I led the end-to-end product and design strategy for a new age cutting-edge Business Intelligence platform that fuses the best of modern BI tools like Tableau, Thoughtspot, and DataGPT, but reimagined for the era of agentic AI. Designed for business users and domain experts across industries, the tool enables seamless transitions between dashboards, manual report building, and a dynamic chat-based interface.

The north star? To make insights conversational, workflows intelligent, and data exploration as easy as talking to a colleague! All without compromising analytical power.

The BI ecosystem has hit a complexity wall. Today's tools are powerful but require training and technical setup. The other natural language platforms aim to democratize data access; but fall short when users need control or want to understand the “why” behind AI-generated insights.

We saw an opportunity to build an agentic system! One that doesn't just answer questions, but proactively surfaces insights, lets users interact on their terms, and empowers everyone (not just analysts) to think with data.

Our goals:

• Collapse the learning curve for business users
• Let users seamlessly toggle between Chat, Build, and Data views
• Introduce dynamic insights based on business KPIs
• Build trust by allowing users to debug, verify, and retrace AI logic

I led product and design, owning everything from the initial vision to detailed UX flows and interaction models.

• AI Engineer: IP owner and tech architect
• Backend Developer: Data pipelines and model integration
• Frontend Developer: UI development using React + Tailwind
• Leadership Advisor: Business strategy and investor conversations

Our workflow was async, agile, and deeply iterative. I worked closely with the AI engineer to ensure design constraints respected model architecture, and synced with frontend regularly to translate prototypes into dynamic, production-ready interfaces.

My design approach merged competitive intelligence, user empathy, and future-forward UX thinking:

• Competitive Benchmarking: I broke down workflows across DataGPT, Tellius, Tableau, and Thoughtspot to identify friction points and missed opportunities.
• Community Insights: As a local Tableau community champion, I crowdsourced feedback from domain experts and analysts via LinkedIn to understand how people actually navigate data tools.
• Design Sprinting: Early click-through prototypes helped us validate mental models and fine-tune the tri-mode interaction design (Chat / Build / Data).
• Agentic UX Experiments: Introduced features like dynamic KPI-based suggestions, recent analysis memory, and intelligent handoffs between views.

The biggest challenge was designing for cognitive transition.

We had to:

• Maintain trust and familiarity for dashboard-first users
• Offer manual control without overwhelming non-technical users
• Introduce a conversational UI as the hero — without alienating users trained on legacy BI flows

Solving this meant building a fail-safe UX model: if chat didn’t work, users could switch to Build mode. If they wanted to peek under the hood, Data mode revealed every calculation, transformation, and model used — empowering even super users to debug or iterate further.

We built a modular, agentic BI platform with four core components:

🏠 Home Screen (3-tab interface)

• Chat-Based Analyses: Auto-saved conversational threads
• Dashboards: Horizontal scrollable cards
• Data Pipelines & Models: Quick access for power users

📊 Dashboard Builder

• Drag-and-drop components with customizable chart types
• Chart configs saved as reusable blocks
• Export as dashboard, report, or raw dataset

📑 Interactive Report View

• Chat Mode: Ask questions naturally
• Build Mode: Traditional BI setup with dimensions/measures
• Data Mode: View raw tables, models, calculations, and pipeline lineage

💬 AI-Powered Chat Interface

• Start new analysis or pick up past sessions
• Dynamic prompts based on KPIs + recent data trends
• “Read More” expandable summaries with context
• Auto-switch to Build or Data if user confidence drops

This tool doesn’t just analyze; it collaborates, educates, and adapts.

Proof of Concept success: The first click through prototype landed a paying customer.

Adoption Enthusiasm: During the POC demo itself, business users especially non-technical sales leads; were eager to get their hands on it.

Zero-Training Onboarding: Users could start with chat, then deepen analysis via toggles, without needing walkthroughs or external help.

This wasn’t just a tool. It was a paradigm shift.