This analysis is part of Amazon’s AI Business Model Pivot, a deep dive by The Business Engineer.

AgentCore is AWS’s answer to the critical enterprise question: how do you deploy, govern, and scale AI agents in production? Internally described as “Kubernetes for agents,” it provides the governance and control plane that makes agentic AI viable for regulated industries.
The Four Pillars
Runtime: Framework-agnostic deployment. Bring LangChain, CrewAI, AutoGen, or any custom framework—AgentCore deploys them all without vendor lock-in. This directly contrasts with Microsoft’s M365-integrated approach.
Policy (Cedar): Cedar-based guardrails enforce agent boundaries and permissions. Agents can only access what they’re authorized to. This is critical for financial services, healthcare, and government deployments.
Memory (Episodic): Persistent context across sessions. Agents remember previous interactions, decisions, and outcomes—enabling multi-day autonomous workflows rather than stateless single-shot executions.
Evaluations: Pre-production testing and validation. Before any agent goes live, it’s tested against defined criteria—reducing risk of autonomous AI failures in production environments.
Why Governance Is Central
Agent boundaries, permissions, and auditability are first-class design constraints, not afterthoughts. This makes AWS agents viable in GDPR, HIPAA, SOX, SOC 2, and FedRAMP environments. Governance-first design is what differentiates AWS from consumer-first agent platforms.
The Lock-In Strategy
AgentCore is led by David Richardson, who returned from Stripe specifically for this role. The control plane creates deep platform lock-in: once enterprises deploy agent governance on AgentCore, switching costs become prohibitive. This is AWS’s real moat in the agent economy.







