ShipStation Global resolves 97.5% of IT queries with Bedrock agent

The architecture uses a Strands agent on AgentCore runtime, with tools exposed via Model Context Protocol (MCP) through AgentCore Gateway. A key feature is the mining of historical Slack threads using DSPy’s Recursive Language Model to extract verified fixes, creating a high-quality knowledge base. The agent is designed to stay silent on social messages or when humans are already assisting, ensuring it does not disrupt natural conversations. It also handles after-hours queries, answering 94% of those that arrive outside US business hours.

This case study highlights how companies can use AWS infrastructure to automate internal support. By integrating with existing tools like Slack, Confluence, and Jira, ShipStation Global reduced operational friction. The system uses strict validation and access controls to ensure security, such as requiring explicit confirmation for ticket creation and restricting image analysis to Slack-hosted files. The result is a self-service option that frees IT staff from repetitive tasks while improving employee experience.