AWS Blueprint Uses AgentCore and Strands Agents to Automate Natural‑Language Dashboard Changes

AWS published a solution blueprint demonstrating how to build AI‑powered dashboard automation agents that accept natural‑language requests and perform discovery or modification operations against dashboards. The blueprint targets a common bottleneck where business analysts must wait days for IT teams to interpret requirements and deploy dashboard changes, and it presents an approach intended to speed those cycles and make scattered dashboard data more actionable for end users.

The architecture uses a multi‑agent design with three specialized agents. The Find Dashboard Agent searches dashboards and retrieves column metadata; the Modify Dashboard Agent validates columns, updates visuals and creates new dashboard versions; and the Orchestrator Agent routes requests by intent. Amazon Nova classifies incoming queries as conversational or operational: conversational requests receive direct LLM responses, while operational requests are forwarded to the specialized agents for autonomous execution.

The implementation runs on Amazon Bedrock AgentCore alongside the Strands framework. AgentCore supplies an agentic platform with intelligent memory, a gateway for secure tool and data access, production‑grade security, dynamic scaling and observability. Strands Agents provides a code‑first framework to integrate agent logic with AWS services. The solution also relies on AgentCore Memory to maintain session state and on AgentCore observability to log decisions and trace API interactions.

Operational flows validate requested dashboard changes against available dataset columns, preserve original dashboards to enable rollback, and maintain audit trails and security controls during modifications. The Modify Dashboard Agent can update table visuals and create new dashboard versions so teams can automate repeated UI changes while retaining governance and the ability to revert to prior definitions.

To adopt the blueprint, implementers are instructed to build the three agents, deploy them to Amazon Bedrock AgentCore and test via the AWS Management Console. Prerequisites listed in the reference implementation include an AWS account with permissions for Amazon Bedrock, Amazon Quick and IAM; an active Amazon Quick account with existing dashboards; and IAM permissions granting the agent access to Quick APIs (quicksight:ListDashboards, quicksight:DescribeDashboard, quicksight:DescribeDashboardDefinition, quicksight:DescribeDataSet, quicksight:CreateDashboard). The reference implementation requires Python 3.10–3.13 for direct code deployment.

The blueprint emphasizes preserving governance and observability while automating dashboard changes, offering a path to reduce manual handoffs between analysts and IT without sacrificing rollback capability or auditability.