As agent adoption scaled, we saw a common pattern emerge across enterprises, including our own sales organization: specialized agents deliver value, but without orchestration, users carry the cognitive load of choosing between them. At AWS Sales, this meant more than 20 domain-specific agents deployed across the global organization, with representatives context-switching between systems instead of focusing on customer conversations. In this post, we show you how we built Field Advisor on Amazon Bedrock AgentCore to solve this, the architecture decisions we made, and the measurable results that we’ve seen.

The challenge: Agent proliferation without orchestration

AWS sales representatives faced a significant challenge as AWS scaled AI adoption. With more than 20 domain-specific agents handling customer relationship management (CRM) operations, meeting scheduling, customer insights, product recommendations, and compliance checks, representatives needed to know which agent to invoke for each task. They also had to manage context across fragmented conversations and manually combine outputs from different systems. This overhead consumed time that could be spent understanding customer needs and delivering solutions.