Good afternoon, everyone. Or I guess good morning for many of you. Thank you all for joining us today. My name is Scott Schmitz, with Investor Relations here at SailPoint, and we are very excited to show you all the highlights from our Navigate conference next week, which was a great event. Before I get started, a little bit of housekeeping.
Today's presentation and discussion are strictly reserved for research analysts, shareholders, and members of the financial investment community. If you do not fall into these categories, please disconnect from the webcast at this time. Additionally, please note that the development, release, and timing of any features or functionality described for our products that are not currently available remains at our sole discretion on a when and if available basis. It may not be delivered at all and should not be relied on in making, purchasing, or investing decisions.
We will not discuss the financial impact of products on this call. Our agenda today is fairly simple. Chandra, our EVP of Product and Chief Technology Officer, will recap key messages from Navigate, and then in the spirit of showing versus telling, Levent and Dan will demo the platform. For Q&A, the chat function is enabled, so please type your questions as we go and we'll aggregate and address at the end. Let me turn it over to Chandra.
Thank you, Scott. Welcome everyone to the super intelligence security briefing, is the way I would describe it. Look, we are at AGI, or at least near AGI, and I think this is pretty important context for the conversation today because the speed at which these machines are operating, the scale and the complexity all is being driven by the fact that we are at or near AGI. With that, I'll really start with what are the key messages that we really delivered last week and that we would like you to take away today. One is, look, given we are at AGI, you can't really solve, or you can't really secure AI by having human in the loop all the time. I'll kind of walk you through more in detail.
We announced our autonomous agents vision, a class of red, blue, and green agents. I'll talk to you more about it, but that was a big announcement last week. We delivered and announced the very first operational identity platform that can truly deliver least and zero standing privilege. This has been an ideology for about 20, 25 years, getting to least privilege and zero standing privilege as the true north of being very secure. There's never been an operational platform that delivered it. With what we GA'd at Black Hat and what we announced last week, the combination is the very first operational platform, both for human and for AI. Atlas, which is our platform, that is the core service layer that underpins everything we do. We have completely rewired it. I'll walk you through the changes.
We have also made Atlas extensible, because our platform has got to be extensible. We announced some major capabilities around plugins and application builder and marketplace. In fact, we conducted a hackathon with 250 - 300 developers, hugely successful on using Atlas extensibility. We have continued to innovate on the core because we have more than 2,000 cloud customers. The core innovation has been going just as fast as these other things. Lastly, on the operational side, we announced a few new regions, support for things around EU, Sovereign Cloud, FedRAMP High, as well as we showed a completely reimagined user experience. Those are the key messages that you should be taking away by the time we are done today. I want to start with how we are solving the problem.
The foundation is really solving it as a human plus AI problem, because you cannot really have two platforms, one for human, one for AI anymore. You have to solve both of those in the same place, which is what we are doing. When you really introduce AI into the mix, you have speed, machine speed. The platform has to be real time, and you have to be in line with threats. On the scale side, we have done the math. We are beginning to see this in our platform already. There are somewhere between 75 million- 150 million agentic interactions a day at any of our large customers. Think of a Global 2000 corporation. When you give an agent a task, it calls multiple LLMs, APIs, data, and so on.
When you do the math on all of that, it is somewhere close to 100 million interactions a day. At that scale, it is not just the scale, it is the topology and the complexity of those interactions, the surface areas that you have to deal with. This concept of zero trust, which was really invented 20 years ago for the human world, will no longer work. This is why we announced our vision for autonomous identity. We were very, very excited to launch our SailPoint autonomous agents. We really describe them in three classes, to really simplify this. It is not like one red, one blue, and one green agent. It is a class of red agents, a class of blue agents, and a class of green agents. Let me really bring this to life for you because we showed our vision around about 10 of these.
