SailPoint Says Entro Integration Lands November 1 as It Bets Fine-Grained Runtime Authorization, Not a Kill Switch, Is Its Edge in Agentic Security
Post-Navigate analyst call, October 14, 2026: management details autonomous agents, rebuilt Atlas platform and a three-tranche release schedule
SailPoint used a post-Navigate briefing for analysts and shareholders to argue that identity security is moving from discovery to enforcement, and that the company is the only vendor with an operational platform that can deliver least and zero standing privilege for both humans and AI agents. Management said roughly 80% of the capabilities shown at Navigate and Investor Day will be generally available between November 1 and year-end, with the remainder in February or March 2027. Executives declined to discuss financial impact. The call was a product walkthrough, and the claims about competitive uniqueness are management's own and were not independently tested.
Release timing and the Entro integration
The most concrete new information was the delivery schedule. Chief Technology Officer Chandrasekar Gnanasambandam said the roadmap falls into three tranches. A "good majority" of the capabilities will be generally available on November 1, the start of the fourth quarter. A second group will follow in December, and a few items will slip into the first quarter of 2027. He estimated that "80% of what we talked about is November 1 end of year."
SailPoint also named its combined agent product the SailPoint Agentic Fabric. It will fold in Entro, acquired roughly 90 days ago, with the fully integrated product generally available in November. Management called this "one of the fastest software integrations you will see." That is an aggressive pace for a deal this young, and investors should watch whether the November date holds, since the integration underpins the agent-side story.
Autonomous agents: red, blue and green
SailPoint introduced a framework of autonomous agents in three classes. Red agents are offensive and look for anomalies. Gnanasambandam's example was an agent from Microsoft Copilot, Salesforce Agentforce or a customer-built system touching data its human owner was never entitled to while that owner is on vacation. Blue agents are defensive. They investigate what red agents flag and stop confirmed breaches automatically. Green agents handle routine governance work.
The green agent has the clearest commercial anchor. Large customers run certification campaigns with 5 million to 10 million line items, and management argues that 95% of those should be automated under policy-based governance. SailPoint said it signed one of the world's largest banks as a design partner at Navigate to build a green certification agent, targeted for a December or January delivery. Management said a few of these agents will launch within months. A single design partner and a timeline of a few months are early evidence, and the red and blue agents remain at the vision stage.
Management sized the problem at 75 million to 150 million agentic interactions per day at a Global 2000 company, since each agent task can fan out across multiple LLMs, APIs and data stores. Gnanasambandam said that at this volume and topology, the Zero Trust model built for humans "will no longer work."
The competitive argument: authorization, not authentication
The most pointed question from investors was why SailPoint should win runtime enforcement against identity vendors that already sit in line with human access traffic. Management gave two answers. The first was context. Gnanasambandam argued that many vendors claim an agent kill switch, but a kill switch must know which agent is bad. "It's like giving a security guard a gun and saying, shoot the bad players. They've got to know how to distinguish the good and the bad players." SailPoint says it combines about 10 risk parameters, including policy drift, behavioral patterns over weeks, the intent behind agent prompts and the human owner's risk profile. He said "80% of the traffic is in actually doing all the analysis," with enforcement as "the last mile."
The second answer was coexistence. Management positioned Okta and Microsoft Entra as the front door that authenticates humans and agents and exchanges tokens. SailPoint adds a policy decision point behind that door. The policy decision point evaluates whether a credentialed identity should reach a given row or column of data at that moment. Levent Besik, the Chief Product Officer, offered an analogy: an access layer sees a valid visitor pass, while SailPoint asks whether the person holding it has a history of malicious behavior. Gnanasambandam summarized it this way: "The IDPs do authentication, and we do authorization, fine-grained authorization."
The coexistence claim is a hedge as much as a strategy. It concedes that SailPoint is not trying to displace the IDPs. Its position depends on the IDPs not extending their own authorization depth, and the call offered no evidence on that point. SailPoint does plan to integrate with Snowflake, Databricks and BigQuery, and with data security posture management and data loss prevention vendors including Cyera, Proofpoint, Varonis and Microsoft Purview, to pull in existing data classifications.
Product demonstrations: discovery, kill switch and just-in-time access
Besik's demonstration showed a console that inventories agents, unsanctioned AI platforms, agents without owners and non-human identities. Telemetry comes from thousands of connectors, endpoint sensors, a browser plug-in, and SIEM and XDR feeds. In the demo, a Cursor agent created on a developer's laptop held a service account, a token, and access to S3, Figma and a production SQL database through an unsanctioned MCP server. Behavioral monitoring flagged it as critical, and a single click deactivated it. Besik said customers running proofs of concept are routinely "surprised" at how many more agents and credentials they have than expected, and how many credentials are exposed, idle or owned by former employees. He showed an AWS credential exposed in a Slack channel as an example.
He also highlighted pre-built, schedulable audit reports covering GDPR, HIPAA and the EU AI Act, which he said no other vendor offers. Policies can be fine-grained, for example blocking an agent from sending Slack messages while still allowing it to read them. Management added that enforcement does not depend on MCP. An endpoint sensor monitors processes directly, and plug-ins cover tools such as Claude Code, Cursor and Codex. Gnanasambandam said a new non-MCP endpoint agent would take "like a week or something" to support, provided it has a published API.
Dan Martillotti, VP of Product for Human Fabric, demonstrated just-in-time privilege. A user requests payroll access in natural language, and the system recommends a role and pre-approves it. Access is provisioned only when the user clicks to start a session, and it is revoked when the session ends. Certification reviews arrive with recommended decisions based on policy guardrails, such as revoking privileged access unused for 30 days. Reviewers can override those decisions.
Platform rebuild and human-side additions
Management said it has rewired Atlas, its core service layer, across four layers: a data model moved from human-only to human plus AI, a context and intelligence layer with risk and intent engines, a policy engine, and an enforcement layer for dynamic provisioning and SOC response. Atlas is now extensible through plug-ins, an application builder and a marketplace. A hackathon drew 250 to 300 developers.
On the human side, SailPoint is adding real-time risk scores for every person and "effective privilege," which captures access inherited through groups and roles rather than only granted access. Gnanasambandam called indirect access "a hidden problem" that human-speed attackers could not exploit at scale but machine-speed agents can. The company also pointed to new regions, EU sovereign cloud support, FedRAMP High and a redesigned, more action-oriented user interface. Management acknowledged that user experience had long been a weakness: "our capabilities are actually very good. The user experience needed some work." A new design lead joined about nine months ago.
Legacy IdentityIQ base and customer demand
Investors asked about the installed base on IdentityIQ, the customer-hosted product. Management said many customers remain on it and that it received modernization work, mostly on the underlying web server and security environment. The agent product can run standalone and connect directly to IdentityIQ to pull human context, so customers can adopt it without migrating. That lowers adoption friction, but it also means the agent opportunity does not necessarily accelerate the cloud migration of the legacy base.
Gnanasambandam said that in roughly 125 customer conversations at Navigate, the theme shifted from "discover, discover, discover" two months ago to acting on what is found, with customers saying they are "overwhelmed with discovery." A second theme, which he said surged in the last two weeks, was audit and compliance, following what he described as a White House meeting with technology leaders and an executive order called the Superintelligence Accord, which he said centers on accountability. Management's framing of "we are at AGI, or at least near AGI" is a strong claim that investors should treat as positioning rather than fact, and the call did not quantify any revenue effect from either demand shift.