AWS re:Invent 2025 - Migrate and Modernize Legacy Workflows to Intelligent Agents (PEX304)

Migrating and Modernizing Legacy Workflows to Intelligent Agents

Transforming Legacy Processes with Intelligent Agents

• The presentation discusses the importance of modernizing legacy business workflows using intelligent agents.
• It uses a story about an old post office in the jungle to illustrate the challenges of outdated processes and the benefits of adopting a more efficient, agent-based approach.
• The key message is that AWS partners can help customers identify opportunities to modernize their legacy workflows and transform them into autonomous, intelligent solutions.

The Shift to Business Process Transformation

• The business process services (BPS) market is undergoing a fundamental shift, with the potential to grow from $214 billion in 2014 to $289 billion by 2028.
• This landscape presents an opportunity for partners to leverage AI and intelligent agents to reimagine existing processes and create autonomous, intelligent solutions.
• Partners can bring their AWS expertise and domain knowledge to help customers modernize their workflows, moving from manual, inefficient processes to production-ready, agent-based solutions.

Modernizing a Legacy Order Management Workflow

• The presentation uses a use case of a premium grocery item distributor to demonstrate the transformation from a legacy, manual order management process to an agent-based solution.
• The legacy workflow involved multiple manual steps, poor customer experience, and efficiency/productivity issues, leading to risks of customer churn.
• The modernized, agent-based workflow reduced the number of steps from 12 to 8, automated key tasks, and provided real-time updates to customers, significantly improving the overall experience.

The Typical Agentic Journey

The presentation outlines the key steps in a typical agentic journey:

1. Assess: Evaluate the existing workflow from both a business and technical perspective.
2. Enable: Modernize the technical stack to be ready for integration with agents.
3. Design and Build: Design and implement the agent-based solution.
4. Scale: Deploy the agentic solution at scale, ensuring it is secure and reliable.

Design Considerations for Agentic Solutions

The presentation covers several important design considerations for building effective agentic solutions:

Orchestration Patterns

• Agent as a Tool: A single agent orchestrates the process and uses various tools to make autonomous decisions.
• Swarm Pattern: Multiple specialized agents work independently to solve complex problems.
• Graph Pattern: A deterministic workflow with dependencies and branching logic.

Reasoning and Intelligence

• Selecting the appropriate language model (LLM) based on task complexity, latency, and cost requirements.
• Leveraging domain-specific models for industry-specific use cases.
• Evaluating model performance and cost-effectiveness.

Context and State Management

• Implementing context management to minimize information overload and ensure relevance.
• Utilizing short-term and long-term memory to enable agents to remember and adapt to user interactions.

Observability

• Incorporating observability to trace agent workflows, monitor performance, and optimize efficiency.
• Leveraging tools like Amazon CloudWatch and third-party solutions for comprehensive observability.

Developing Agentic Tools and Integrations

• The presentation highlights the importance of providing agents with the right information and tools to interact with various data sources and applications.
• It introduces the Model Context Protocol (MCP) as a way to unify data integration and enable agents to access the information they need.
• Amazon Agent Core Gateway is discussed as a managed service that simplifies the deployment and management of MCP-based integrations.

Scaling Agentic Solutions with AWS

• The presentation outlines a three-layer architecture for scaling agentic solutions:
  1. Business Process Layer: Captures the knowledge and flows of the business processes.
  2. Enterprise Agentic Platform: Provides the governance, operations, and tools for agent-based solutions.
  3. Core Platform Services: Leverages AWS services to power the underlying infrastructure.
• This architecture allows for independent scaling and flexibility in modifying each layer.

Modernizing Data Collection with Intelligent Agents

• The presentation showcases a use case from the AWS partner Smart Communication, which used intelligent agents to modernize data collection in regulated industries.
• Smart IQ, a low-code enterprise form automation solution, integrated with agents to transform manual, paper-based processes into dynamic, conversational data collection.
• This resulted in significant improvements, such as reducing loan package turnaround time from 20 days to 5-7 minutes and decreasing "not in good order" rates from 67% to under 3%.

AWS Partner Enablement and Resources

• The presentation introduces the AWS "Process to Agent" (P2A) initiative, which aims to help partners and customers transform traditional business processes into agent-powered, AI-driven solutions.
• It highlights the Orchestrate AI platform from Capgemini as an example of a partner-built enterprise agentic platform that leverages AWS services.
• Partners are encouraged to learn more about the Strands agent, Amazon Bedrock agent, and other AWS resources to build and scale their agentic capabilities.

Key Takeaways

• Modernizing legacy workflows with intelligent agents can significantly improve efficiency, productivity, and customer experience.
• AWS partners can leverage their expertise to help customers identify opportunities, design agent-based solutions, and scale them securely on AWS.
• Careful consideration of orchestration patterns, reasoning and intelligence, context management, and observability is crucial for building effective agentic solutions.
• Integrating agents with the right tools and data sources, as well as leveraging AWS services and partner platforms, can enable scalable, enterprise-grade agentic deployments.
• The AWS "Process to Agent" initiative and partner enablement programs provide resources and support for partners to develop and monetize their agentic capabilities.