Webapp to WhatsApp Chatbot for Senior Management

Our customer is a large insurance service company serving insurance companies of India related to crop and rural insurance. It conducts crop cutting experiments, surveys for insurance claims and related services. The mobile app to conduct surveys is used by field agents all over India and the webapp for overall governance and reporting is used by management in head quarter, at customer locations when meeting in person or in district offices. 

While web application has query screens to extract required data and generate required reports, it comes with a few challenges: 

• Requires login to a separate web application and UI is tied to a web interface 
• Users must remember the right screens, menus, and query filters 
• One gets the whole report even though, he needs only a couple of metrics 

This makes retrieving insights repetitive, time-consuming, and not the most efficient way. 

Solution

We have created a conversation agent. Please watch the video in the frame. Senior Management can ask the agent 

1. Specific data / metric that they need to see 
2. The agent asks what specific information (query parameter) it needs 
3. Perform specific analytics. It can even use publicly available data (e.g. crop yield in a district in last 10 years) to create a wholistic picture. 

Send report / information on intended channel – email, WhatsApp, SMS etc

Why a Conversational Agent? 

The challenge isn’t about technical complexity; it’s about usability and experience. 

• Accessing the same interface repeatedly is boring. 
• Another login, another app to remember, another set of steps. 
• User must always know exactly where to click. 

A chatbot changes the game:  

• It feels like having a conversation with the system 
• It’s fine to forget the steps, just ask what you want 
• Analytics and external insights can be pulled in seamlessly 
• Works across channels like web and WhatsApp, Slack, Teams 

In short, the conversational agent makes interaction simpler, faster, and smarter.

Functionality  

Our CCE Agent allows authorized users to interact naturally with related data, and even ask other relevant questions anytime in between: 

• Understands natural queries 
  • “Show me CCE experiments in Ahmednagar.” 
  • “List Localized experiments for Rabi 2024.” 
• Asks for missing inputs
  • “Which year are you interested in?” 
  • “Do you mean Rabi or Kharif season?” 
  • “Which district should I fetch results for?” 
• Frames inputs and executes API 
  • Collects required parameters based on query type (e.g., year, season, district) 
  • Builds the correct GraphQL request 
  • Executes the AWS AppSync GraphQL API 
• Returns results instantly 
  • Displays the top results clearly 
  • Works across different query types 
  • Can handle other questions mid-conversation 

From conversation → query → results, all in a single, seamless flow. 

Technology Stack 

This stack ensures scalability, clean data handling, and an interactive user experience: 

• React JS (web app) | WhatsApp (over Twillio) 
• FastAPI (Lightweight backend for the agent) 
• LangGraph (Orchestration layer) 
• LangChain + gpt-4o (Model) 
• PostgreSQL with pgvector (as RAG) 
• DuckDuckGo Search (External information) 
• GraphQL (AWS AppSync) – Data access layer. 
• CCE App Installed in AWS 

This setup enables dynamic adaptation to different query types while staying simple and intuitive. 

Why It Matters

This isn’t just about CCE data.  

The same approach applies to any domain where structured data is valuable and natural language makes access easier; agriculture, BFSI, healthcare, logistics, or finance. 

It empowers authorized users to get answers instantly, whether for CCE or any other relevant information, without needing technical expertise. 

AI agents like this are no longer experiments; they’re becoming practical interfaces for real-world data systems. 

We’d love your feedback – how else would you imagine using such an agent?