        SQL Meets GenBI: AI-Powered

            Dashboards in a Declarative World

                          Applied AI and Data Meetup, June 2025

                                       Simon Späti

                                      Welcome

                                         ▓▓▓ Today's Topic

                                         GenBI - something that is dear to my heart

                                         Why? Because it solves the self-serve BI

                                         challenge I've been trying to solve since

                                         the beginning.

                                         ▓▓▓ The ChatGPT Moment for BI

                                         ChatGPT taught us it's possible to ask

                                         complex questions and get reasonable

                                         answers with little effort.

█████████████

                                     About Me

        ▓▓▓ Simon Späti, web: (https://ssp.sh)

           •  20 years working with data

           •  Started with BI in 2003 (Oracle, SQL Server)

           •  Transitioned to data engineering 10 years

              later

           •  Writing on #dataengineering since 2015 at

              ssp.sh

           •  Working for myself since 2024

█████████████████

                                   Why GenBI?

        ▓▓▓ The Long-Standing Challenge

        ▍ Self-serve BI - something I've heard about and been trying to solve

        ▍ since the beginning

█████████████████████

                                   Why GenBI?

        ▓▓▓ The Long-Standing Challenge

        ▍ Self-serve BI - something I've heard about and been trying to solve

        ▍ since the beginning

        ▓▓▓ A little bit of History

           •  20 years ago: We moved from pixel-perfect to interactive dashboards

           •  15 years ago: Drag-and-drop tools promised self-service through

              cloud first

           •  7 years ago: dbt where analysts could create SQL metrics &

              transformations

           •  Today: The full AI boom helps us improve the workflow

█████████████████████████

                                   Why GenBI?

        ▓▓▓ The Long-Standing Challenge

        ▍ Self-serve BI - something I've heard about and been trying to solve

        ▍ since the beginning

        ▓▓▓ A little bit of History

           •  20 years ago: We moved from pixel-perfect to interactive dashboards

           •  15 years ago: Drag-and-drop tools promised self-service through

              cloud first

           •  7 years ago: dbt where analysts could create SQL metrics &

              transformations

           •  Today: The full AI boom helps us improve the workflow

        ▓▓▓ Why Now?

           •  BI is in the best spot for using LLMs and AI power

           •  BI provides the otherwise missing domain knowledge

           •  LLMs and MCP can empower non-technical business experts

█████████████████████████████

          What is GenBI? (and Self-Serve BI?)

        Changes the way people interact with data by combining the power of

        natural language and analytics.

        ▓▓▓ Before GenBI: Linear & Manual        ▓▓▓ With GenBI: Iterative &

                                                 AI-Powered Workflow:

        ▒▒▒▒ The key difference: Continuous, iterative human-AI collaboration

█████████████████████████████████

                               GenBI Examples

        Let's see GenBI in action...

█████████████████████████████████████

        Create Models/Dashboards with OpenAI

                                       - Rill

        ▒▒▒▒ Generate with OpenAI (https://rilldata.com/):

        Generation of BI artifacts (dashboards, models, data sources), based on

        BI-as-Code (YAML)

█████████████████████████████████████████

        Presentation: Chat Interface with MCP

                                       - Rill

        ▒▒▒▒ MCP (https://docs.rilldata.com/explore/mcp):

█████████████████████████████████████████████

        Presentation: Chat Interface with MCP

                                       - Rill

        ▒▒▒▒ MCP (https://docs.rilldata.com/explore/mcp):

██████████████████████████████████████████████████

        Use MCP Chat interaction to

                         live-query OLAP cube

        --> Try at llm.clickhouse.com/ (https://llm.clickhouse.com/)

██████████████████████████████████████████████████████

                         Three Key Components

        For GenBI to work effectively, we need three critical components.

        ▓▓▓ 1. Quality Data → DWH/Organized Lake

        ▍ No data, no insights

        ▓▓▓ 2. Speed → OLAP

        ▍ No speed, no one uses it

        ▓▓▓ 3. BI-as-Code → Declarative way

        ▍ Human-AI Interface

██████████████████████████████████████████████████████████

                             Why Declarative?

        ▓▓▓ The Language of AI

           •  LLMs and AI-powered workflows work best with declarative interfaces

           •  They pick up context around your business.

        ▓▓▓ Data Modeling Languages

        LookML, MDX, MAQL, SQL and many more: A way for humans to describe and

        model metrics/KPIs for AI to understand.

        The declarative language is usually:

           •  YAML - for configuration & artifacts (dashboards)

           •  SQL - for metrics and business logic

██████████████████████████████████████████████████████████████

                             Why Declarative?

        ▓▓▓ The Language of AI

           •  LLMs and AI-powered workflows work best with declarative interfaces

           •  They pick up context around your business.

