AI is reshaping how teams explore and reason about data. Business users increasingly expect natural-language answers. Data engineers want analytics they can version, test, and deploy like software. AI researchers need structured semantic models that agents can reliably operate on. To serve all three, analytics must evolve into something more semantic, more interactive, and more code-driven.

This workshop introduces AI + BI — an approach where dashboards, metrics, and models are created and explored through natural language, grounded in a robust semantic layer. At the core is BI-as-Code, the practice of expressing metrics, transformations, and dashboards declaratively using SQL and YAML. Inspired by the success of Infrastructure-as-Code, this approach makes your analytics stack reproducible, transparent, and ready for collaboration between humans and AI systems.

In this hands-on workshop, you’ll build an end-to-end AI + BI workflow using your own GitHub commit history. You’ll ingest commit data into DuckDB or ClickHouse, model it using SQL and YAML, and construct a governed metrics layer in Rill that becomes the shared contract for both humans and conversational AI.

You’ll then use natural-language prompts to explore engineering productivity questions—receiving responses that include SQL, citations, visualizations, and clear explanations. You’ll also learn how to secure your dashboards and metrics by user attributes, ensuring that both human users and AI agents see only what they’re authorized to see.

By the end, you’ll experience what it feels like to move from raw data to real decisions with a modern, AI-native, code-driven analytics workflow.

What You’ll Build

• A ClickHouse dataset created from your Git commit logs
• SQL + YAML analytics models following BI-as-Code practices
• A governed, semantic metrics layer designed for conversational AI
• Natural-language analytical workflows with SQL, citations, and charts
• A GitHub-backed BI project with GitOps deployment and governance
• A complete engineering productivity dashboard for your team

What You’ll Learn

• How the principles of Infrastructure-as-Code can be applied to BI-as-Code
• How a semantic metrics layer enables trustworthy conversations with AI
• How to implement governance and secure data access in metrics and dashboards
• How to deploy BI projects using GitHub + GitOps workflows
• How to build an AI-ready end-to-end analytics project

Target Audience

Level: Intermediate developers and AI practitioners
Prerequisites: Basic Python knowledge, familiarity with LLMs/prompting concepts
Ideal for: Data Engineers, AI/ML developers
Hardware requirements: Laptop with 8G RAM, Linux/MacOS/Windows with WSL

High level Workshop Flow (1.5-2 hrs)

Welcome & Context (15 min)

• The shift from traditional BI to Generative BI
• Why BI-as-Code provides the technical foundation
• What we will build end-to-end
• Ensure every participant has the environment setup properly

Data Extraction & Engine Setup (15 min)

• Extract GitHub commit logs from your own repos
• Load into ClickHouse
• Quick exploration of the dataset by connecting ClickHouse to Claude using MCP

Modeling with BI-as-Code (10 min)

• Clean and model commit data in SQL
• Define transformations in YAML

Build the Semantic Metrics Layer (15 min)

• Use GenAI to define metrics (velocity, churn, PR cycles)
• Preview metric definitions
• Add time semantics, descriptions, and metadata
• Deploy your dashboards

Adding Security (10 min)

• Configure access policies
• Implement row-level and column-level security
• Test with mock users
• Validate that dashboards and AI responses apply the same rules

Conversational AI Insights (10 min)

• Ask natural-language questions about engineering activity
• Review responses with SQL, charts, citations, and explanations
• Iterate on insights, refine prompts, and explore deeper

Further tune the dashboards (10 min)

• Generate charts from metrics
• Add filters, dimensions, and time controls
• Assemble a dev-productivity dashboard end-to-end

Closing & Q&A (20 min)

• Review Semantics, Speed, Stewardship
• Overview of Real-world projects handling large scale datasets
• How to take BI-as-Code into production

Here is a demo project what audience can build on top of their own git commit datasets -
https://ui.rilldata.com/demo/rill-github-analytics