Artificial intelligence is no longer just a tool for analysts or an assistant for writing, coding, and creating content. It’s becoming a co‑worker, co‑defender, and in some cases, the first line of defense. With the growing shift to the agentic workspace, where AI agents act on behalf of people, CISOs must rethink their approach to security. We need to think not just about our users and systems, but these AI agents themselves.
In this post, I’ll explore how agentic AI is transforming the SOC, how security operations must evolve, and how Proofpoint is stepping in with new capabilities announced last month.
What is Agentic AI and Why It Matters for Security
Traditional AI in security tends to be used for predictive analytics, anomaly detection, or automated response triggers. But agentic AI describes autonomous agents that can take actions, understand context, and collaborate across workflows.
In an agentic workspace:
- AI agents interact with users, systems, and data. They are making decisions (e.g. summarising, drafting, automating tasks).
- They consume and generate data, potentially triggering or influencing risk.
- They can themselves be targeted through prompt injection, social engineering, or data manipulation.
Because these agents behave more like autonomous collaborators than passive tools, they expand the attack surface. Your security model must evolve to treat agents as privileged human users including monitoring, policy enforcement, and oversight.
How SOCs & Security Teams Evolve with Agentic AI
Here’s how I see SOCs and security organisations shifting under the influence of agentic AI:
- Alert Triage will become Agent-Assisted: Rather than routing, deduping, or enriching alerts, agents can classify, prioritise, and suggest next steps in real time. These agents can learn from analyst decisions, progressively improving triage accuracy.
- Automated Repetitive Remediation: Low-risk, repeatable tasks (e.g. blocking a known malicious IP, applying a patch, isolating endpoints) can be delegated to AI agents. Analysts will supervise or audit these agents, so they can focus mental effort on higher-risk or more nuanced cases.
- Threat Hunting & Hypothesis Generation: Agents can assist in generating hypotheses, pulling correlated data, automating pivot paths, and surfacing anomalies across large datasets. Particularly when combined with threat intelligence, they will lead in proactive detection.
- Explainability & Oversight: Because agents act, their decisions must be explainable. Teams will need audit trails, decision logs, and guardrails to prevent drift, bias, or unintended data leakage.
- Trust Calibration: Not all agents should be “fully autonomous.” Some should operate in human-in-the-loop mode, escalating uncertain decisions. Over time, trust is calibrated: more mature environments might allow greater autonomy for proven agents.
To summarise, a modern SOC will become more orchestration, oversight, and governance than pure execution. The tools become more intelligent, but human judgment remains vital.
Proofpoint’s Innovations & How They Address Security in the agentic AI Space
During our flagship Protect event in Nashville in September 2025, Proofpoint made multiple announcements signaling how we intend to lead in the agentic AI security space. These are not minor tweaks - they are material advances in how the agentic workspace is secured.
Agentic AI for Human Communications Intelligence (HCI)
- Proofpoint launched the industry’s first agentic AI solution for Human Communications Intelligence (HCI).
- Unlike legacy connectors or post-incident archiving, HCI agents interpret, reason, and flag emerging risk in real time across 80+ communication channels (email, collaboration, voice, chat, etc.).
- This gives compliance, legal, and security teams a proactive, contextual view of intent, behavior, and risk before it escalates.
Industry-first Innovations to Secure the Agentic Workspace
At Proofpoint Protect ‘25, we also announced four new innovations aimed at protecting both people and agents in the agentic workspace.
Key components of these innovations include:
- AI exploit detection over email (Prime Threat Protection): Blocks malicious prompt injections, AI-targeted lures, or crafted content that could trick AI assistants before it reaches inboxes.
- Proofpoint Data Security Complete: A unified data classification, discovery, and protection solution that covers both human and agent data flows. It uses Autonomous Custom Classifiers, and tracks both sanctioned and unsanctioned AI usage.
- Secure Agent Gateway (using Model Context Protocol / MCP): Controls how AI agents access data, enforces usage policies, and can block or redact sensitive data before it reaches agents or back to humans. It works with Data Security Complete to provide consolidated agent/human data controls.
- Proofpoint Satori Agents & Agent-to-Agent Access (Satori MCP Access): Internal AI agents that help automate tasks: DLP alert resolution, recommending simulations, handling threat responses. It allows other compliant agents (e.g., external agents or third-party LLMs) to invoke Satori for cross-system orchestration.
These announcements make Proofpoint’s vision clear: secure interactions not just among people, but among people, agents, and data in a unified, defensible architecture.
Strategic Implications & Recommendations for CISOs
Given these developments, here’s what CISOs should be thinking about now:
- Upgrading Architecture to Be Agent-Aware: Security strategy must evolve beyond people and devices - agents are now integral. Start categorising agents (internal, third-party, human-initiated vs. autonomous) and mapping their data and communication paths.
- Enforcing Policy Guardrails & Least Privilege for Agents: Apply the same zero trust, least privilege, and policy enforcement paradigms used for humans to agents. Use gateways like MCP-based proxies to mediate access and redactions.
- Leveraging AI Agents for Scale, but Oversight Always: Deploy agents in SOC and governance workflows but with human-in-the-loop modes initially, metrics for accuracy, and gradual autonomy calibration.
- Integrating Threat Intelligence, Behavior Signals & Agent Context: An alert from a human account is meaningful. An alert from an agent executing suspicious action is more so. Build pipelines to correlate and reason across them.
- Preparing for Explainability, Auditing & Compliance: Agent decisions will be scrutinised. Ensure logs, rationale, and overrides are stored. Use governance tools and AI models that provide reasoning, not just black-box outputs.
- Embracing Proofpoint’s Agentic Capabilities: With the latest launch of HCI, Satori agents, Secure Agent Gateway, and agent exploit detection, Proofpoint is uniquely positioned to help you to monitor communications intent in real-time, enforce data and agent policies holistically, automate SOC tasks with internal AI agents and bridge human-agent collaboration with trusted oversight
As the agentic workspace becomes the new norm, these tools will be foundational to futureproofing your organisation.
The Next Frontier of Trust
Agentic AI is not a fad - it’s a transformation of work. And as agents become collaborators, the distinction between human and machine security risk will blur.
For CISOs, the task is urgent: move beyond legacy toolsets, adopt an architecture that embraces agents safely, and build human-AI agent co‑defense as your posture. Because when your adversary can exploit AI agents or trick your digital co-workers, defending only human actions is no longer enough.
