# Build a Business Dashboard with AI in MINUTES (for dummies...)

## Executive summary

This tutorial demonstrates a full-cycle approach to building a business analytics dashboard using AI coding agents (Codex and Claude Code) within the Hugging Face ecosystem. The process involves uploading a dataset, exploring it via Data Studio, defining key performance indicators (KPIs) and business questions, generating an initial dashboard draft, and iteratively refining the visualization using expert feedback. The final dashboard is then deployed to a private Hugging Face Space for secure team sharing.

## Key takeaways

- AI Agent Setup and Authentication: To use coding agents like Codex or Claude Code, users must authenticate the Hugging Face CLI/plugin. This requires a Hugging Face Pro/Academia account and involves granting organizational permissions to ensure the agent can access and manage private resources.
- Data Exploration via Data Studio: The Hugging Face Data Studio allows users to chat directly with the dataset, enabling quick understanding of the data structure, identifying the target variable (e.g., 'Y' in the bank marketing dataset), and suggesting potential analytical axes without local data loading.
- Iterative Dashboard Improvement: The process is not linear; after the AI generates a first draft, the human analyst must provide specific feedback (e.g., 'replace bar plots with a heatmap,' 'move filters to a collapsible sidebar') to refine the dashboard and maximize its business utility.
- Deployment to Private Spaces: The final dashboard, which is initially running locally, must be uploaded to a private Hugging Face Space. This ensures that the application is hosted on the platform and can be securely accessed by the organization's team members.

## Technical details

- Coding Agent Integration: Codex and Claude Code can be integrated using the Hugging Face plugin or the dedicated Hugging Face CLI. The CLI approach is detailed as more robust for advanced operations.
- Data Preparation and Access: Datasets (like the Bank Marketing dataset) can be uploaded to Hugging Face and accessed by agents via a link, eliminating the need to download the CSV locally. The dataset must be set to private for organizational security.
- Business Intelligence Workflow: The workflow emphasizes defining specific business questions and proposing a small set of KPIs (e.g., Baseline Take Up Rate, Group Size and Share, Lift) *before* asking the AI to build the dashboard. This ensures the output aligns with stakeholder needs.
- Dashboard Visualization Best Practices: The tutorial highlights improving visualizations by suggesting alternatives (e.g., replacing bar plots with heatmaps) and improving UI/UX by moving filters to a sticky, collapsible sidebar.

## Practical implications

- The process outlines a complete data product lifecycle, from raw data ingestion to interactive, shareable dashboard deployment, which is highly relevant for DataOps and MLOps engineers.
- It demonstrates how AI agents can accelerate the initial prototyping phase of data visualization, allowing engineers to focus on the strategic refinement and business logic.
- The use of private Hugging Face Spaces provides a structured, version-controlled environment for sharing data applications within an organization.

## Topics

Business Intelligence, Data Visualization, AI Agents (LLMs), DataOps, Hugging Face Ecosystem, Workshop slides (PDF), Bank Marketing dataset, Data Studio Viewer

Source: https://www.youtube.com/watch?v=44RXU8Cxoas