AWS re:Invent 2025 - Building enterprise-grade Agentic AI factory (PEX302)

AWS re:Invent 2025 - Building enterprise-grade Agentic AI factory (PEX302)

Overview and Industry Trends

• Enterprises are doubling down on a "new paradigm of AI" - generative AI and agentic AI
• Analyst research shows the AI agent market will grow from $8 billion today to $98 billion by 2032, a 45% annual growth rate
• VCs invested 50% of all funding in the first 3 quarters of 2025 into AI-first organizations
• Gartner predicts 40% of enterprise software will have embedded AI agents by next year

The Evolution from Generative to Agentic AI

• Generative AI started with chatbots and language models that responded to prompts
• Simpler agents emerged that could break down tasks, make API calls, and take actions
• Multi-agent agentic systems can now handle complex, multi-step business processes
• Gartner forecasts 1.3 billion agents will be in use across industries by 2028

Challenges with Scaling Agentic AI

• Lack of technical foundations to deploy and govern agents at scale
• Siloed and inaccessible data needed for agents to make informed decisions
• Security, trust, and governance challenges in regulated industries
• Leadership misalignment and talent gaps in critical thinking skills
• Agents getting stuck in proof-of-concept purgatory without clear ROI
• Lack of repeatable patterns and blueprints to scale agent deployments

The Agentic AI Factory Approach

• Provides a repeatable framework and platform to take agentic AI use cases from concept to production
• Includes four key processes: Assess, Plan, Implement, and Iterate/Learn
• Integrates with external systems, data sources, and identity providers
• Supports agent blueprints that encapsulate business workflows
• Built-in security and governance controls across six pillars:
  1. Identity and access
  2. Data security
  3. Model security
  4. Operational safety
  5. Governance
  6. Runtime controls

Technical Architecture and Implementation

• Standardized patterns and prescriptive guidance for building agentic systems
• Leverages AWS best practices, partner ecosystem knowledge, and customer constraints
• Knowledge base and shared memory layer to inform agent reasoning and planning
• Automated assessment, planning, implementation, and iterative feedback loops
• Centralized UI for managing the factory, agents, and agentic workflows
• Agentic orchestration using a supervisor pattern for emergent behavior
• Observability and analysis to monitor agent health and refine the system

Customer Case Study: Slalom's Agentic Factory Implementation

• Fragmented AI pilots, readiness gaps, and wasted investments at a healthcare customer
• Used the factory approach to:
  • Align leadership and catalog current state
  • Assess talent and AI readiness gaps
  • Plan high-value use cases with defined ROI
  • Build a center of excellence and governance model
  • Rapidly deploy and validate agents in production
• Reduced 50 initial use cases to 14 high-priority, data-ready, and ROI-defined projects
• Enabled rapid business validation and scaled agent deployments

Getting Started with the Agentic AI Factory

• AWS is launching the Agentic AI Factory assets, including:
  • Early adoption program for guided onboarding and co-development
  • Agentic AI Platform starter kit available on GitHub
  • Playbooks, agent blueprints, and prescriptive guidance
• Roadmap includes:
  • Full operational experience within the factory
  • Recursive learning and iteration capabilities
  • Compliance automation and controls
  • Agent cost tracking and AWS Marketplace integration

Key Takeaways

• Enterprises are rapidly adopting agentic AI, but scaling these systems poses significant challenges
• The Agentic AI Factory provides a repeatable framework and platform to deploy agentic AI at scale
• The factory integrates security, governance, and compliance controls from the ground up
• Technical architecture includes standardized patterns, automated processes, and observability
• Customers can leverage the factory to align stakeholders, assess readiness, plan high-value use cases, and rapidly deploy agentic systems
• AWS is providing the foundational assets, but customers can build their own differentiated factories on top