It support · Production

Thomson Reuters builds an agentic platform engineering hub with Amazon Bedrock AgentCore

The problem

Thomson Reuters' Platform Engineering team relied on semi-automated, manual processes for cloud infrastructure and operational tasks. Engineers repeatedly answered the same questions and executed identical workflows across multiple teams, creating delays and preventing innovation.

Workflow diagram · grounded in source
1
User request via web portal
trigger
“The portal authenticates users against TR's enterprise single sign-on (SSO) and provides access to agent flows based on user permissions”
2
Orchestrator routes to specialist agent
routing
“The orchestrator retrieves context from their agent registry to determine the appropriate agent for each situation”
3
Service agent executes operational task
ai_action
“Multiple service-specific agents handling specialized tasks such as AWS account provisioning and database patching”
4
Human-in-the-loop validation
human_review
“Aether Greenlight, a validation service that makes sure critical operations receive appropriate human oversight. This service extends beyond basic requester approval, so that team members outside the initial conversation can participate …”
5
Memory-backed context retention
feedback_loop
“Short-term memory maintains context within individual conversations, while long-term memory tracks user preferences and interaction patterns over time. This dual-memory approach allows the system to learn from past interactions and avoid…”
Reported outcome

Thomson Reuters achieved a 15-fold productivity gain and 70% automation rate at first launch, with autonomous agents handling complex operational workflows end-to-end around the clock.

Reported metrics
Productivity gain15-fold
Automation rate at first launch70%
Reported stack
Amazon Bedrock AgentCoreLangGraphAmazon DynamoDBAmazon API GatewayAmazon S3AgentCore MemoryAgentCore RuntimeAgentCore GatewayTRACKReact
Source
https://aws.amazon.com/blogs/machine-learning/how-thomson-reuters-built-an-agentic-platform-engineering-hub-with-amazon-bedrock-agentcore?tag=soumet-20
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Thomson Reuters achieved a 15-fold productivity gain and 70% automation rate at first launch, with autonomous agents handling complex operational workflows end-to-end around the clock.

What tools did this team use?

Amazon Bedrock AgentCore, LangGraph, Amazon DynamoDB, Amazon API Gateway, Amazon S3, AgentCore Memory, AgentCore Runtime, AgentCore Gateway, TRACK, React.

What results were reported?

Productivity gain: 15-fold; Automation rate at first launch: 70% (source-reported, not independently verified).

How is this it support AI workflow structured?

User request via web portal → Orchestrator routes to specialist agent → Service agent executes operational task → Human-in-the-loop validation → Memory-backed context retention.

Related it support cases