HEMA, a 100-year-old Dutch retailer, built an internal AI assistant using Amazon Bedrock AgentCore to streamline knowledge access across 750 stores. The solution, called HAL, consolidates fragmented knowledge into one governed source of truth, improving efficiency for engineers and analysts. The initiative is part of HEMA’s broader digital transformation strategy, aimed at reducing friction in internal knowledge management.
HEMA’s challenge stemmed from fragmented internal knowledge, where answers were scattered across disconnected wikis, service catalogs, and IT portals. This made onboarding new engineers slow and inconsistent, as teams relied on informal networks rather than structured documentation. HAL addresses this by centralizing knowledge and delivering it where users already work, such as in chat windows and IDEs.
The solution leverages Model Context Protocol (MCP) and Amazon Bedrock AgentCore to create a unified knowledge layer. MCP provides a standardized interface for AI clients to access backend capabilities, while AgentCore enables scalable agent development without custom infrastructure. HAL’s architecture integrates with existing tools like Kiro and Claude, allowing seamless access to internal knowledge without additional engineering.
"Finding an answer that once meant navigating three or four portals, sometimes across an entire afternoon, now happens in seconds," said Mauro Rallo, a co-author of the post. This shift from portal-hopping to instant answers has transformed how HEMA teams collaborate and access critical information.
The announcement follows HEMA’s ongoing efforts to modernize its technology landscape. By using MCP and Amazon Bedrock AgentCore, HEMA is setting the foundation for future enhancements, including turning HAL from a read-only knowledge layer into an action layer.
HEMA did not specify the timeline for this next step, but emphasized the importance of security and access control in the current implementation.
Source: awsml
