AWS, pour une nouvelle révolution de l'intelligence artificielle d'entreprise... l'essentiel réside dans "l'agent intelligent"

Artificial intelligence systems progress is highly regarded, and Amazon Web Services (AWS) is targeting the next phase of enterprise AI. During this year’s AWS re:Invent event, “agent artificial intelligence” emerged as the core architecture to replace traditional cloud strategies, shifting the discussion beyond mere model performance towards how to integrate AI into real-world business processes. Although this momentum has been forming for several years, industry evaluations consider that this event marks the official start of a structural overhaul of AI architectures and the entire cloud infrastructure.

AWS’s new CEO Matt Garman stated: “80% to 90% of the value of future enterprise AI will come from ‘agent’ architectures,” positioning the agent concept as the core axis of the new cloud era. He emphasized that through a comprehensive tech stack—from scalable custom semiconductors to new model series, learning-based architectures, and runtime operations—businesses can build valuable intelligence systems. AWS designed AI factories to accommodate the needs of a very small number of high-performance clients, which is also an extension of this strategy.

This direction is not merely a vision. AWS has proposed goals that go beyond AI assistance towards the evolution of “team member AI” capable of collaboratively handling overall organizational tasks. Swami Sivasubramanian, Vice President of the Agent AI division, explained: “Most AI systems are like interns, needing to learn from scratch every day; what we want is AI that can understand the work context over the long term.” Extending this vision, new storage services such as S3 Vector, capable of directly processing internal enterprise data, reducing latency, and improving accuracy, are also included.

Leading enterprises are also very enthusiastic about this transformation. Jerry Chen, Partner at Grelock Partners, pointed out: “Cloud architecture has begun evolving from traditional ‘cloud’ structures to ‘cloud + AI’,” emphasizing the need to prepare for a sovereign cloud environment driven by AI in the future. In this trend, AWS has also launched the Nova Forge platform, supporting cutting-edge model tuning across various industries, demonstrating a move to accelerate model training based on enterprise-specific data.

The agent-based architecture is transcending mere automation, spreading towards reimagining the overall customer experience. AWS Vice President Collin O’Brien used the “Just Walk Out” shopping tracking technology as an example: “A system that makes real-time decisions through visual reasoning in a fully non-identifiable state can meet diverse privacy protection requirements in the future.”

This shift is also evident among developer trends. Tools are moving from simple code assistance to integrated agent-type tools capable of handling everything from work planning to documentation and task execution. According to AWS Vice President Deepak Singh, the Kiro platform focuses on converting conversational commands into explicit specifications to quickly produce results-rich products.

The wave of transformation also affects the entire cloud ecosystem. Nick Johnston, Vice President at Salesforce, stated: “Customers want an open platform that avoids vendor lock-in and can activate data anywhere,” announcing the integration plan “Agentforce 360” with AWS. Oracle is also accelerating its multi-cloud strategy, and Snowflake is porting Cortex-based financial and security agents to AWS Bedrock Core, aiming to enhance intelligent analytics capabilities.

To ensure consistent AI performance across different environments, multi-cloud integration strategies are also accelerating. Companies are freely migrating critical task data between clouds to effectively harness the potential of evolving agent AI. User-centric customized insights are also being applied more rapidly within such hybrid architectures.

Cloud infrastructure is being redefined. The system itself is shifting towards an AI-first architecture, where agent AI is no longer just a tool but is positioned as a “digital colleague.” Led by AWS and its partners, this trend is likely to establish a new benchmark for enterprise technological strategy.