Agentic Platform Engineering Hub for Cloud Operations Automation

thomson_reuters · Tech · 2026 · source Agentic: yes · promotional bias: high · source-backed · extracted with prompt 0.5.2 (schema v2)

Generated from extraction.json — do not edit by hand. Every claim cites verbatim source evidence; stated means the quote directly asserts the claim, implied means it supports it indirectly.

Problem

Domain: Platform engineering

Thomson Reuters platform engineering needed to automate repetitive cloud operations while maintaining secure self-service for internal teams.

Constraints:

Multiple operational domains
Security and compliance
Non-technical users
Repetitive engineer work

Architecture

Thomson Reuters built Aether, an AgentCore-backed platform-engineering hub. A React web portal authenticates users with enterprise SSO and exposes permitted agent flows. A LangGraph-based Aether orchestrator retrieves context from a custom A2A registry, routes natural-language requests to specialized agents for account provisioning, database patching, networking, and architecture review, and uses AgentCore Runtime/Gateway for tool execution. AgentCore Memory keeps conversation/user context while the service remains stateless, TRACK standardizes agent deployment and registry onboarding, and Aether Greenlight adds human validation and audit trails for sensitive operations.

Diagram confidence: explicit (topology described by the source).

flowchart LR
  custom-web-portal[Custom web portal] --> aether-orchestrator[Aether orchestrator]
  aether-orchestrator --> a2a-registry[A2A registry]
  aether-orchestrator --> service-specific-agents[Service-specific agents]
  aether-orchestrator --> agentcore-memory[AgentCore Memory]
  a2a-registry --> service-specific-agents
  service-specific-agents --> agentcore-runtime-and-gateway[AgentCore Runtime and Gateway]
  service-specific-agents --> aether-greenlight[Aether Greenlight]
  track-deployment-engine[TRACK deployment engine] --> a2a-registry
  track-deployment-engine --> agentcore-runtime-and-gateway

Drag to pan · scroll or use +/− to zoom · reconstructed from the source's described component topology

Components

Component	Role	Interacts with
`custom-web-portal`	React/S3 user interface for secure agent interactions, enterprise SSO, and permission-based access to agent flows.	`aether-orchestrator`
`aether-orchestrator`	LangGraph orchestrator that reads registry context, routes requests, and makes secure tool calls to the appropriate service agent.	`custom-web-portal`, `a2a-registry`, `service-specific-agents`, `agentcore-memory`
`a2a-registry`	Custom DynamoDB/API Gateway registry for agent discovery, cross-account calls, version history, auditability, and production validation.	`aether-orchestrator`, `service-specific-agents`, `track-deployment-engine`
`service-specific-agents`	Domain agents for cloud account provisioning, database patching, network configuration, and architecture review.	`aether-orchestrator`, `agentcore-runtime-and-gateway`, `aether-greenlight`
`agentcore-runtime-and-gateway`	Managed AgentCore infrastructure for runtime, tool management, gateway connectivity, deployment, and operation.	`service-specific-agents`, `track-deployment-engine`
`agentcore-memory`	Conversation- and user-level memory that preserves short-term context and long-term preferences while the orchestrator remains stateless.	`aether-orchestrator`
`aether-greenlight`	Human-in-the-loop validation service with approvals, outside-the-conversation review, and audit trails for critical operations.	`service-specific-agents`
`track-deployment-engine`	TR-AgentCore-Kit deployment framework that packages baseline agent setup, runtime/gateway connection, and registry registration for service teams.	`agentcore-runtime-and-gateway`, `a2a-registry`

Data flow

A user enters a request through the secure web portal, which controls access through enterprise SSO and permissions. The Aether orchestrator reads registry context to select the right service agent, then makes a tool call populated from the registry to communicate securely with the endpoint. Specialized agents run through AgentCore Runtime/Gateway for tasks like cloud account provisioning, database patching, networking, or architecture review. For critical operations, Aether Greenlight brings additional approvers into the flow and records approvals and actions for compliance.

4 supporting quotes

“The portal authenticates users against TR’s enterprise single sign-on (SSO) and provides access to agent flows based on user permissions.”article, stated

“The orchestrator retrieves context from their agent registry to determine the appropriate agent for each situation.”article, stated

“When an agent’s actions are required, the orchestrator makes a tool call that programmatically populates data from the registry”article, stated

“The system includes Aether Greenlight, a validation service that makes sure critical operations receive appropriate human oversight.”article, stated

Why this architecture

The system is organized as a hub-and-specialist architecture because TR needed self-service automation across many operational domains while keeping enterprise security, reliability, and compliance. AgentCore supplies managed agent infrastructure, TRACK standardizes how service teams deploy and register agents, and Greenlight preserves human oversight for sensitive operations.

