AgentDalal

Our LLM-powered WhatsApp bot makes Zepto shopping effortless with voice, breaking digital literacy barriers. Expanding quick commerce from 26M tech-savvy users to 500M+ via simple conversations.

Comment

team: Pavan Kalyan: code Nirmit: code

Inspiration

https://www.loom.com/share/dddc02c70b454952a33d0cf2e62091e1?sid=dca4fc9e-ac33-4e21-a8e7-023e0c4f7484

Our inspiration stems from two key observations in the Indian market. First, while India has over 500 million WhatsApp users, only a fraction of them actively use quick commerce platforms. Second, many potential users face barriers to entry due to complex app interfaces and varying levels of digital literacy.

We were particularly inspired by local grocery stores in India, where shopkeepers maintain personal relationships with customers, remember their preferences, and handle orders through simple phone calls or WhatsApp messages. This traditional model of convenience and personalization, combined with modern technology, formed the foundation of our solution.

Additionally, seeing how WhatsApp has become the default communication platform for millions of Indians across different age groups and technological comfort levels, we recognized its potential as a universal interface for e-commerce. This led us to envision a solution that could make online shopping as natural and conversational as chatting with a local shopkeeper.

What it does

Our WhatsApp bot for Zepto offers several key functionalities:

  • Login to Zepto
  • Search for products using text or voice
  • Suggest weekly groceries
  • Compare prices
  • Find best offers and discounts
  • Add items to cart and process payment
  • Voice-first interface for natural conversations

The bot essentially transforms the entire Zepto shopping experience into a conversational interface through WhatsApp, making it more accessible and user-friendly.

How we built it

  1. We leveraged the WhatsApp Developer Platform
  2. We developed automated functions by crawling the Zepto website
  3. We implemented AI agents to coordinate these functions based on user queries
  4. We integrated Eleven Labs APIs to enable voice-based user interactions

Challenges we ran into

  1. Without access to Zepto's official APIs, we had to develop custom crawling functions using Selenium to access the platform.
  2. WhatsApp's text and voice-only interface lacks the visual elements common in shopping apps. To overcome this, we innovatively used stickers to display product search results to users.

Accomplishments that we're proud of

  1. We successfully built a WhatsApp bot that can place Zepto orders in real time
  2. We developed creative crawling functions to overcome the lack of official Zepto APIs
  3. We implemented an innovative way to display product results as stickers to users

What We Learned

  1. Working with the WhatsApp Business API and its integration capabilities
  2. Implementing text-to-speech functionality using Eleven Labs' API
  3. Developing AI agents using LangGraph and browser automation with Selenium

What's next for AgentDalal

Expand to other quick commerce platforms like Blinkit

Establish partnerships for official API access

Create a one-stop WhatsApp bot that can compare and purchase items across multiple platforms

Built With

  • anthropic
  • elevenlabs
  • langgraph
  • python
  • sarvam
Share this project:

Updates