Visual Representation of Idea Prioritization (Ease of Prioritization)
User Interface
ReAct Style Reasoning for Ranking

Pluto – AI Agent for Prioritizing Innovation Ideas

Case #1 – ReAct-based Idea Portal Assistant
Team Pluto
Team Members: Mukund Komati, Parvez Shaik, Vijay Sunkugari, Yamini Ane

🚀 Inspiration

We’ve all seen how organizations collect hundreds of ideas through internal portals, but very few ever make it to implementation. The challenge isn’t a lack of creativity — it’s figuring out which ideas are actually worth the time, resources, and effort.

We were inspired to solve this decision-making bottleneck using AI.

We wanted to build a system that could think like a product manager, balancing effort, impact, and strategy — but also explaining why it made its decisions.

💡 What We Built

We built Pluto, a GPT-powered AI agent that helps organizations:

Evaluate and prioritize innovation ideas
Rank the top 3 ideas to pursue
Provide transparent reasoning
Let the user tune priorities (ROI, effort, alignment)
Learn from past feedback

We implemented Pluto using:

Python & Ollama for backend
Streamlit for the user interface
LLaMA 3.2 as our language model

Note: Pluto doesn’t just score ideas — it justifies its reasoning using the ReAct (Reasoning + Action) framework.

👥 Team Collaboration

We’re a team of four. We divided responsibilities based on strengths but collaborated closely throughout:

Frontend and interface design
GPT prompt engineering and logic flow
Data modeling and simulation
Feedback loop and ReAct integration

Everyone contributed to multiple parts of the stack.

🗃️ Custom Dataset

We created a realistic idea portal dataset of 116 innovation ideas based on challenges in a health tech company.

Each idea included fields such as:

Effort-related: Ease of implementation, resource availability, dependency freedom
Impact-related: Value created, user demand, business impact
Strategic & feasibility factors: Scalability, regulatory fit, adoption likelihood, brand alignment, technological feasibility

We also factored in Strategic Alignment, which adjusted based on whether the organization was in a growth or profit-maximizing phase.

We used GPT to simulate and estimate these values where needed.

🧠 Reasoning Logic with ReAct

Pluto uses a layered approach for scoring:

Custom weights for each field (ROI, effort, etc.)
GPT evaluates each idea’s value holistically
ReAct logic applies structured reasoning before action

🧭 Here’s how it works:

The user uploads a CSV of ideas
Pluto parses each idea’s fields
GPT evaluates parameters and provides a score
Pluto ranks the Top 3 Ideas
Each ranked idea includes a clear explanation of why it was chosen
Users can adjust sliders to tweak the weights
Optional feedback allows users to share past context or organizational priorities
Pluto incorporates feedback into future analysis

📊 Key Features

CSV Upload: Load your idea portal data easily
Interactive Dashboard:
- Top 3 recommendations with detailed reasoning
- Full idea list with scores
- Sliders to adjust importance of ROI, effort, etc.
Effort vs ROI Plot:
Visualizes trade-offs and quick wins
Live Feedback Loop: Users can input contextual feedback (e.g., “We are in scaling phase”)
GPT uses this to adapt how it evaluates future ideas