Auto-Insights AI Agent
A GenAI agent that explores datasets, extracts business insights, and generates executive-ready reports and visualizations.
Demo
Project walkthrough
This project explores how generative AI can support the first moments of data analysis: understanding what is inside a dataset, identifying promising analytical directions, and turning early findings into a clear business narrative.
Business issue
Why I was brought into the project
Business teams receive new datasets regularly, but the first analysis step is often slow, manual and dependent on technical profiles. I stepped in to design an AI-assisted experience that could turn raw data into a first readable business understanding.
Context
The environment around the project
Business teams often need to understand a dataset quickly before deciding where to invest analysis effort.
Functional environment
The functional context involved business users who needed to evaluate datasets, identify relevant analytical angles and understand whether the data could support decision-making.
Technical environment
The technical environment combined dataset profiling, LangGraph multi-agent orchestration, RAG-style reasoning patterns, Dataiku workflows and a Vue.js interface.
Challenges
The agent had to profile data, decide which analyses were relevant, generate insights and keep enough traceability for users to trust the output.
Solution
My contribution and its impact
My contribution to the project
I designed the agent workflow, structured the analysis steps, orchestrated tools and exposed the generated insights in an interactive Vue.js application integrated with Dataiku.
- LangGraph-based multi-agent workflow
- Interactive Vue.js insight exploration interface
- Traceable insight generation flow
- Executive-ready report and visualization generation
Impact
The intervention transformed a slow exploratory phase into a guided analysis experience where users could move faster from raw data to first business conclusions.
- Faster first understanding of business datasets
- Clearer prioritization of analysis opportunities
- More transparent AI-generated recommendations
Impact metrics
Approach
How I structured the work
- Profile datasets automatically to identify structure, quality signals, missing values and usable dimensions.
- Use a planning agent to decide which analyses are relevant before generating insights.
- Expose results through a web interface so users can inspect findings, understand reasoning and continue exploration.
Takeaways
What I learned from this project
- A useful AI agent is not only a model call: it needs a clear workflow, checks and a user experience that makes the output understandable.
- For business users, traceability matters as much as speed because generated insights need to be trusted before they can influence a decision.
- The strongest product value came from connecting AI automation with familiar analytics patterns.