4 supporting quotes

“TR chose Amazon Bedrock AgentCore because it provides the complete foundational infrastructure needed to build, deploy, and operate enterprise-grade AI agents at scale without having to build that infrastructure from scratch.”article, stated

“The Platform Engineering team gained the flexibility to innovate with their preferred frameworks while designing their autonomous agents operate with enterprise-level security, reliability, and scalability”article, stated

“TRACK maintains a seamless flow for developers by offering deployment capabilities to AWS and registration to the custom-built services in one package.”article, stated

“This governance model facilitates conformance with TR’s ISRM standards while providing flexibility for future expansion.”article, stated

Models used: Amazon Bedrock AgentCore · LangGraph

Scale

TR's platform-engineering unit is geographically distributed and oversees service availability.
“This blog post explains how TR’s Platform Engineering team, a geographically distributed unit overseeing TR’s service availability”article, stated
The team supports multiple lines of business with cloud infrastructure and enablement services.
“the Platform Engineering team supports multiple lines of business by providing essential cloud infrastructure and enablement services”article, stated
The launch reported a 15-fold productivity gain and 70% automation rate.
“Productivity and efficiency * 15-fold productivity gain through intelligent automation of routine tasks”article, stated
“Productivity and efficiency * 15-fold productivity gain through intelligent automation of routine tasks * 70% automation rate achieved at first launch, dramatically reducing manual workload”article, stated

Alternatives considered (roads not taken)

Continue semi-automated, human-executed operational processes

Rejected because: They were repetitive, slow to deliver value, and consumed engineering resources across coordination, execution, and validation.

“Current processes are semi-automated and are:- * Repetitive and labor intensive – Because of the nature of the workflows and multi-team engagement model, these operational processes tend to be labor intensive and repetitive.”article, stated

“Manual execution requires dedicated engineering resources whose time could be better spent on innovation rather than repetitive tasks.”article, stated

Build enterprise agent infrastructure from scratch

Rejected because: AgentCore supplied managed infrastructure to build, deploy, and operate enterprise-grade agents at scale.

“complete foundational infrastructure needed to build, deploy, and operate enterprise-grade AI agents at scale without having to build that infrastructure from scratch”article, stated

Let each service team wire runtime, gateway, and registry integration separately

Rejected because: TRACK packages AgentCore connection, tool management, gateway connectivity, baseline setup, deployment, and registration so teams can focus on business logic.

“so developers can focus on implementing business logic rather than dealing with infrastructure concerns. AgentCore Gateway provided a straightforward and more secure way for developers to build, deploy, discover, and connect to tools at scale.”article, stated

Patterns

`multi-agent-orchestration` (platform-ops)

A central orchestrator routes requests across service-specific agents.

“A central orchestrator agent that routes requests and manages interactions 3.”article, stated

`human-in-the-loop-gate` (governance)

The platform maintains security and compliance with validation in operational workflows.

“The solution delivered a 15-fold productivity gain, achieved 70% automation rate at launch, and freed engineering teams from repetitive tasks to focus on higher-value innovation work while maintaining security and compliance standards through human-in-the-loop validation.”summary, stated

Tradeoffs

Keep human validation for sensitive platform operations.

Chose Aether Greenlight approvals and audit trails over Fully autonomous execution for critical cloud provisioning and database patching changes.

Rationale: Thomson Reuters needed automation while preserving compliance evidence and oversight for sensitive operations.

“The system includes Aether Greenlight, a validation service that makes sure critical operations receive appropriate human oversight. This service extends beyond basic requester approval”article, stated

How much agent infrastructure to build internally.

Chose Amazon Bedrock AgentCore as managed foundational infrastructure over building the enterprise runtime, deployment, and operations layer from scratch.

Rationale: TR wanted framework flexibility and control while avoiding low-level infrastructure complexity for production-grade agents.

“complete foundational infrastructure needed to build, deploy, and operate enterprise-grade AI agents at scale without having to build that infrastructure from scratch”article, stated

How service teams should onboard new agents into Aether.

Chose TRACK plus a custom A2A registry with validation and audit history over per-team bespoke deployment and registration paths.

Rationale: A shared deployment kit and registry keep onboarding repeatable while preserving compliance and auditability.

“TRACK maintains a seamless flow for developers by offering deployment capabilities to AWS and registration to the custom-built services in one package.”article, stated

“The A2A registry maintains a comprehensive history of agent versions for auditing purposes and requires human validation before allowing new agents into the production environment.”article, stated

Gotchas (what bit them)

Manual operations consumed engineering time on repetitive requests.

Root cause: Many workflows depended on humans executing predefined steps.

Resolution: TR automated service-domain workflows with specialized agents.

“* Repetitive and labor intensive – Because of the nature of the workflows and multi-team engagement model, these operational processes tend to be labor intensive and repetitive. The Platform Engineering team spent a lot of time doing work that is undifferentiated heavy lifting.”article, stated

Evaluation

TR reports productivity gain, automation rate, and engineer time freed.

Metrics: productivity gain · automation rate

Outcomes

The summary reports 15x productivity gain and 70% automation at launch.

Vendor framing is high because this is an AWS customer story.