23 December 2025

At the AWS re:Invent event, announcements are plentiful and attention is scarce. Yet some stories only reveal their full significance once the press releases fade and the architecture choices are explained in plain language.

That was the case with SentinelOne’s expanded partnership with Amazon Web Services, unveiled in early December. Beyond the headlines — new marketplace listings, deeper integrations, and bold claims about autonomous AI — the announcement signals a shift in how security vendors expect enterprises to operationalise AI inside AWS.

In an exclusive interview with MoveTheNeedle.news, SentinelOne Chief Product Officer Ana Pinczuk explains how the company defines “agentic AI,” what customers are actually doing with it today, and why data discipline — not bigger models — will determine whether autonomous security systems succeed.

The expansion spans four tightly connected elements: the Purple AI MCP Server on AWS Marketplace; the Observo AI data pipeline on the AWS GenAI Marketplace; a new integration streaming AWS Security Hub findings into Singularity AI SIEM; and bidirectional data flows with Amazon CloudWatch, aligned with the OCSF open data standard. Together, they form what SentinelOne sees as the practical foundation for AI-driven security at enterprise scale.

From AI insight to AI action

“Agentic AI” has quickly become a buzzword across the security industry. In SentinelOne’s view, the term only has meaning if it translates into action.

“We define agentic AI as AI that can autonomously act on security insights — making decisions, orchestrating workflows and executing responses across endpoints, cloud workloads, identities and AI applications.”

Pinczuk’s definition draws a clear boundary. Traditional security tools — even those branded as AI-powered — typically stop at analysis or recommendation. Agentic systems are expected to go further: to decide and act, without waiting for human approval at every step.

That shift raises immediate questions about trust, governance, and scale. According to Pinczuk, those concerns can only be addressed if autonomous systems operate with full, unified context — and if their actions remain explainable and auditable.

Purple AI MCP Server: giving AI agents real security context

A central piece of SentinelOne’s AWS expansion is the Purple AI MCP Server, now available via the AWS GenAI Marketplace. In the exclusive interview, Pinczuk describes it as a way to embed SentinelOne’s security intelligence directly into modern AI agent frameworks.

“Within Amazon Bedrock AgentCore, agentic AI serves as a foundation for integrating SentinelOne’s intelligence directly with AI-driven operational environments. Through tools like the Purple AI MCP Server, now available on AWS GenAI Marketplace, organizations can build custom AI agents using frameworks such as Amazon Bedrock, OpenAI AgentKit, or Google’s Agent Development Kit, giving these agents the full context of SentinelOne’s Singularity platform to make informed, automated decisions across complex, multi-cloud environments.”

The emphasis on context is deliberate. AI agents operating without access to live security telemetry are limited to theoretical reasoning. By contrast, agents connected to Singularity can reason over endpoint data, cloud workload signals, identity activity, and threat intelligence in real time.

A useful analogy is the difference between a consultant working from reports and an analyst embedded in the SOC with access to every system. The MCP Server is designed to move AI agents closer to the latter — while still operating within enterprise controls.

Observo: solving the data economics of AI security

While Purple AI focuses on autonomy, SentinelOne’s Observo acquisition addresses a more fundamental constraint: cost.

In the interview, Pinczuk repeatedly returns to the problem of data volume. Autonomous systems depend on large amounts of telemetry, but ingesting everything is neither affordable nor operationally useful.

“AI-driven observability and cost optimization: Organizations are filtering high-volume, low-value data before ingestion, prioritizing only actionable insights to reduce operational cost and improve incident response times.”

Observo’s AI-driven data pipeline, now available on the AWS GenAI Marketplace, is positioned as a way to reduce security and observability costs by more than 50 percent while filtering out up to 80 percent of data with no analytical value.

This may be the least flashy part of the announcement — but arguably the most important. Without aggressive data reduction, agentic AI risks becoming an academic exercise rather than a deployable system.

Bridging cloud-native alerts and enterprise response

Another element of the AWS expansion focuses on correlation rather than autonomy. SentinelOne now enables prioritized AWS Security Hub findings to stream directly into Singularity AI SIEM, where they can be correlated with endpoint, identity, and AI telemetry.

