AWS re:Invent 2025 - Automate insights and drive innovation with cloud and AI solutions (IND384)

Automating Insights and Driving Innovation with Cloud and AI Solutions

Addressing the Challenges of Rapid Change in Retail and CPG

• Retail and CPG companies are facing unprecedented change, with higher consumer expectations and tighter margins
• The pace of technology change is accelerating, forcing companies to maintain an agile approach
• Traditional IT teams are struggling to keep up with the speed of business and technology innovation

Manderly's Cloud Engineering Journey

Building a Self-Service Cloud Platform

• Manderly started its cloud journey in 2018 with a lift-and-shift approach, relying on managed service providers
• In 2021, Manderly built a dedicated cloud engineering team to create a new cloud platform from scratch
• Key principles:
  • Self-service for developers
  • Secure by default
  • Full automation and infrastructure as code
  • Micro-segmentation and cost transparency

Implementing GitOps and Policy Enforcement

• All cloud resources are provisioned through automated, infrastructure-as-code pipelines
• Static branches in Git repositories are mapped to AWS environments, enforcing GitOps principles
• OPA (Open Policy Agent) is used to enforce security and best practice policies across all deployments
• This enables rapid updates to standards and controls, with changes propagating across the entire environment

Achieving Cost Transparency and Efficiency

• Micro-segmentation of AWS accounts allows for granular cost visibility and attribution
• Developers and product owners can easily access real-time cost data for their applications
• The platform enables Manderly to drive cost optimization and efficiency across the organization

Next Steps and Future Plans

• Manderly is transitioning to a product-centric, platform-based organizational model
• Completing the migration of remaining on-premises systems to the cloud
• Refactoring and re-engineering applications to leverage more serverless and cloud-native services
• Exploring further integration and synergies with AWS as a strategic partner

Gringanger's PHOPS Transformation with Generative AI

The Challenge: Scaling PHOPS Insights to Enterprise Stakeholders

• Critical cloud cost and usage insights were trapped in complex systems, with stakeholders unable to interpret the data
• Manual communication and reporting could not keep up with the pace of enterprise cloud adoption

Initial Approach: Executive Summaries Delivered to Inboxes

• Targeted product and technical domain leaders as the highest-value audience
• Delivered concise, executive-friendly cloud cost and usage summaries to their inboxes
• Manual creation of these summaries proved unsustainable at scale

Evaluating Solutions: Manual, Programmatic, and Generative AI

• Programmatic solutions lacked the nuance and customization of the manual summaries
• Generative AI emerged as a promising solution to meet the scale requirement while maintaining quality

Implementing Generative AI with Agentic Workflows

• Developed a multi-agent architecture using Agent Core and Amazon Bedrock
• One agent performs the detailed cloud cost analysis, another synthesizes the insights into executive summaries
• Leveraged custom MCP (Model Composition Protocol) tools to fetch data and ensure consistency
• Deployed the solution in a headless, enterprise-ready architecture

Benefits and Future Expansion

• Thousands of specialist hours saved by automating the summary generation process
• Ability to scale PHOPS insights to a broader set of stakeholders, including product owners, finance, and executives
• Plans to extend the agentic workflow to other domains, such as right-sizing recommendations and integration with developer tools

Key Takeaways

• IT teams must evolve from reactive service providers to strategic partners that drive business innovation
• Automation, infrastructure as code, and policy enforcement are critical to enabling agile cloud adoption
• Generative AI can be leveraged to scale specialized insights and analysis to enterprise stakeholders
• Agentic workflows and custom MCP tools can help overcome the limitations of traditional programmatic solutions
• Combining cloud engineering best practices with generative AI unlocks new possibilities for IT-business collaboration and transformation