        ▓▓▓ Data Modeling Languages

        LookML, MDX, MAQL, SQL and many more: A way for humans to describe and

        model metrics/KPIs for AI to understand.

        The declarative language is usually:

           •  YAML - for configuration & artifacts (dashboards)

           •  SQL - for metrics and business logic

        ——————————————————————————————————————————————————————————————————————————

        ██ > Think of it as the common language between humans and AI.

██████████████████████████████████████████████████████████████████

                     The Developer Experience

        ▓▓▓ GUI Approach                         ▓▓▓ Code-First Approach

           •  Initial: many minutes                 •  Initial: minutes

        ▓▓▓ GenBI Approach

           •  Seconds to generate

██████████████████████████████████████████████████████████████████████

                     The Developer Experience

        ▓▓▓ GUI Approach                         ▓▓▓ Code-First Approach

           •  Initial: many minutes                 •  Initial: minutes

        ▓▓▓ GenBI Approach

           •  Seconds to generate

        ⚠️ Humans are still crutial to steer and adjust.

██████████████████████████████████████████████████████████████████████████

         Faster Iterations; and Bridging Gaps

        ▓▓▓ Improving the BI Workflow with..

           •  BI as the core driver of domain knowledge

           •  Dashboards and Chat interfaces as the presentation layer

           •  More context for LLMs and MCP that help us humans to iterate much

              faster, and easier

██████████████████████████████████████████████████████████████████████████████

         Faster Iterations; and Bridging Gaps

        ▓▓▓ Improving the BI Workflow with..

           •  BI as the core driver of domain knowledge

           •  Dashboards and Chat interfaces as the presentation layer

           •  More context for LLMs and MCP that help us humans to iterate much

              faster, and easier

        ▓▓▓ Wrapping up...

           •  GenBI bridges 20 year gap - LLMs enable self-serve BI

           •  MCPs provide the missing piece on top of LLMs for plug-and-play AI

              integration

           •  BI's semantic business layer provides the context it needs to

              understand your business

           •  Time-to-insight goes from days to minutes, especially for

              non-technical users

██████████████████████████████████████████████████████████████████████████████████

         Faster Iterations; and Bridging Gaps

        ▓▓▓ Improving the BI Workflow with..

           •  BI as the core driver of domain knowledge

           •  Dashboards and Chat interfaces as the presentation layer

           •  More context for LLMs and MCP that help us humans to iterate much

              faster, and easier

        ▓▓▓ Wrapping up...

           •  GenBI bridges 20 year gap - LLMs enable self-serve BI

           •  MCPs provide the missing piece on top of LLMs for plug-and-play AI

              integration

           •  BI's semantic business layer provides the context it needs to

              understand your business

           •  Time-to-insight goes from days to minutes, especially for

              non-technical users

        And Finally:

        ██ > Self-Serve BI achieved in 2025?!

██████████████████████████████████████████████████████████████████████████████████████

                           Want to Know More?

        ▓▓▓ Resources

        ▍ 📈 Slides: Slides (https://ssp.sh/slides)

        ▍ 📝 My Full Article: BI-as-Code and the New Era of GenBI (

        ▍ https://www.rilldata.com/blog/bi-as-code-and-the-new-era-of-genbi)

        GenBI Examples:

           •  GenBI: Blog (https://rilldata.com/) and MCP Rill: Docs (

              https://docs.rilldata.com/explore/mcp)

           •  Chat-MCP with ClickHouse: Application (https://llm.clickhouse.com/)

           •  AI Use-Case with MCP: SQL and DuckDB/MotherDuck: Blog (

              https:/motherduck.com/blog/open-lakehouse-stack-duckdb-table-formats

              #appendix)

           •  GenBI with Cursor: Demo YT (

              https://www.youtube.com/watch?v=Th5Krj14DCI)

        Related:

           •  Declarative Data Stack: Blog (

              https://www.ssp.sh/blog/rise-of-declarative-data-stack/)

           •  And Remember: "Everything becomes BI"; the interface to the

              business. Benn Stancil (

              https://benn.substack.com/p/no-really-everything-becomes-bi)

        ——————————————————————————————————————————————————————————————————————————

        Slides made with Presenterm (https://github.com/mfontanini/presenterm/).

██████████████████████████████████████████████████████████████████████████████████████████

                                   Thank You!

        ▓▓▓ Questions?

        SQL Meets GenBI: AI-Powered Dashboards in a Declarative World

        ——————————————————————————————————————————————————————————————————————————

        The future of BI - declarative, contextual, and AI-powered.

        GenBI is reducing manual work, enabling faster iterations, and expanding

        BI creation.

██████████████████████████████████████████████████████████████████████████████████████████