CrowdStrike used its RSA Conference 2026 presence on March 25 to launch Charlotte AI AgentWorks, a no-code development platform that allows security teams to build, orchestrate, and scale custom AI agents within the Falcon platform. The announcement signals a strategic pivot from offering AI as a vendor-controlled copilot toward enabling organizations to construct their own autonomous security workflows with enterprise-grade guardrails.
The launch partners tell the story of where the market is heading. Accenture, Amazon Web Services, Anthropic, Deloitte, Kroll, NVIDIA, OpenAI, Salesforce, and Telefonica Tech all joined as inaugural ecosystem contributors. The breadth of that coalition, spanning consultancies, cloud hyperscalers, frontier model providers, and managed security firms, suggests that the agentic SOC is no longer a concept paper. It is an integration surface that CrowdStrike intends to own.
Architecture and Technical Approach
AgentWorks sits on top of the existing Falcon platform and exposes a visual builder where analysts and security engineers can define agent logic, assign data sources, configure decision boundaries, and connect output actions without writing code. The platform integrates directly with frontier AI models including Anthropic Claude, NVIDIA Nemotron, and OpenAI GPT, while supporting infrastructure services like Amazon Bedrock and Amazon SageMaker for model hosting and fine-tuning.
The key architectural decision is bounded autonomy. Every agent built through AgentWorks operates within explicit permission scopes, and CrowdStrike has layered its AI Detection and Response (AIDR) capability as a supervisory system that monitors agent behavior for drift, hallucination, or actions that exceed defined policy boundaries. Charlotte Agentic SOAR handles multi-agent coordination, routing tasks between specialized agents and human analysts based on confidence scores and escalation thresholds.
Why This Matters for Security Operations
The security operations center has spent the past three years absorbing copilot functionality, starting with natural language queries over telemetry data and graduating to AI-generated investigation summaries. AgentWorks represents the next step: moving from AI that assists an analyst to AI that executes repeatable workflows autonomously while the analyst supervises at a strategic level.
The practical implication is that organizations with mature detection engineering teams can now encode their institutional knowledge into persistent agents that run continuously, rather than depending on analysts to remember and manually execute the same investigative playbooks across every shift change. A Tier-3 analyst’s methodology for triaging a particular alert class can become an agent that handles the repetitive portion while escalating only the decisions that require human judgment.
Partner Ecosystem Strategy
CrowdStrike Chief Business Officer Daniel Bernard stated that AgentWorks enables every Falcon user to build their own agentic security workforce. The framing is deliberate. Rather than positioning AI agents as a premium feature controlled by CrowdStrike’s own development team, the company is creating a marketplace dynamic where partners and customers generate value on the platform.
Accenture plans to deliver customized agentic SOC implementations through the ecosystem. Kroll will operationalize AgentWorks within its managed detection and response service. Telefonica Tech described the integration as accelerating its transition to an AI-Native SOC. Each partner relationship extends the platform’s reach into different market segments while creating switching costs that compound over time.
Competitive Positioning
The announcement positions CrowdStrike directly against Palo Alto Networks’ Cortex XSIAM and its recently integrated Protect AI capabilities, as well as Microsoft’s Security Copilot ecosystem. The differentiator CrowdStrike is claiming is openness at the model layer combined with security-first governance. Where competitors tend to lock customers into a single AI provider, AgentWorks supports multiple frontier models and allows organizations to select based on task characteristics, cost, or data residency requirements.
The simultaneous launch of Agentic MDR through Falcon Complete extends the same technology into CrowdStrike’s managed services tier, ensuring that mid-market organizations without dedicated detection engineering teams can access agentic automation through a managed delivery model.
Implications for Enterprise Security Teams
Three operational considerations emerge from the launch. First, organizations evaluating agentic AI for security operations now face a build-versus-buy decision that did not exist six months ago. AgentWorks provides the build path with vendor-supported guardrails, but building effective agents still requires deep domain expertise about what decisions can safely be automated.
Second, the multi-model architecture means that security teams need to develop evaluation frameworks for which AI models perform best against specific security tasks. A triage agent may require different model characteristics than a threat hunting agent or a compliance reporting agent.
Third, the AIDR supervisory layer introduces a new category of operational concern. Organizations must now monitor not just their infrastructure for threats, but their defensive AI agents for behavioral drift, policy violations, or adversarial manipulation by attackers who understand that compromising an autonomous agent grants persistent and potentially invisible access to security operations.
Source: CrowdStrike Press Release.
