a working example to point at. Short of sitting at your keyboard, it's the closest thing to having our team build it with you, and our aim is to help every partner move faster.

That's the multiplier effect we're after together. Enabling one partner to build faster is linear; enabling every partner to build deep, differentiated products is how we aim to turn that into exponential growth across our joint customer base.

The Window is Open

The innovation window is open right now, and it rewards partners who build deep and differentiated: using more of our platform to build something your competitors can't match, and that you couldn't build anywhere else.

It's all live today. Point your coding agent at the architecture center, pull in the Dev Kit, and clone Firefly. Tell us what you ship, or what we're missing, through the Partner Portal.

We'll keep shipping new patterns, skills, reference implementations, and demos. Bookmark the architecture center and point your tools at it, because this is an ever-evolving framework, one that moves fast and is only going to move faster. It's early days, but we're excited about what we're building, and that we get to do it with a partner network this strong. Let's build it together, brick by brick.