The DIVA logistics agent, powered by Amazon Bedrock

DTDC is India’s leading integrated express logistics provider, operating the largest network of customer access points in the country. DTDC’s technology-driven logistics solutions cater to a wide range of customers across diverse industry verticals, making them a trusted partner in delivering excellence. DTDC Express Limited receives over 400,000 customer queries each month, ranging from tracking requests to serviceability checks and shipping rates. Their existing logistics agent, DIVA, operated on a rigid, guided workflow, forcing users to follow a structured path rather than engaging in natural, dynamic conversations. This lack of flexibility increased the burden on customer support teams, led to longer resolution times, and resulted in poor customer experience. DTDC sought a more flexible, intelligent assistant—one capable of understanding context, managing complex queries, and improving efficiency while reducing reliance on human agents. To achieve this, DTDC enhanced DIVA with generative AI using Amazon Bedrock. ShellKode, an AWS Partner specializing in modernization, security, data, generative AI, and machine learning, collaborated with DTDC. ShellKode empowers businesses through advanced technology solutions addressing complex challenges and unlocking new opportunities. This article discusses how DTDC and ShellKode used Amazon Bedrock to build DIVA 2.0, a generative AI-powered logistics agent.

Solution overview

To overcome the limitations of the original agent, ShellKode developed an advanced agentic assistant using Amazon Bedrock Agents, Amazon Bedrock Knowledge Bases, and an API integration layer. DIVA 2.0 offers a seamless conversational interface that understands and responds naturally to customer queries. Whether tracking a package, checking shipping rates, or verifying service availability, users can ask questions in their own words without following a rigid script. The AI’s enhanced capabilities allow it to understand context, handle complex requests, and provide accurate, personalized responses—improving customer experience and reducing human intervention.

Architecture and workflow

The logistics agent uses a modular, scalable architecture integrating AWS services:

• Hosted as a static website on Amazon CloudFront and Amazon S3, integrated into the DTDC website.
• User queries are processed by AWS App Runner, which runs the web application and API services.
• App Runner invokes Amazon Bedrock Agents API based on user requests.
• Amazon Bedrock Agents interpret user intent and trigger AWS Lambda functions for:
- Tracking consignments - Pricing information - Location serviceability checks - Support ticket creation These Lambda functions call client APIs to retrieve real-time data:
• Tracking System API: Provides shipment status updates.
• Delivery Franchise Location API: Checks parcel delivery service availability.
• Pricing System API: Calculates shipping rates.
• Customer Care API: Creates support tickets.
The agent passes data to the large language model (Anthropic’s Claude 3.0 on Amazon Bedrock), which generates meaningful, context-aware responses.

Knowledge base

The knowledge base includes web-scraped content from the DTDC website, internal support documentation, FAQs, and operational data. Stored as vector embeddings in Amazon OpenSearch Service, it enables fast, relevant responses. Semantic similarity search retrieves relevant information for Amazon Bedrock to generate accurate replies. If no relevant data is found, a fallback response is returned.

Data storage and monitoring

• Queries and responses are stored in Amazon RDS for PostgreSQL for scalability and relational data handling.
• Amazon CloudWatch Logs captures events like intent recognition and API responses for auditing.
• AWS CloudTrail logs AWS account activity for governance and compliance.
• Amazon GuardDuty monitors for security threats.

Logistics agent dashboard

The dashboard is a static website hosted on CloudFront and Amazon S3, accessible only to DTDC admins. It uses Amazon API Gateway and Lambda to retrieve data from Amazon RDS. The dashboard provides real-time insights into agent performance, including accuracy, unresolved queries, query categories, session statistics, and user interactions. Features such as heat maps, pie charts, and session logs enable continuous improvement and quick issue resolution.

Challenges and benefits

Challenges

• Integrating real-time data from multiple legacy systems to provide accurate tracking, rates, and serviceability.
• Training AI to understand complex logistics terminology and multi-step queries.
• Transitioning from the old rigid system while maintaining service continuity and preserving historical data.
• Scaling to handle over 400,000 monthly queries without performance degradation.

Benefits

• Enhanced conversations and real-time data access: Improved natural language understanding and multi-step reasoning powered by Amazon Bedrock Agents.
• Intelligent data processing and accurate FAQ responses: LLMs process complex queries; Knowledge Bases handle FAQs, reducing wait times.
• Reduced live agent dependency: Customer support query load reduced by 51.4%, enabling focus on critical issues.

