Vibe coding needs context and control to work in the enterprise

The rise of coding agents has been hard to miss. In the last six months alone, Databricks Apps has seen major growth in the number of customers building apps, leading to a near doubling of the number of active running apps and more than 3x the number of users interacting with apps every week.

Anyone with a clear idea, whether it be a business analyst, a domain expert, or an operations lead, can describe what they need in plain language and get a functional application in hours — and this is transforming how our customers interact with their data.

But in the enterprise, teams need more than a fast starting point to capture value. Most vibe coding tools are optimized for speed and ease. But without the context to build something grounded in real business data–or the control to make sure what gets built is safe to deploy and cost-effective—apps cannot actually make a widespread impact in an organization. 

At Data + AI Summit 2026, we talked about how Databricks is bringing apps vibe coding to the enterprise with three new key capabilities for Databricks Apps: 

• App Spaces, a new governance boundary to configure resources, access management, and security policies for groups of apps.
• Genie App Builder, a purpose-built AI app authoring tool for Databricks with native awareness of your data assets, Unity Catalog semantics and workspace context.
• A new class of Serverless Micro Apps that start up fast when needed and scale down to zero when idle, so organizations can support a broad portfolio of apps without the cost of always-on infrastructure. 

App Spaces: a governance-first approach

Our goal is to enable any user in the organization to safely create and deploy apps on Databricks.

But this means that Databricks admins need to ensure that developers–who may not be application security experts–are able to work securely with data.  As more people build more apps, governance cannot happen one application at a time. Per-app configuration does not scale and creates bottlenecks for platform teams.

App Spaces is designed to solve that. Admins define resource and data access, API scopes for “on-behalf-of-user” access, and security policies at the App Space level. Every app in that space inherits those settings automatically. The guardrails now exist before the first app is created. 

This shifts governance from reactive to systematic. Instead of reviewing each new app as a one-off, platform teams can create pre-approved environments for builders to work within. Builders get more autonomy and admins get consistent visibility–what apps exist, who owns them, how they are being used–across the entire portfolio.

Genie App Builder: AI-assisted creation with all Databricks context 

Genie App Builder gives teams–both technical and less technical–an AI-assisted path from a plain-language description to a working internal app. Users describe what they want to build or provide a screenshot for context, review a generated plan and iterate from there with a live preview of the app updating in a side pane as they go. 

Under the hood, apps are built on AppKit–a TypeScript SDK designed specifically for production-ready Databricks applications. AppKit takes care of built-in caching, telemetry, retry logic and seamless integration with the data and resources in Databricks out of the box.

Because Genie App Builder is built on Databricks, it has direct awareness of your workspace: the tables you have, the semantic layer you have defined in Unity Catalog and the governance policies already in place, so when you describe what you want in a prompt, the builder agent can find the right data and surface it in the app without you having to wire anything up manually.

Serverless micro apps: infrastructure built for scale-to-zero apps 

The economics of traditional app infrastructure are less favorable for smaller, more specialized apps. Many of these types of apps–like departmental tools, and line-of-business workflows–are used in bursts and idle most of the time. Using always-on, dedicated compute means paying for capacity that is often not in use. The alternatives, though, have historically come with their own tradeoffs: slow cold starts, shared environments without isolation, or apps that never get prioritized because the infrastructure cost is hard to justify.

Serverless micro apps are built on a new micro VM-based runtime that resolves this. Each app runs in its own lightweight virtual machine, which means it can start up fast when needed, scale all the way down when idle and stay isolated from other apps. The practical outcome is that deploying a lightweight app on Databricks becomes usage-based, not reserved capacity-based. This makes it much more viable to create entire classes of apps that would otherwise get deprioritized due to cost concerns. 

Summary