Databricks Genie: Ask, Build, Compose

Databricks' fifth customer hackathon showcased the evolving capabilities of its Databricks Genie platform, focusing on three distinct user pathways: asking questions, building solutions, and composing intelligent agents. The event underscored how conversational analytics, governed by Unity Catalog, is shifting from a niche feature to a foundational element of data strategy.

StartupHub data

Companies working on this

Profiles of the companies named in this story, with founding year, headquarters, and a short description from our database.

Databricks

A unified data analytics and AI platform built on the lakehouse architecture.

Founded

2013

Location

San Francisco, United States

Valuation

$190.0B

The hackathon structure mirrored Genie's three core functions. First, business users can interact with their data through Genie Agents, chat interfaces curated over specific datasets. Second, analysts and builders leverage Genie Code to automate the creation of data pipelines, metric views, and dashboards. Finally, developers can integrate Genie Agents as tools within larger autonomous agents.

Track 1: Talk to your data with Genie Agents

Genie Agents are designed for business users seeking direct insights from governed data. Analysts curate these agents, defining business vocabulary and pinning trusted functions to ensure accuracy. Users then interact by typing natural language questions, receiving results, charts, and the underlying queries.

One project demonstrated how to overcome the 30-table limit per agent by creating a supervisory layer. This system shards data across multiple focused agents, routes queries intelligently, and consolidates answers, extending self-service analytics across an entire enterprise without compromising governance.

Another team applied Genie Agents to loan records, meticulously defining terms like "cure" and "DNC." This allowed the collections team to query in plain English, uncovering insights such as the average resolution time for delinquent loans. The agent even began suggesting relevant questions the team hadn't considered.

Track 2: Build with Genie Code

Genie Code targets analysts and builders who can write SQL but need assistance with more complex data workflows. It generates metric views, Unity Catalog functions, pipelines, and dashboards based on natural language descriptions, directly within Databricks.

A team used Genie Code to build a governance intelligence platform. This tool identifies dormant reports, clusters duplicate assets using lineage and SQL logic, and scores data readiness for AI applications, a project typically requiring significant time and planning.

Procore developed a comprehensive analytics experience for a vacation-rental platform. Leveraging built-in AI functions, they classified and scored listings, then delivered a dashboard of key performance indicators, trends, and forecasts. A companion Genie Agent answered portfolio managers' questions about optimizing satisfaction.

Fanatics Betting and Gaming used Genie Code to create a customer-experience tool providing managers with ranked, ROI-justified action lists. They also stress-tested their churn model, discovering that two history-based features contained most of the signal, leading them to adopt a simpler, more effective approach. This workflow was packaged as a reusable analyst skill.

Track 3: Compose Genie into agents

This track pushes the boundaries, enabling developers to use Genie Agents as tools within larger autonomous systems. Through Genie Conversation APIs and managed MCP servers, Genie Agents can be invoked by other agents to answer natural language data queries.

ShipBob built an "11 PM Ops Brief" system. This agent queries the warehouse via Genie, fuses data from 17 live public feeds, identifies recurring patterns, and drafts a factual brief detailing revenue at risk and potential actions. This transforms a lengthy operational review into a concise, actionable summary.

Reach Mobile's DBX Lens provides cost and governance insights for Databricks usage. It pairs an embedded Genie Agent with an MCP server, allowing users to ask questions like "show DBUs by SKU over the last 30 days" in plain English. It even translates natural language governance rules into SQL for monitoring.

Kin Insurance developed a growth and marketing agent that researches new markets and runs analyses using Genie. This integrates autonomous planning with data querying, streamlining multi-step research processes into a single request.

Ripple created a KYC briefing agent for regulated finance. Genie provides internal CRM context, while the agent screens against external sanctions and adverse media sources, collapsing hours of manual research into a single, cited brief.

Fanatics Betting and Gaming also built FirstBet Coach, an onboarding guide for new sportsbook customers. It combines a Genie Agent with a custom sports-data MCP server, persistent memory, and MLflow tracing for an audit trail.

The underlying theme across all three tracks is the critical role of Unity Catalog. It provides consistent governance, ensuring that data access and insights are secure and compliant, regardless of how users interact with the data, whether asking questions, building solutions, or composing complex agents.