Let me really bring this to life. In a red agent, a good example is, let us say you are on PTO, you have a bunch of agents working on your behalf. You have a Microsoft Copilot agent that is taking some notes in the meetings that your teams are doing. You might have an Agentforce agent inside Salesforce and a Claude agent built for you on top of Google Cloud. What if those agents are accessing data that they were not supposed to, that you do not have permission for? This is what a red agent will catch. It will basically say, "Wait a second. This agent is accessing data, critical data, that its human owner was not even supposed to access, and while they are on PTO. This does not look right." A red agent will catch that. A blue agent, now, is actually a defensive agent.
A red agent is an offensive agent. A blue agent will be like, "Oh, okay. Looks like the red agent found something. Let me go do further diagnosis on this." If it actually determines there was a breach, it will stop that breach. Meaning, if this agent that is working on your behalf is accessing critical data that it's not supposed to, a blue agent will automatically stop it. A green agent is what we call a peacetime agent. A good example is automating certification. Look, our customers, these are the large ones, the complex ones. When they run a certification campaign today, even just on the human side, not even including the AI side, a lot of them have between five, eight, 10 million line items in a cert campaign. That's huge complexity.
Our contention is 95% of that should be automated, because we are moving to policy-based governance. In a policy-based world, you should be able to apply intelligence and automate 95%. That's what a green agent does. It's not just a vision. We announced that we will be launching a few of these, at least in the next few months. In fact, last week at Navigate, we signed up one of the largest banks in the world as a design partner to actually build and deliver on the green certification agent by the December-January timeframe. We are quite excited. This was extremely well-received. The reality of today is 99% of the world is in standing privilege. It's going to take a journey to go from where we are today to get to autonomous identity.
That's why going through least and zero standing privilege is critical. This is why our announcement with what we GA'd at Black Hat and what we announced last week, we are the very first operational platform that can truly deliver on least and zero standing privilege. We spoke to you. I walked you through this exact picture on Investor Day. I think many or most of you would have been there. As well as at Black Hat, we announced a lot of this, or we generally made these available. What I'm going to do in the next few minutes is really give you a glimpse of what's new that we announced last week so that you know what's coming. I'm not going to rehash what is already in the market because we have talked with you about that already.
The first of the newness I'm going to talk about, I'll start here. Which is, look, the SailPoint value proposition across both agentic and the human side is discover, govern, and protect. On Investor Day and at Black Hat, I actually talked about protect, and I said, "Protect is where all the runtime controls are." This was extremely well-received last week. In fact, there was a massive reinforcement that the world is sort of moving from doing discovery to really wanting to act on things they are finding and starting to do audit and governance, which is under govern. This was extremely well-received. But I also want to say, because there continues to be a question about when SailPoint says runtime controls, how does it play with other IDP players?
It could be an Okta, it could be a Microsoft Entra or a Ping, so on. What they do is, the IDPs do authentication. Right? Both for human and for AI. They are really the front office or the front gate, or the front door. Once they come in, what we have done now is with all the runtime controls, we have actually introduced another gate. Right? Which is what we call a policy decision point. Right? In the flow of traffic, you can get authenticated and get token exchanged, and a policy decision point will actually evaluate the risk of that access. Right? And determine whether, even though you have the right credentials for accessing a critical piece of data, whether or not you should at this moment. Right? That's what a policy decision point does.
Protect, there are five runtime controls. This is a big deal because you want to have discover, govern, and protect in the same platform. You don't want them in different platforms. In terms of newness, let me quickly give you a sense, and then Levent and Dan will actually show you. As you know, we acquired Entro about 90 days now. I'm thrilled to be announcing that SailPoint Agentic Fabric and Entro will be completely integrated come November, and the fully integrated product will be generally available. At least in my experience and our experience, this is one of the fastest software integrations you will see. We have been very ambitious here. Given the pace of innovation in the market, we wanted to make sure that we had one product that actually brought everything together.
