Overview

Engagement overview

A leading Australian Wagering & Gaming Company sought to scale AI agent adoption across multiple business units while maintaining consistent security, governance, cost control, and operational visibility. As more teams began building AI solutions independently, separate infrastructure, model access, and observability for every agent became difficult to manage.

TO THE NEW worked with the organization to design a centralized AI Core Platform on AWS, using Amazon Bedrock AgentCore. The platform introduced self-service onboarding, governed tool integrations, AI guardrails, observalility and FinOps controls as a common foundation for enterprise AI agents. 

Our client

Wagering & Gaming
 Australia

The client is a leading Australian Wagering & Gaming Company operating large-scale digital platforms across multiple business functions. As generative and agentic AI adoption grew, teams across engineering, operations, FinOps, and productivity functions began exploring independent use cases, creating a need for a scalable platform that avoided separate governance and infrastructure models for every solution.

Wagering & Gaming AI Platform

Business objective

The engagement focused on scaling AI agent adoption across the client's business units while maintaining consistent security, governance, and cost control.

01

Scale AI across multiple teams without requiring new infrastructure and governance design for each agent

02

Standardize foundation model access to avoid model sprawl while maintaining controlled access for teams

03

Govern enterprise tool access so agents could securely reach APIs, repositories, databases, and platforms

04

Automate security and compliance controls instead of relying on manual implementation by each team

05

Improve developer experience through self-service onboarding rather than manual infrastructure setup

06

Evaluate agent responses, tool usage, safety, and quality before changes were promoted

07

Improve AI cost visibility by attributing model consumption to the correct team and cost centre

08

Apply consistent governance standards across agents as enterprise AI adoption expanded

Business solution

The platform engineering effort focused on building a governed AI foundation, standardizing model and tool access, and strengthening observability for enterprise-scale agent adoption.

  • Established a centralized AI platform foundation using Amazon Bedrock AgentCore for reusable AI capabilities

     

  • Implemented AgentCore Runtime and Harness patterns to support custom and managed orchestration-based agents

     

  • Introduced declarative agent onboarding through a self service portal with approvals or a single manifest capturing ownership

  • Automated infrastructure provisioning through CI/CD and infrastructure as code to reduce repeated setup
     

  • Implemented policy-as-code controls to validate approved models, regions, and platform policies before deployment

  • Established centralized model governance with controlled access to approved Bedrock foundation models

  • Implemented governed tool and MCP integrations using AgentCore Gateway to replace point-to-point integrations

  • Implemented identity and least-privilege access controls so agents accessed only required tools and data
     

  • Integrated AI guardrails and data protection controls to manage sensitive information and content risks

  • Automated agent evaluation capabilities were implemented to assess quality, safety, and regression before promotion

  • Standardized observability using AgentCore capabilities and OpenTelemetry across agent, model, tool, and system flows

  • Embedded FinOps and lifecycle governance to support cost attribution, versioning, and controlled rollback

Business impact

The initiative improved agent governance, operational visibility, and provisioning speed while supporting enterprise-scale growth in AI adoption.

50+ governed AI agents supported through a common platform with standardized controls

5+ business units enabled to independently build AI solutions on the same foundation

10+ enterprise tools integrated through reusable, secure agent connections

<5 minute agent provisioning achieved through automated onboarding instead of manual setup 

Automated agent evaluations enabled for quality, safety, and regression checks before promotion

Single governed AI foundation established for model governance, security, and lifecycle management