Thomson Reuters, the 150-year-old information powerhouse serving legal, tax, accounting, risk, trade, and media professionals, has long embedded AI into its operations to deliver trusted content and workflow automation. Now, its Platform Engineering team—a geographically dispersed group ensuring service availability—has pioneered an AI-driven self-service hub using Amazon Bedrock AgentCore, slashing manual toil and accelerating operations.

The core of this transformation is Aether, a central orchestrator agent that routes natural-language requests to specialized agents handling tasks like cloud account provisioning, database patching, network services, and architecture reviews. This setup allows non-technical users to interact seamlessly without grappling with underlying complexities, as detailed in the January 21, 2026, AWS Machine Learning Blog.

“Our engineers were spending considerable time answering the same questions and executing identical processes across different teams,” said Naveen Pollamreddi, Distinguished Engineer at Thomson Reuters. “We needed a way to automate these interactions while maintaining our security and compliance standards.”

Manual Bottlenecks Give Way to Agentic Automation

Prior to Aether, Platform Engineering relied on human coordination for repetitive workflows: provisioning AWS accounts with organizational units, security policies, and networking; managing database lifecycles and upgrades; setting up VPCs and subnets; and reviewing architectures for best practices. These processes, adhering to ITIL standards and integrating third-party SaaS tools, suffered from delays due to inter-team dependencies, high resource costs, and prolonged validation cycles.

Thomson Reuters selected Bedrock AgentCore for its managed infrastructure, framework flexibility (including LangGraph), model agnosticism, and enterprise-grade security, scalability, and reliability. The architecture features a custom React-based web portal hosted on Amazon S3, authenticating via enterprise SSO for permissioned access; Aether as the modular orchestrator using AgentCore Memory for short-term conversation context and long-term user patterns; service-specific agents; and Aether Greenlight for human-in-the-loop oversight on sensitive actions.

The agent registry, built with Amazon DynamoDB and API Gateway, enables agent-to-agent communication across accounts, tracking versions for audits and requiring human validation for production deployment to meet ISRM standards.

TRACK Framework Accelerates Agent Deployment

Central to scalability is TRACK (TR-AgentCore-Kit), a homegrown framework customizing the Bedrock AgentCore Starter Toolkit for Thomson Reuters’ compliance needs, such as asset identification and resource tagging. TRACK manages AgentCore Runtime connections, tool orchestration via AgentCore Gateway, and baseline setups, allowing developers to focus on business logic. It automates agent registration into Aether, generating “agent cards” for easy onboarding.

Development followed a three-phase workflow: discovery and architecture planning to map AWS resources; core development migrating existing solutions and building TRACK; and enhancements like UX refinement and team training. This dual-track approach minimized disruption while building new capabilities.

Bedrock AgentCore’s composable services—Runtime, Gateway, Memory, Identity, Observability—handled infrastructure, letting Thomson Reuters emphasize agent intelligence. As noted in the AWS Bedrock AgentCore page, it supports any framework or model, powering production agents without custom plumbing.

Quantifiable Gains in Efficiency and Security

Launch metrics are striking: a 15-fold productivity boost from automating routine tasks, 70% automation rate, 24/7 reliable runbooks, faster time-to-value through self-service enforcement of standards, and optimized resource allocation. Security improved with guardrails, default patching, and audit trails via human oversight.

“Using Amazon Bedrock AgentCore, TR transformed their platform engineering operations from manual processes to an AI-powered self-service hub,” the AWS blog states. “This approach not only improved efficiency but also strengthened security and compliance controls.”

Developer satisfaction rose with streamlined workflows, consistent standards, and reduced cognitive load, freeing talent for innovation. Cost savings stem from infrastructure optimization and less manual intervention.

AgentCore’s Broader Enterprise Momentum

Thomson Reuters joins early adopters like Ericsson, Cox Automotive, Druva, and Sony, where AgentCore has downloaded over a million SDK times. Dag Lindbo, Head of AI at Ericsson, praised its framework flexibility: “AgentCore also lets us use any agent framework, which is critical to help us scale across many teams and use cases,” per the AWS announcement of general availability on October 20, 2025.

Recent innovations, announced December 2, 2025, at AWS re:Invent, include AgentCore Policy for real-time action boundaries, episodic Memory for experiential learning, and Evaluations for quality monitoring—enhancements trusted by firms like Workday and S&P Global, as reported by About Amazon.

On X, reactions highlight the impact: Sergio Cuéllar posted on January 26, 2026, “Thomson Reuters saw a 15x productivity gain and 70% automation rate, boosting speed, security, and cost efficiency.”

Path to Organization-Wide Adoption

Aether establishes a replicable pattern beyond Platform Engineering, targeting broader teams for productivity gains. Thomson Reuters plans expansion, leveraging AgentCore’s low-latency runtimes (up to eight hours), secure tool integrations, and CloudWatch observability for token usage, latency, and error rates.

This builds on prior Bedrock uses, like Open Arena for no-code AI and CoCounsel for tax research, showcasing a maturing AI strategy. As Pollamreddi and colleagues—Seth Krause, Pratip Bagchi, Sandeep Singh—envision, Aether redefines operations, proving agentic AI’s enterprise viability.

Industry watchers see this as a blueprint: Ernest Chiang noted in his blog that AgentCore moves agents “from prototype to production,” with Thomson Reuters exemplifying cross-industry traction.