Come November, the fully integrated product will be generally available. On the SailPoint Agentic Fabric, there were a number of net new innovations we announced. Let me give you my own personal favorites, and Levent and Dan will walk you through it. On the discover side, the main message that we delivered was really, or what we announced was around posture, autonomous posture. I'll tell you, look, we have 20+ real-time signals through which we calculate a SailPoint risk score and get posture.
The main differentiating part of the SailPoint posture relative to everyone else in the world is the fact that we can actually act on the posture. It's not just, hey, there is an exposed credential. Our platform will allow you to go rotate that credential or delete it. Right? You can act on the vulnerabilities you find. That's a main differentiator. On the govern side, we are bringing certification for agents and human together. This is a big deal, right? You can no longer certify either human or an agent without knowing who's, or the human or the agent associated with them and what they have access to.
We are bringing the certification experience for human and agent together in one platform. On the protect side, we are really delivering on Just-in-Time . On top of, it's not like we have four or five controls in the market today. We have added more depth and expanded the surface area, right? All around Just-in-Time , real-time authorization. The main advancement we announced last week and a ton of customer excitement was all around the ability to do real-time authorization to fine grain data.
And the way we are doing it is through integration with all the big data cloud platforms, be it Snowflake, Databricks, BigQuery, and so on, as well as all the existing DSPM providers, right? So Cyera to Proofpoint, right? Varonis, right? So if a customer has already done classification and has policy on data in those DSPM, DLP providers, we will automatically integrate with them, grab that context and map it to the identity context, right? So those were the things that we talked, sorry, we announced, and my personal favorites. On the Human Fabric side, I think, in fact, I would characterize our innovation on the Human Fabric side to be just as rich, actually, as what we have done on the agent side.
On the discover front, one of the things that I am personally most excited about is we are really bringing the element of risk and posture to humans, right? This is a big deal. So now, in our platform, if you go to the SailPoint Human Fabric, for every human, you will have a risk score. And the risk score is not what they are supposed to access, what they are accessing at the very moment now in actually real time, and the risk associated with their access, right? So this is a huge advancement. On govern, we introduced, we announced Effective Privilege. This is a huge deal because what you have access to, it is not just what is granted to you, it is all of the indirectly what you inherit through all the groups and roles you are part of, right? And it has always been there.
It has been a hidden problem, and we kind of the world got away with it because at human speed, you could not really find these things. But these machines are running around and finding all this indirect access, and they are becoming a major risk vector, right? A breach vector. And that is why now when we show access graph in SailPoint's Identity Graph, you will see not only what the granted privileges, but also the effective privileges. And we announced on protect, not just JIT, but doing Just-in-Time authorization with policy, continuous evaluation and so on. So that is quickly what to expect. And Dan will give you a glimpse of it. On Atlas, like I said, I think massive innovations, right? So there are four layers of Atlas where we have completely reinvented. One is a data model, right?
We realized you cannot solve this AI problem without really completely redoing the data model. So we have really redone the data model to be from human to human plus AI. We have built a whole context and intelligence layer with the risk engine and intent engine, as well as all of the privilege classification we are doing, because that is the foundation for doing Just-in-Time .
And the policy engine. The policy engine is core to actually us being the policy decision point, right? So that is a net new innovation we announced. And the enforcement layer, where we are the dynamic provisioning engine, as well as the ability to respond right away from the SOC. So that is quickly what we announced in Atlas. Melissa, why do not we just go to the user experience because I am being mindful of the time. I just want to get through this.
Look, for the longest time, the feedback has been like, our capabilities, excuse me. Our capabilities are actually very good. The user experience needed some work, right? We have heard it. About nine months ago, we went and got a world-class designer to actually lead our design team. He has done an incredible job, he and his team. We announced and sort of unveiled a bunch of our newer, the new SailPoint experience, on both the agent side as well as the human side. It is not just the look and feel. We have taken a fundamentally different approach. It is all about our UI is now not just about who has access to what, not just intelligence, but the ability to act on it, right? A very action-oriented UI. We also think the users are expanding.
