The modern enterprise stands at a historic crossroads. For decades, the primary charter of identity and access management (IAM) was clear, if complex: verify the human user, provision appropriate access based on the principle of least privilege, and revoke credentials when employment or contract terms ended. Even as service accounts, application-to-application integrations, and cloud infrastructure entitlements complicated this paradigm, the architecture remained fundamentally tethered to deterministic, human-driven workflows.
Today, that foundational security assumption has been fundamentally shattered.
According to landmark research published in SailPoint’s Horizons of Identity Security report, artificial intelligence has rapidly transitioned from speculative proof-of-concept projects to deeply embedded enterprise infrastructure. An astonishing 79% of global enterprises now report running autonomous AI agents in production environments. Yet, in a staggering display of security lag, a mere 2% of these same organizations have deployed identity security tools explicitly designed to govern, monitor, and audit these powerful digital actors.
This gaping chasm between deployment and governance has created an unprecedented vulnerability landscape. Modern AI agents are no longer passive assistants summarizing documents or drafting emails; they are active execution engines capable of autonomous reasoning, multi-app integration, and real-time decision-making. Armed with sensitive API tokens, corporate credentials, and unfettered access to proprietary enterprise data repositories, these agents represent an entirely new class of digital identity.
To confront this existential crisis head-on, enterprise identity security leader SailPoint used its flagship Navigate conference to announce a sweeping suite of capabilities designed to tame the chaos of the agentic era. By expanding its proprietary Agentic Fabric, introducing specialized AI security guards, and deploying advanced discovery mechanisms for "shadow AI," SailPoint is attempting to establish a new operational baseline for autonomous governance. This comprehensive feature-level rollout aims to shift enterprise security from reactive, human-bottlenecked oversight to continuous, machine-speed automated compliance.
Detailed Chronology: The Evolution of the Agentic Security Crisis
To fully grasp the magnitude of SailPoint’s recent announcements, one must trace the rapid, almost exponential trajectory of enterprise AI adoption over the past thirty-six months.
Phase 1: The Copilot Era and the Illusion of Control
The journey began with the generative AI boom, characterized by conversational copilots. Initially, these tools functioned as isolated extensions of individual productivity. They ingested prompts, generated text or code, and relied entirely on human oversight to execute any downstream action. Security teams treated these early copilots much like traditional software extensions—evaluating them at the perimeter, checking vendor compliance, and assuming that human-in-the-loop workflows would naturally mitigate catastrophic errors.
However, enterprise appetite for efficiency quickly outgrew static chat interfaces. Organizations began demanding tools that could not only answer questions but execute complex, multi-step workflows autonomously.
Phase 2: The Rise of Autonomous Agents and Shadow AI
By late 2024 and throughout 2025, developers and lines of business began deploying autonomous agents—software entities programmed with specific goals, equipped with tool-calling capabilities, and granted programmatic access to enterprise databases, SaaS platforms, and cloud infrastructure.
Crucially, this evolution outpaced traditional IT procurement and security review processes. Employees and department heads, seeking immediate productivity gains, began spinning up agents using low-code platforms, open-source frameworks, and decentralized cloud services without the knowledge of central IT or security operations centers (SOCs). This phenomenon—dubbed "shadow AI"—created millions of unmapped, unmanaged non-human identities. These shadow agents frequently held over-provisioned credentials, granting them broad permission scopes capable of reading, writing, and deleting critical corporate assets.
Phase 3: The Operational Bottleneck
As these agents proliferated, they began interacting with enterprise software at speeds and scales utterly incompatible with human administration. An active AI agent orchestrating a supply chain workflow, for example, might execute tens of thousands of API calls, database queries, and tool invocations within a matter of minutes.
Traditional identity governance and administration (IGA) frameworks rely heavily on periodic access reviews, manager approvals, and manual ticket-based provisioning. When applied to autonomous workflows, this legacy model completely breaks down. Security administrators drowning in an ocean of automated requests are forced to either rubber-stamp permissions blindly or grind business operations to a grinding halt. Recognizing that human-speed governance is mathematically incapable of managing machine-speed autonomy, SailPoint accelerated the development of its counter-framework, culminating in the Navigate unveilings.
Supporting Context & Metrics: The Anatomy of the Governance Gap
The scale of the agentic identity crisis is illuminated by data points gathered across global enterprises. The metrics released in SailPoint’s latest research underscore an alarming operational vulnerability that threatens to derail digital transformation initiatives if left unchecked.
79% Production Adoption: Nearly four out of five enterprises have moved AI agents beyond sandboxed testing environments and integrated them directly into core business operations. These agents touch customer data, financial systems, and intellectual property daily.
The 2% Security Chasm: Only 2% of surveyed organizations possess purpose-built identity tools to govern these AI agents. The remaining 98% are relying on legacy service account management or improvised scripts that lack context-aware authorization.
10,000+ Operations Per Second: The raw velocity of autonomous agents makes manual oversight obsolete. As SailPoint Chief Technology Officer Chandra Gnanasambandam noted, expecting a human security analyst to manually review and approve access requests running at thousands of tool calls per second is not merely outdated—it is a dangerous fantasy.
