From 30% to Zero Support Backlog: How Agentic AI Transformed Customer Support for XpressBees

About Customer

‍

XpressBees, a premier Indian logistics provider, delivers end-to-end supply chain solutions for B2B and B2C segments. Founded in 2015, serving over 19,000 pin codes, XpressBees manages millions of daily shipments with unmatched efficiency. Its expansive network includes 4,500+ fulfillment centers, 260+ hubs, and a dedicated workforce of over 20,000 Last Mile Delivery Partners, ensuring swift and reliable deliveries. Committed to excellence, XpressBees continuously adapts to meet the dynamic logistics demands of customers across India and beyond. 

‍

Business Challenge

‍

The business handles a substantial volume of customer emails daily, typically between 3,000 to 5,000, rising to 15,000 during peak seasons. Without a structured prioritization system, only a fraction of these emails are manually reviewed, often at random. This makes it difficult to identify and respond to high-value or time-sensitive enquiries in a timely manner, impacting SLA adherence and overall customer satisfaction.

‍

Operational challenges such as inconsistent handling of Non-Delivery Reports (NDRs) and the absence of ticket workflow automation have added to the complexity, resulting in delayed resolutions and a fragmented support experience. Compounding this is the lack of real-time visibility and granular insights across customer queries, which limits the ability to track support performance, manage escalations efficiently, and make data-driven decisions aligned with business objectives.

‍

Shellkode Solutions

XpressBees, which handles a large number of customer queries, was using an in-house ticketing tool for support requests. As the queries grew more complex, recurring issues like late deliveries and false shipment complaints started overwhelming the system. To solve this, ShellKode implemented an AI-powered email automation solution using Amazon Bedrock’s Foundation Model and a RAG approach. This helped reduce the backlog and speed up issue resolution.

‍

The solution started by training a Large Language Model (LLM) on past customer support data covering over 25 recurring categories. We then built an email categorization and draft generation engine using a pretrained model and RAG, and integrated it with the XpressBees ticketing system through an API.

‍

When new emails arrive, the engine extracts key entities like customer IDs and dates, classifies tickets into high, mid, or low priority, and enriches them with contextual data. Using chain-of-thought processing, the solution connects to internal systems to generate real-time, personalized replies automatically sent to live agents via a callback URL for review and ticket status updates.

‍

Beyond automation, ShellKode’s solution enforced SLA compliance, improved visibility for high-priority issues, and dramatically improved response and resolution times.

‍

Results and Benefits

‍

• Achieved 90% overall accuracy, significantly enhancing response quality and SLA compliance across 20 support categories.
• Improved operational performance 5x and reduced costs through strategic automation, enabling smarter resource utilization.
• 100% of incoming emails are successfully routed through the AI system, ensuring consistent and timely responses.
• Delivered a 50% boost in efficiency for customer support representatives by streamlining workflows.
• Eliminated support backlogs, bringing them down from 30% to 0%, improving customer satisfaction and resolution time.