It is not just an administrator or a practitioner, it is every business user. We are also driving more of a consumerization of our end user experience. Excuse me. I am fighting a little cold. Look, hopefully that gave you a glimpse of what we announced last week. I know I went through it quickly, but what I want to do is hand it over to my good friend and colleague, Levent, our Chief Product Officer, who will actually show you some of the innovations we launched on the agentic side. Levent, please.
All right. Thank you, Chandra. Hi, folks. Really excited to show you the product that we just talked about in action. What I am going to do is I am going to walk you through our agentic capabilities, our Agentic Fabric. I am going to walk you how a security professional and an administrator can discover, govern, and protect their agents, their non-human identities, their credentials, their secrets through our capabilities. Then I am going to hand it over to Dan, our VP of Product for our Human Fabric. He is going to walk you through our capabilities and all the innovations that we have done across the board on the human side. I am going to start with our agentic overview. This is where a security administrator, when they launch our Agentic Fabric, would land. Here, they can see all the agents that they have.
All the unsanctioned AI, those AI platforms that maybe were not approved by the administrator beforehand, all the agents without owners. Some major, basically critical risks here that they want to know about in numbers and all the NHIs that they have. The risk widget here shows you the critical, the very well-organized, and sort of critical high and other findings here you can see in one quick snapshot. A critical attack vector has been, in both the frontier model exploits, but also in general in the cyber attacks lately, has been exposed credentials. These are the credentials that, for some shape and form, leaked either publicly or in other places that are publicly available or accessible to other customers or other users. You can quickly see all your exposed credentials here that we have detected, also sorted by criticality.
This is basically just a great dashboard that gives you a quick glimpse of your landscape, the top risks, the top pieces that you have to pay attention to. We have risky AI agents here again. One of our key capabilities is behavioral monitoring. We monitor the agents and their behavior, as well as NHI and their use, in almost near real-time to see the spikes in their activity. I will show you more about this in detail in a second. We have some great data here about what we found, some of the anomalies that maybe we flagged, and some really useful information here that you can see in a dashboard. Let me dig in to a little more detail about what we actually show. Let us start with the agent inventory.
Our inventory is aimed to be the single source of truth across all your digital landscape. As Chandra said, we have multiple detectors and sensors. We plug into the platforms. We have thousands of connectors, first of all. In addition to that, we plug into all the top AI platforms, all the critical SaaS platforms that are used in enterprise that you can imagine. As well as, of course, we have our own sensors on the endpoint, so we can detect any kind of endpoint agent activity ourselves, as well as on the web browser. We have a very great easy-to-deploy, centralized web browser plugin that basically detects any kind of user and agent interaction.
That is very critical because that means that we can basically detect the full life cycle of an agent and their use of NHI from beginning to end, from the point that the user maybe prompted an agent with a specific task where we can understand the intent and how it is planning its activity, all the way to actually accessing the resource and executing the mission. All along, we can track all that activity and detect behaviors. Let me show you how we do that here. Here is, for instance, these are all of our agent inventory. Let us click on one of these. Here, very quickly, you can see who owns this agent, what are the other users or owners that does this. This was clearly a Cursor agent that was discovered in an endpoint. It is accessing AWS.
Let us look at the critical actual insight that our customers love, which is the lineage, the usage, and the lineage of this agent. Here you can see Ella Peters created this agent on her local MacBook Pro. It was a Cursor agent. Some interesting things. This agent is using multiple identities. We caught all of them. There is a service account here, there is a token here that we kind of created. You can also see what these agents, and through these identities, are accessing to. In this case, it is very interesting. It is accessing a storage S3 storage account. It has access to Figma, the design system. It has a production SQL database. These are some really high privilege access that this agent has through Cursor. That is why it was, if you look at it was probably ranked as critical.
