Transforming Platform Engineering with AI: A Case Study on Thomson Reuters’ Agentic System Using Amazon Bedrock AgentCore
Co-Authors: Naveen Pollamreddi and Seth Krause, Thomson Reuters
Introduction to Thomson Reuters and the Role of AI in Operations
Business Challenge: Addressing Inefficiencies in Platform Engineering
Current State: An Overview of Service Delivery and Challenges
Solution Overview: Building an Autonomous Agentic System
Solution Workflow: Steps to Develop and Deploy the Innovative System
Aether: The Orchestrator Agent and its Features
Service Agent Development Framework: Simplifying Agent Deployment
Agent Discovery and Registration: Ensuring Seamless Communication
Aether Web Portal Integration: A User-Friendly Interface
Human-in-the-Loop Validation Service: Enhancing Security and Compliance
Outcome: Transformative Results and Key Metrics
Conclusion: A Roadmap for Operational Excellence through Automation
About the Authors: Meet the Experts Behind the Innovation
Transforming Platform Engineering with AI: The Aether Project at Thomson Reuters
This post was co-written with Naveen Pollamreddi and Seth Krause from Thomson Reuters.
Thomson Reuters (TR) stands tall as a leading AI and technology company, dedicated to delivering trusted content and workflow automation solutions. With over 150 years of expertise, TR provides essential solutions across the legal, tax, accounting, risk, trade, and media sectors in a fast-evolving world. AI plays a crucial role at TR, enhancing how information is created, connected, and delivered to customers. This blog post delves into how TR’s Platform Engineering team transformed its operational productivity by moving from a manual to an automated agentic system using Amazon Bedrock AgentCore.
Business Challenge
Platform engineering teams face a myriad of challenges in delivering seamless, self-service experiences for their internal customers at scale. TR’s Platform Engineering team was tasked with operational activities like database management, information security and risk management (ISRM), and infrastructure provisioning, among others. Unfortunately, manual processes and repeated coordination between teams led to unnecessary delays, hindering innovation.
Naveen Pollamreddi, Distinguished Engineer at TR, emphasizes the issue: "Our engineers were spending considerable time answering the same questions and executing identical processes across different teams. We needed to automate these interactions while maintaining our security and compliance standards."
Current State
The Platform Engineering team provides essential services to various product teams within TR, utilizing internal, home-grown solutions to build and run applications at scale on AWS. However, many processes relied heavily on human execution, leading to significant dependencies and delays.
These limitations manifested in three primary ways:
- Repetitive and Labor-Intensive Workflows: Engineers spent too much time on undifferentiated tasks, slowing down productivity.
- Longer Time to Value: Interdependencies within operational workflows resulted in drawn-out processes, lacking true autonomy.
- Resource and Cost Intensity: Manual execution consumed dedicated engineering resources, diverting them from more strategic innovation tasks.
Solution Overview
To address these challenges, TR’s Platform Engineering team turned to autonomous agentic solutions powered by Amazon Bedrock AgentCore. They focused on scalability, extensibility, and security, enabling non-technical users to quickly create and deploy AI-powered automation. The solution architecture allows business users to interact with specialized agents through basic natural language requests.
With AgentCore’s foundational infrastructure, TR gained the flexibility to innovate while ensuring enterprise-level security and reliability. The outcome? A self-service agentic platform engineering hub tailored to meet TR’s unique needs.
Key Components of the Solution
- Custom web portal for secure agent interactions.
- Central orchestrator agent (Aether) managing requests.
- Service-specific agents for specialized tasks, like AWS account provisioning and database patching.
- Human-in-the-loop validation service to ensure oversight on critical operations.
Solution Workflow
The journey to building the agentic AI system comprised several key steps:
-
Discovery and Architecture Planning: The team evaluated existing resources to design a comprehensive solution using AgentCore.
-
Core Development and Migration: They adopted a dual-track approach: migrating existing solutions to AgentCore while building a deployment engine (TRACK) for rapid agent creation.
-
System Enhancement and Deployment: Focused on refining functionalities and executing a team onboarding process for new agentic systems.
Orchestrator Agent and Framework
The orchestrator agent, aptly named Aether, retrieves context from the agent registry to determine appropriate actions. Using the LangGraph Framework, it facilitates secure communication while integrating with the AgentCore Memory service to maintain conversation context and learn from interactions.
Moreover, TR developed the TR-AgentCore-Kit (TRACK) framework, simplifying agent deployment and ensuring compliance with internal standards. The TRACK project streamlined the registration of service agents, facilitating seamless integration with Aether.
Outcome
By leveraging Amazon Bedrock AgentCore, TR transitioned to a self-service agentic system that automated complex workflows. Let’s explore the results:
Productivity and Efficiency
- Achieved a 15-fold increase in productivity through intelligent automation of routine tasks.
- Realized a 70% automation rate at launch, significantly reducing manual workloads.
Speed and Agility
- Accelerated product delivery by automating environment setup and operations.
- Empowered teams with self-service workflows.
Security and Compliance
- Enhanced security posture with automatic policy enforcement.
- Maintained human oversight with a validation service.
Cost and Resource Optimization
- Improved cost efficiency by optimizing infrastructure usage.
- Freed teams to focus on high-value tasks, minimizing operational toil.
Developer Experience
- Streamlined workflows resulted in improved developer satisfaction and standardization across teams.
Conclusion
The Aether project sets a replicable pattern for teams across Thomson Reuters, enhancing operational excellence through automation. By eliminating manual tasks, TR’s Platform Engineering team has positioned itself to drive innovation in the age of AI.
As Aether evolves, the hope is that this framework can realize broader adoption, assisting teams beyond Platform Engineering to surpass productivity standards.
Ready to transform your platform engineering operations? Explore AgentCore and its documentation to gain insights into enhancing your workflows with AI.
About the Authors:
- Naveen Pollamreddi is a Distinguished Engineer in Thomson Reuters, focusing on the Agentic AI strategy for Cloud Infrastructure services.
- Seth Krause is a Cloud Engineer on the Platform Engineering Compute team, specializing in implementing generative AI solutions for enhanced productivity.
- Pratip Bagchi is an Enterprise Solutions Architect at AWS, passionate about facilitating AI adoption for business transformation.
- Sandeep Singh is a Senior Generative AI Data Scientist at AWS, specializing in optimizing efficiency through AI and machine learning solutions.