In her responses, Pinczuk frames this as a unification problem.

“The expanded integration is all about unifying security and operational data across cloud and hybrid environments. It makes it easier for teams to correlate signals, detect threats faster and automate responses. Flexible, bidirectional data flows reduce latency, control costs and help manage complex environments at scale. Essentially, it’s about making AI-driven security practical, efficient and actionable across the enterprise.”

For enterprises running large AWS estates, this integration addresses a familiar pain point: cloud alerts often live in isolation, slowing investigations and limiting automation. Correlation across surfaces is a prerequisite for meaningful autonomy.

CloudWatch and bidirectional visibility

The Amazon CloudWatch integration reinforces that same logic from an operational perspective. Bidirectional data flows between CloudWatch and the Singularity platform allow security and operational signals to inform each other, using the OCSF open data standard to reduce friction.

Pinczuk emphasises that these capabilities are designed to fit into existing workflows rather than disrupt them.

“AWS customers will experience more seamless access to security and operational data. Expanded integrations allow findings and metrics to flow directly into their existing environments, where alerts can be prioritized and correlated automatically. This reduces manual effort, accelerates investigations and makes AI-driven security insights more practical and actionable without forcing teams to change how they work.”

The subtext is clear: agentic AI only works if it feels invisible to the teams relying on it.

What customers are actually doing with agentic AI

Asked about conversations at re:Invent, Pinczuk points to concrete — and already emerging — use cases rather than speculative future scenarios.

“Automated threat detection and response across cloud and hybrid environments: Enterprises are using agentic AI to correlate signals from multiple sources — endpoints, identities and cloud workloads — to accelerate investigations and reduce manual intervention.”

“Custom AI agent workflows: Developers and service providers are experimenting with autonomous agents that can execute operational or security tasks based on unified telemetry, enabling tailored workflows that scale across teams.”

“Holistic defense across multiple surfaces: Customers are recognizing that attacks now span endpoints, cloud and AI systems themselves. Employee use of GenAI tools, homegrown AI applications and emerging agentic AI workflows introduce new attack surfaces to be monitored. This is driving use cases where autonomous agents monitor and respond across hybrid and multi-cloud environments rather than single-tool silos.”

Across all three, the common denominator is scope. Agentic AI becomes valuable when it crosses boundaries — between tools, environments, and organisational silos.

Infrastructure choices shape autonomy

The interview also touches on the less visible layers beneath AI adoption. Pinczuk notes growing enterprise interest in AWS Graviton processors for AI workloads, driven by performance gains and cost efficiency at scale.

She also highlights how organisations are balancing cloud and on-premises infrastructure to meet privacy and regulatory requirements.

“Organizations are increasingly adopting hybrid AI strategies to balance the benefits of cloud scalability with the control and compliance offered by on-premises infrastructure. Many enterprises use cloud AI for large-scale data processing, rapid deployment and operational efficiency while retaining sensitive workloads on-premises to meet regulatory requirements.”

This hybrid reality complicates autonomy, but it also defines its boundaries. Agentic systems must function reliably across environments with different latency, compliance, and governance constraints.

What still needs to be solved

Despite the momentum, Pinczuk is cautious about declaring agentic AI ready for universal enterprise adoption.

“Before agentic AI becomes mainstream, organizations must ensure the quality and reliability of the data that autonomous systems act on. Trust and governance are critical, with teams requiring explainable, auditable AI actions to stay compliant and accountable. At the same time, enterprises must manage costs, data movement and latency to make deployments practical at scale. Hybrid and multi-cloud environments also require interoperability so autonomous systems can operate consistently across diverse infrastructures. Addressing these challenges is key to adopting agentic AI safely and effectively.”

That assessment captures the deeper message of SentinelOne’s AWS expansion. The future of autonomous security is less about technological breakthroughs and more about execution: clean data, integrated workflows, and AI systems that enterprises can trust to act on their behalf.