So let's kind of go dig into further though here. We can look at the access graph of this agent. So the lineage showed you what it did. The access graph shows what it has access to. That also goes into our risk calculation. So you can see the full access graph of what this agent can do. You can zoom and slice any of these and understand that. And you can see the identities that are using it, the accounts that are associated with it. But here's the most important tab, I would say. We have a dedicated risk tab here for this agent. We figured out the top issues that this agent has demonstrated and classified those risks. In this case, this agent was using an unclassified, unsanctioned MCP server. That means that it had capabilities that you, the administrator, probably didn't know about.
That could be dangerous. It's an unsanctioned platform, so that's clearly another risk factor. The human owner was confirmed, but the ownership is not exactly, the classification is exactly matching the behavior. So some really great details here that you can see right off the bat. And here's another critical piece here, and this is the behavior. I talked about our abnormal behavior detection. This is basically that in action. You can see on this graph all the sessions and activity that this agent has done over time. And when there was specific things like, or just for instance, in the very first one, it was an unsanctioned agent platform that we saw that, and then it was accessing, it did a spike in its planning for a specific activity. All of these basically got flagged on our abnormal behavior detection, and that figured into the high-risk severity.
So we already kind of found out from the very get-go that, hey, look, this is a very high-risk agent, access and production systems, unsanctioned platform. You don't want this. Again, as Chandra said, this is not just a discovery platform. This is an action platform. So you can very easily take action on this, and you can create policies to, first of all, block this action. I'll show you how. But in this very moment, we know that this agent is the go, so we can just simply say deactivate and deactivate that agent, and that would be gone. It's as easy as that. Okay? So that's the power of this platform. You can deactivate and take action any moment. That's our kill switch that we have here. Now, let me move on in the interest of time here.
We talked a lot about agents and how we detect it. But the reality is agents use credentials, and equally important to manage and govern and protect are the credentials. And here we can see all the credentials that your landscape has. You can slice and dice them by idle ones, by exposed ones, maybe ones that are owned by former employees. That's a major flag that you should manage. Those ones that are privileged, and you can easily go through those filters and either rotate them or take action on them. But more interesting about credentials, and one of the top attack vectors, as we mentioned, especially the frontier model incidents exploited, was exposed credentials. So let's go there, because this is one of the key areas that when we do a POC, customers get surprised quite a bit about.
They are so surprised, A, how they have way more agents and how they have way more credentials than they thought they have, and B, how many of those credentials are actually exposed are not secured or stale or idle or owned by a former employee. This is the place that they would come to, in this case, exposed credentials, and they would take action on it. For instance, in this case, let's take for instance this one. Let's take a random one here. This was an AWS credential exposed in a Slack channel. You don't want that because Slack channels are not exactly safe places to expose your credentials. Again, you can see how that credential was used. In this case, it was another user. It was a Cursor agent, and then that credential got used for access to much broader things.
We can easily take action on this as well, as Chandra said, to rotate it or deactivate it. That's basically our exposed credential. Now, to show this was just one risk vector. To really explore all the risks on your landscape, which is, again, one of the top areas that people are very curious about and get surprised when we do a POC with them is the risk center. This is the place that they come to see all the top risks. Again, here you can sort by critical risks and various type of filters here that we can go do abnormal tokens, abnormal vault activity, former employees again. You can basically go ahead and easily see these risks and handle them as you see. You can see the details, all the connected services, all these pieces are here. All right.
Now we talked a lot about activity of agents and NHI, the risks around them, and how we can manage it, how we can govern and protect them. Let me focus into an area that is very critical for our customers, and that's audit and compliance. Usually, audit and compliance is this ceremony that takes, in some cases, honestly, weeks to finish, and months for some of our large, complicated customers. We really thought deep about how we can innovate here, how can we make their job easier, faster, quicker. We came up with a new innovation here that no other vendor has, which is pre-canned, easy-to-use audit reports that automatically run on a schedule that you would like.
