NoCode-Agent: WhatsApp AI Sales & Support

Turn a business catalogue into a WhatsApp AI customer agent. Answer product questions with Amazon Bedrock, AWS End User Messaging Social, and Amazon SES email follow-up.

Comment

NoCode-Agent: WhatsApp AI Sales & Support

Inspiration

Small businesses often receive repeated customer questions about products, prices, and services. Answering every message manually takes time, while generic chatbots may not understand a business's own product information.

We wanted to make AI-powered customer communication easier for businesses without requiring them to build a chatbot from scratch.

What It Does

NoCode-Agent is a no-code platform for creating AI customer agents that communicate through WhatsApp.

Business owners can configure an agent and provide product catalogue information. Customers can then ask questions through WhatsApp and receive AI-generated answers based on the available business knowledge.

The platform connects customer conversations with AI responses, product information, and email follow-up workflows.

How We Built It

We developed the application using a React dashboard and FastAPI backend services.

The main technologies and services include:

  • React: Provides the business-owner dashboard and user interface.
  • FastAPI: Handles API requests and backend operations.
  • Strands Agents SDK: Orchestrates the conversational AI workflow.
  • Amazon Bedrock: Powers AI responses and embedding generation through configured inference profiles.
  • AWS End User Messaging Social: Supports WhatsApp communication.
  • Amazon SNS and Amazon SQS: Carry incoming WhatsApp events through the tested development flow.
  • Amazon S3: Stores business documents and catalogue assets.
  • PostgreSQL: Stores application data, conversation records, and the in-app knowledge index.
  • Amazon SES: Supports email notifications and customer email workflows.

The agent uses conversation context and available business information to prepare relevant responses for customers.

AWS Communication Services Integration

WhatsApp is the primary customer communication channel in NoCode-Agent.

A customer sends a WhatsApp message, the application receives and processes the incoming event, the AI agent prepares a response, and the reply is sent back through AWS End User Messaging Social.

Amazon SNS and Amazon SQS are used in the inbound event pipeline. During development, a local bridge connects incoming events to the application.

We also integrated Amazon SES for email workflows and verified email delivery during development. These integrations demonstrate how AWS communication services can work alongside conversational AI to support customer engagement and business follow-up.

AI and Knowledge Retrieval

A useful customer agent needs access to business product information, not just a general-purpose language model.

NoCode-Agent includes an in-app knowledge retrieval flow that uses PostgreSQL and Titan Text Embeddings V2 through a configured Amazon Bedrock inference profile. This allows the application to retrieve relevant information from the available business knowledge when preparing responses.

The current implementation uses this approach because the intended Amazon Bedrock Knowledge Bases workflow is blocked by the AWS account's organization policies.

What We Tested

During development, we verified the following:

  • The dashboard and live AI chat passed the recorded local smoke checks.
  • Incoming WhatsApp messages reached the application through the configured AWS messaging event pipeline.
  • AI-generated replies were sent back through AWS End User Messaging Social.
  • Product-related questions were answered using the sample skincare catalogue.
  • Email delivery through Amazon SES was verified.

These are development verification results. They should not be interpreted as production-scale performance or business-impact measurements.

Challenges We Faced

One of the main challenges was working within AWS organization policies that restricted the intended Bedrock Knowledge Bases workflow and certain resource-management operations.

We adapted by using an in-app knowledge retrieval approach and configuring the messaging event resources needed for development.

We also worked through challenges involving asynchronous event processing, conversation context, catalogue-based answers, and connecting AI-generated responses with external communication services.

These challenges helped us understand that building a useful AI communication product involves more than generating text. It also requires reliable message handling, relevant business context, and appropriate follow-up workflows.

Potential Impact

NoCode-Agent aims to help businesses reduce repetitive manual responses, make product information easier for customers to access, and connect customer enquiries with business follow-up.

The platform is intended to make AI-powered customer engagement more accessible to businesses that do not have a dedicated AI engineering team.

For future pilots, we plan to measure first-response time, catalogue-question resolution, customer handoffs, and successful follow-up delivery. We have not yet measured business impact at production scale.

Future Roadmap

Our next steps are to complete and verify a secure cloud deployment, validate the Lambda-to-AgentCore runtime integration, strengthen tenant isolation and operational security, and evaluate the platform with real businesses.

We also plan to expand communication channels and business-specific agent workflows as their integrations are implemented and verified end to end.

Our Goal

Make AI-powered customer engagement accessible to businesses without requiring them to build and maintain their own AI infrastructure.

NoCode-Agent brings business knowledge, conversational AI, WhatsApp communication, and email follow-up together in one platform.

Built With

Share this project:

Updates