Intelligent First Response Agent for EDI Support
Event-driven, serverless AI system that automatically analyses incoming EDI support tickets and generates intelligent first responses — using multi-step AI reasoning, internal knowledge enrichment, and human-in-the-loop review.
Impact Metrics
EDI Support Teams Slowed by Manual First-Response Drafting
Cleo's EDI support teams handling high volumes of incoming tickets face a persistent bottleneck: crafting the initial response. Each ticket requires an engineer to read the customer's EDI issue, research relevant handling procedures and past resolutions, and draft a thoughtful, accurate reply — a process that consumes significant time even for experienced engineers.
Response quality and consistency vary across engineers and shifts. Different engineers may address the same type of EDI issue with different levels of detail, tone, or accuracy — leading to an inconsistent customer experience and increased follow-up interactions when initial responses miss key information.
Cleo needed a system that could automatically analyse incoming EDI tickets, enrich them with relevant context from internal knowledge sources, and generate high-quality draft responses for engineers to review and send — reducing drafting time while maintaining quality and consistency.
Key Pain Points
Event-Driven Serverless AI Pipeline with Multi-Step Reasoning
We built an event-driven, serverless system on AWS that automatically processes incoming EDI support tickets through a multi-step AI reasoning pipeline. The architecture uses API Gateway for ticket ingestion, SQS queues for reliable event processing, Lambda functions for distinct processing stages, and DynamoDB for state management — ensuring scalability and resilience.
The pipeline enriches incoming tickets with relevant handling notes and internal EDI knowledge, then applies multi-step AI reasoning using LangChain orchestration with AWS Bedrock (Claude) to analyse the issue and generate a well-structured draft response. Prompt versioning with static templates ensures consistent, controllable AI behaviour across all ticket types.
The system operates with a human-in-the-loop model: AI-generated responses are posted as internal draft notes for engineers to review, edit if needed, and send to the customer. This approach ensures quality control while dramatically reducing the time engineers spend on initial response drafting. Full observability via CloudWatch provides monitoring across the entire pipeline.
Our Approach
Key Features Delivered
Built With
Outcomes Achieved
The Intelligent First Response Agent reduces the time Cleo's EDI engineers spend drafting initial customer replies while improving response consistency — with AI-generated drafts enriched by internal knowledge and reviewed by engineers before reaching the customer.
Related Case Studies
RAG-Powered Past Ticket Intelligence for Zendesk
Ingested historical Zendesk tickets into a knowledge base and built an in-app Zendesk widget that analyses open tickets, cites relevant past tickets, synthesises root cause, diagnostic steps, and resolution — with AI chat for engineers.
AI-enabled Intelligent Search Solutions for Documents
RAG-powered semantic search and Q&A system for MoSPI's vast document archive — enabling natural language querying with voice and text input, multilingual support (Hindi, English, Kannada), and deep citations linking directly to source PDF pages.
EUCfy — Excel AI Productivity Copilot
AI-powered Excel productivity tool enabling natural language data analysis, formula generation, visualization, and automated reporting.