The Threat Vectors of Autonomous Identities
Ungoverned AI agents present unique security vectors that differ fundamentally from traditional malware or compromised human accounts:
Indirect Prompt Injection: Attackers can manipulate data inputs (such as incoming emails, customer support tickets, or web pages) read by an AI agent, tricking the agent into executing malicious commands, exfiltrating sensitive tokens, or altering database records.
Privilege Escalation via Tool Chaining: An agent given access to multiple tools can chain them together in ways unintended by its creators, effectively bypassing intended security boundaries by using tool A to access data that feeds insecurely into tool B.
Orphaned and Dormant Agents: Just as human employees leave behind stale accounts, experimental AI agents are frequently abandoned in cloud environments, retaining valid credentials and API keys that can be discovered and exploited by bad actors.
Official Statements and Technological Architecture
Addressing these complex challenges requires a complete reimagining of identity architecture. SailPoint’s announcements at Navigate introduce a multi-layered defense strategy spanning discovery, continuous runtime control, and automated governance.
The Agentic Fabric and Shadow AI Discovery
To solve the visibility problem, SailPoint is upgrading its Agentic Fabric with advanced discovery mechanisms. By deploying specialized endpoint and browser sensors, leveraging existing Security Information and Event Management (SIEM) and Extended Detection and Response (XDR) telemetry, and applying sophisticated non-human identity (NHI) discovery algorithms, the platform actively hunts down hidden agents, Model Context Protocol (MCP) servers, API tokens, and unauthorized data stores.
Once discovered, these elements are brought under central governance. The Agentic Fabric provides runtime authorization and controls, enabling security teams to instantly isolate and disable rogue or compromised agents without disrupting the broader enterprise ecosystem.
SailPoint Autonomous Agents: AI Governing AI
In a brilliant stroke of recursive engineering, SailPoint introduced SailPoint Autonomous Agents—specialized, highly secure AI agents whose sole operational mandate is to govern other AI agents.
These security agents operate at machine speed, continuously monitoring the runtime activity of production agents, intercepting suspicious behavioral patterns, and dynamically adjusting permissions. Crucially, rather than shutting down entire applications when an anomaly is detected, these autonomous guards can isolate specific risky behaviors while allowing legitimate, verified agentic workflows to proceed uninterrupted.
Complementing the runtime agents is the introduction of Autonomous Identity Security Posture Management (A-ISPM). Powered by SailPoint’s core Atlas platform, A-ISPM aggregates identity telemetry to uncover dormant accounts, orphaned access pathways, shadow administrators, and over-privileged roles.
Once risks are identified, A-ISPM provides flexible remediation pathways: organizations can configure the system to automatically revoke excessive access, disable compromised agents, assign a human owner for review, or enforce pre-set compliance policies without manual intervention.
Just-in-Time (JIT) Provisioning for the Agentic Era
Moving away from the dangerous paradigm of permanent, standing privileges, SailPoint is emphasizing Just-in-Time (JIT) provisioning. Under this model, human users, service accounts, and AI agents are granted access strictly on an as-needed basis. Permissions are tied to rigorous contextual constraints, such as mandatory reauthentication, explicit business justifications, or validation checks performed by SailPoint’s autonomous governance agents.
Leadership Perspectives
Reflecting on the urgency of the platform rollout, SailPoint President Matt Mills emphasized the paradigm shift currently underway across the corporate landscape:
"Autonomous agents are no longer just experimental copilots; they are becoming the primary execution layer of modern enterprises."
Adding technical depth to the announcement, CTO Chandra Gnanasambandam dissected the operational impossibility of legacy security models in an automated world:
"The idea that a security admin is going to review and approve access for a group of autonomous agents making 10,000 tool calls a second isn’t just outdated — it’s a fantasy."
Future Outlook: The Road Ahead for Enterprise Identity
The global rollout of SailPoint’s enhanced capabilities—with core components of the Agentic Fabric, Human Fabric, and Atlas platforms deploying immediately, and A-ISPM slated for the fourth quarter of fiscal 2027—signals a defining moment for the cybersecurity industry.
As enterprises race to harness the productivity gains of artificial intelligence, the differentiator between sustainable innovation and catastrophic security failure will be the maturity of their identity governance infrastructure. The traditional perimeter has dissolved, and human-centric access controls are no longer sufficient to secure an environment where software acts with agency.
In the coming years, identity security will cease to be a back-office compliance checkbox and will instead evolve into the central nervous system of enterprise trust. Organizations that successfully bridge the 79-to-2 governance gap by adopting autonomous, machine-speed identity controls will unlock the full potential of the agentic era. Those that fail to adapt will find themselves navigating a chaotic landscape of shadow AI, unmonitored execution vectors, and catastrophic data breaches.
The agentic revolution is here. The tools to secure it have finally arrived; the burden now rests on enterprise leaders to deploy them before the window of control closes entirely.