Results

• DIVA 2.0 understands natural language queries with 93% response accuracy.
• Over three months:
- 71% of inquiries related to consignments (256,048), 29.5% general inquiries (107,132). - 51.4% of consignment inquiries (131,530) did not result in support tickets. - Of inquiries leading to tickets, 40% started with support before AI, 60% started with AI before involving support. DIVA 2.0 significantly reduces customer support workload and improves overall efficiency.

Summary

Amazon Bedrock transforms a traditional chatbot into a generative AI-powered logistics agent delivering dynamic, natural conversations and improved customer experience. This solution offers a blueprint for businesses seeking to modernize AI assistants while maintaining compliance and scalability. For more information, contact AWS.

---

About the authors

Rishi Sareen – Chief Information Officer (CIO), DTDC A seasoned technology leader with over 20 years in digital transformation and AI-driven logistics innovation. Arunraja Karthick – Head – IT Services & Security (CISO), DTDC Strategic IT and cybersecurity leader driving cloud-native modernization and secure digital transformation. Bakrudeen K – AI/ML Practice Head, ShellKode AWS Ambassador and AI innovator specializing in generative AI and agentic assistants. Suresh Kanniappan – Solutions Architect, AWS Expert in cloud security and industry solutions for logistics and manufacturing. Sid Chandilya – Sr. Customer Relations Manager, AWS Passionate about tech-led business transformation and AI-driven customer experience.

---

Source: Originally published at Amazon Web Services Blog on 07 August 2025.

Frequently Asked Questions (FAQ)

About DIVA 2.0 and Generative AI

Q: What is DIVA 2.0? A: DIVA 2.0 is the enhanced, generative AI-powered logistics agent developed by DTDC, utilizing Amazon Bedrock. It is designed to handle customer queries more intelligently and naturally than its predecessor. Q: How does generative AI improve a logistics agent? A: Generative AI allows the agent to understand context, handle complex and varied queries, and provide more natural, human-like responses, moving beyond rigid, pre-scripted interactions. Q: What were the main limitations of the original DIVA agent? A: The original DIVA operated on a rigid, guided workflow, forcing users into structured paths and lacking flexibility for dynamic conversations, which led to increased customer support burden and poorer customer experience.

Technology and Architecture

Q: What AWS services were used to build DIVA 2.0? A: Key services include Amazon Bedrock Agents, Amazon Bedrock Knowledge Bases, Amazon CloudFront, Amazon S3, AWS App Runner, AWS Lambda, Amazon OpenSearch Service, Amazon RDS for PostgreSQL, Amazon CloudWatch Logs, AWS CloudTrail, and Amazon GuardDuty. Q: How does DIVA 2.0 access real-time logistics data? A: DIVA 2.0 integrates with client APIs such as the Tracking System API, Delivery Franchise Location API, Pricing System API, and Customer Care API to retrieve live data. Q: What kind of data is included in DIVA 2.0's knowledge base? A: The knowledge base comprises web-scraped content from the DTDC website, internal support documentation, FAQs, and operational data, stored as vector embeddings for efficient retrieval.

Performance and Results

Q: How accurate is DIVA 2.0's response accuracy? A: DIVA 2.0 understands natural language queries with a 93% response accuracy. Q: What was the impact of DIVA 2.0 on customer support workload? A: DIVA 2.0 reduced the customer support query load by 51.4%, allowing human agents to focus on more critical issues. Q: How many customer queries does DTDC handle monthly? A: DTDC Express Limited receives over 400,000 customer queries each month.

Crypto Market AI's Take

The integration of generative AI, like that powering DIVA 2.0 with Amazon Bedrock, exemplifies the growing trend of leveraging AI to enhance customer service and operational efficiency across various industries. In the crypto space, similar AI advancements are being used to develop sophisticated AI crypto trading bots that can analyze market sentiment, execute trades at optimal times, and manage risk more effectively than human traders. Furthermore, the concept of building a robust knowledge base, as DTDC did, is directly applicable to creating advanced AI market analysis tools that can sift through vast amounts of data, news, and regulatory information to provide actionable insights for cryptocurrency investors. This case highlights how AI can transform complex, data-intensive operations into streamlined, intelligent processes.

More to Read:

• AI Agents: Revolutionizing Business Automation and Customer Engagement
• Understanding the Impact of AI on the Cryptocurrency Market
• The Rise of AI-Powered Trading Bots in Crypto