TLDR: Amazon Web Services (AWS) is significantly enhancing its AI agents by integrating predictive machine learning (ML) models via Amazon SageMaker AI and the Model Context Protocol (MCP). This strategic move aims to empower AI agents to make more data-driven business decisions, combining the strengths of traditional ML for precise predictions with generative AI for creative tasks. The initiative, highlighted at the AWS Summit New York 2025, underscores AWS’s commitment to advancing agentic AI with substantial investments and new developer tools.
Amazon Web Services (AWS) is pushing the boundaries of artificial intelligence by integrating predictive machine learning (ML) models directly into its AI agents, leveraging Amazon SageMaker AI and the innovative Model Context Protocol (MCP). This development, detailed in a recent AWS Machine Learning Blog post by Saptarshi Banerjee and Davide Gallitelli, is set to revolutionize how AI agents operate by enabling them to make sophisticated, data-driven business decisions through ML predictions.
AI agents, defined as Large Language Model (LLM)-powered applications, function as autonomous systems that observe environments, plan actions, and execute tasks with minimal human intervention. They integrate reasoning, memory, and tool use to tackle complex, multi-step problems by dynamically creating and revising plans, interacting with external systems, and learning from past interactions.
While generative AI has emerged as a transformative force for creative tasks like content generation and product design, traditional ML models remain indispensable for specific predictive applications. These include precise sales forecasting, which relies on algorithms such as random forests, gradient boosting machines (like XGBoost), ARIMA models, LSTM networks, and linear regression. Traditional ML also excels in customer segmentation and churn prediction using methods like K-means and hierarchical clustering. AWS emphasizes that combining both generative AI and traditional ML approaches yields optimal results, delivering accurate predictions while maintaining cost efficiency.
Central to this enhancement is the Model Context Protocol (MCP), an open protocol designed to standardize how applications provide context to LLMs. MCP is rapidly becoming an industry standard for ‘tool calling,’ allowing LLMs to decide which actions to take, perform them, and evaluate their results through a standardized interface. AWS is further supporting developers with new MCP resources, including a local AWS API MCP server with comprehensive knowledge of the AWS API surface and an AWS Knowledge MCP server offering up-to-date AWS documentation.
Developers can build these advanced AI agents using the open-source Strands Agents SDK, which facilitates rapid development from simple assistants to complex workflows. The solution offers flexible integration paths, including direct endpoint access via tool annotations and MCP-based integration, allowing developers to choose the most suitable approach for their specific use cases.
At the AWS Summit New York 2025, Swami Sivasubramanian, AWS VP for Agentic AI, highlighted the profound impact of AI agents, describing them as a ‘tectonic change’ that will reshape software development and human-software interaction. AWS is backing this vision with a significant investment of an additional $100 million into its Generative AI Innovation Center, aimed at accelerating the development and deployment of agentic AI systems globally.
Also Read:
- Omdia Report Highlights AWS Agentic AI’s Role in Global Business Expansion
- AI Agents Realize Semantic Web’s Vision for Intelligent Data Interconnection
Further innovations announced include Amazon Bedrock AgentCore, a comprehensive suite of seven core services designed to deploy and operate secure AI agents at enterprise scale. Additionally, the AWS Marketplace now features ‘AI Agents and Tools,’ providing a streamlined platform for customers to discover, purchase, deploy, and manage AI agents and tools from leading providers. Enhancements to Amazon Nova models, including customization capabilities with SageMaker HyperPod and improved security via the Nova Act SDK, also contribute to AWS’s robust AI ecosystem.
