The customer currently has an in-house developed chatbot managing customer inquiries on two of their websites.
This chatbot handles approximately 120 different intents, they have a team of four analysts who manually evaluate the chatbot’s performance and are responsible for its optimization.
Our objective is to implement an AI Assistant solution for the client, replacing the current chatbots to improve operational costs, overall customer satisfaction, chatbot performance, streamline the optimization process with greater automation, and enable them to scale the solution to their other websites.
Customer interactions were spread across multiple channels, including the website, WhatsApp, and Facebook, without a unified interface, resulting in fragmented experiences and high operational effort.
The client also managed an in-house chatbot across two websites, which required four analysts to continuously review performance, update more than 120 intents, and maintain duplicated configurations. Operating several disconnected systems and redundant workflows significantly increased maintenance costs and limited the customer’s ability to scale efficiently.
Customer needed to significantly reduce dependency on human intervention by introducing automation and generative AI improvements; also, centralize chatbot flows into a single advisor interface and deploy the solution across seven additional websites with no duplicated configurations.
A conversational AI assistant was deployed using an AWS serverless architecture. The SPA frontend runs on S3 + CloudFront, with a chat interface connected to backend services through API Gateway and Lambda. DynamoDB (on-demand) stores chat history and the Q&A base, while Claude via Amazon Bedrock handles model inference. CI/CD is managed with GitHub Actions and AWS CDK (SST). Twilio was used for WhatsApp integration.
The AWS-based assistant automates FAQs, reduces manual workload, and integrates all channels into one scalable backend. It delivers real-time quotes, processes queries using Bedrock LLMs, and guides users through reservation and financing flows. The solution improved response accuracy and increased conversion rates by over 25%.
Key features include Twilio WhatsApp integration, Dialogflow intent routing, Claude models via Amazon Bedrock, an SPA agent dashboard on S3 + CloudFront, segmented Markdown knowledge bases on S3, automated specialized-agent orchestration, CI/CD with GitHub Actions + CDK, and compliance with AWS Well-Architected and internal security standards.
The customer is an online travel-services platform that offers car rentals and hotel bookings by partnering with major international rental agencies and accommodation providers.
They provide users with a wide selection of vehicles and lodging options, often promoting competitive pricing and bundled travel-insurance benefits.
The platform aims to simplify travel planning by combining transportation and lodging reservations in a single interface, making travel more convenient and cost-effective for customers across the Americas.
The solution was implemented using an AWS-based architecture including Lambda for compute, Amazon Bedrock for model inference, API Gateway for API management, DynamoDB for chat history and knowledge base, S3 + CloudFront for the SPA interface, and Twilio for WhatsApp integration. CI/CD was configured using GitHub Actions and AWS CDK (SST).
The customer maintained a team of four analysts who handled chatbot performance optimization and intent management. They collaborated by providing chatbot flows, intents, and operational context for automation and scaling.
The solution reduced manual workload for analysts, automated FAQ handling, improved response quality and enabled deployment across multiple channels and websites. Chatbot-driven conversions improved by over 25%.
Using a fully serverless architecture simplified scaling and maintenance; integrating Bedrock with existing intent systems proved to be effective for reducing manual intervention, and also, having aunified dashboard greatly improved the agent workflow.
The solution lowered operational costs, improved customer satisfaction through faster response times, and provided a scalable foundation for adding new brands and channels. Centralized analytics improved decision-making and overall efficiency.
Plans include expanding the assistant to additional websites, deploying specialized AI agents, and scaling the multichannel support architecture to more regions and brands. The solution is also designed to be extensible, enabling future expansion into additional use cases and product lines as the customer’s needs evolve.
We are an AWS cloud-native service provider focused on helping customers harness the power of the cloud to drive business growth, in a secure and cost-effective way.
Whether you’re just getting started with the cloud or looking to enhance your digital experience, we offer a team of experts to help you take advantage of cloud ecosystems.
We are a customer-centric company, and for this reason we choose the cloud service or solution that best suits your needs.
Copyright © 2022 Nub8 LLC – All Rights Reserved.
