A simple business question shouldn’t turn into an hour of manual reporting. Pull the data. Clean it. Check it. Build the chart. Write the summary. Export the PDF. Send it. That’s the workflow I wanted to remove. So I built Plemdo AI Enterprise Analytics. A conversational analytics system where the starting point is simply a business question. The production version I delivered a more advanced version of this system to a client for $3,500. The production implementation went beyond the public demo and included: - Automated data preprocessing before analysis - Claude-based prompt and reasoning skills - Microsoft Fabric / Power BI skills for visual dashboards - Multi-channel interaction across WhatsApp, Telegram and Slack - A real-time Streamlit dashboard - Cloud PDF archival through Google Drive - Subscription and billing workflows The real challenge wasn’t calling an LLM. It was making the data, reasoning, visualization, reporting and delivery layers work together reliably. The base version I’m releasing publicly I wanted developers to be able to study and build on the core idea without publishing the full client implementation. So I created a simpler open-source version: WhatsApp → Conversation Router → Gemini → Data Analysis → Visualizations → Executive PDF → Email + WhatsApp Delivery The public workflow includes: ✅ intent-based conversation routing ✅ context-aware analytics ✅ dynamic chart generation ✅ automated executive PDF creation ✅ email and WhatsApp delivery ✅ full n8n workflow and setup documentation One demo query is: “Show me top beaches in Galway with pubs.” The dataset is just an example. The interesting part is the workflow behind it. Instead of: Question → Analyst → Spreadsheet → Charts → Report → Email you get: Question → Automated Analytics Workflow → Decision-Ready Output That’s the part of AI automation I find most useful — not just getting a model to answer a question, but building the system around it so the answer becomes something the business can actually use.