Here we have basically covered most of the frameworks that our customers are trying to audit and comply to, like GDPR and HIPAA and the EU AI Act. The great thing about these is, while we have these pre-canned reports that we can just simply export, that just create the text, and that's going to basically be in progress, and it's going to be ready probably in a few minutes. You can do this, you can export that easily. You can also schedule these on a time basis. This is a great innovation that we added. Customers love it so much that, "Hey, I want this on the human side, too." We're going to now unify our audit capability and create that on the human as well, these capabilities, and it's really been a very exciting capability for our customers.
And each of these are also customizable. If those pre-canned reports do not fit your needs, you can actually go ahead and create your own customized evidence reporting that is as easy to export. All right, so I am mindful of the time here. Let me come to a very important protect capability, and that is our agent authorization capabilities. One of the key things our customers, of course, face is once they discover an agent and their agents in their landscape. Once they discover the NHIs, the first thing they want to do is start putting guardrails. For instance, they can be very simple, they can be very complicated. We can do all of the guardrails that people want to create. But for instance, one guardrail that people want to create is, hey, look, first, I am deploying a coding agent for my developers.
Obviously, those coding agents have some critical access to GitHub and other resources. I do not want that to be used by people in marketing department. We can do that. We have authorization policies that enable that. So here, we have this policy center that enables you to create all these great policy types. I am going to show you a quick one here.
In this case, for instance, we have another one that says, say, look, I do not want our agents to send Slack messages, because you do not want to maybe spam the user or share credentials or anything like that. So let us walk through this. This policy basically says, hey, make sure that Claude Code agent using any MCP servers, it could be a server, it could be a gateway, it does not matter. Whatever they use, again, we enforce it on multiple layers. That is the power of SailPoint.
Cannot send any Slack messages. And remember, notice that it is a very fine grain policy. It just says, "Do not send messages." You can still read messages on the platform with this policy. So this one is enabled. That means you already have the protection. So if I were now to create an agent that tried to create a Slack message, it would not work. So that is a very quick glimpse of all the capabilities that we have on the SailPoint Agentic Fabric. We have way more than that. The time does not allow us to show it, but these are some of the top customer favorites. I am going to hand it over to Dan Martillotti, our VP of Product for Human Fabric, where he is going to walk us through the innovations that we have on the human agentic side, or on Human Fabric side.
Thanks, Levent. I am going to go ahead and look at what we see from our Autonomous Identity Security Posture Management that Chandra talked about earlier. It is the ability for somebody from a security and risk perspective to be able to see what is happening across their identity program for all types of identities, whether it be human identities or non-human identities. And look at a variety of posture dimensions to understand what is happening within the environment and what risk may exist. Now, as Chandra mentioned earlier, the key thing around the posture management work that SailPoint is doing here is it is not just visibility, it is about remediation and action.
When we identify where there is a potential issue within the system, we are going to use AI to be able to analyze the risk and provide you with a recommended remediation of that risk that you can then execute from within this single dashboard. The other interesting thing about this whole program is as organizations are expanding their identity program and expanding the types of identities that they are governing, understanding their capability adoption of the program, and how that impacts their ability to truly govern what is happening in their environment. This gives them some insights into the adoption of capabilities they have across the SailPoint platform and how that impacts their potential risk posture. Now let us take a shift and look at an administrator and the work that they are doing to onboard entitlements into the identity program so that they can be properly governed.
A couple of years ago, we rolled out the ability to just create entitlement descriptions. The idea of entitlement descriptions are, as an application owner, when I define the description of an entitlement, I do it from the perspective of an expert, and it does not necessarily translate into descriptions that end users understand about that application. We created capabilities to create descriptions, but now we have automated that process. When I onboard a new application and we discover new entitlements, we are going to automatically generate entitlement descriptions for all of those entitlements. As the administrator of the program, you can review those entitlement descriptions and make adjustments to them. But now, by providing that level of context, when anybody needs to interact with the application, they have better understanding of what that entitlement grants access to.
Another view of this is looking at what we are doing from a privilege perspective. As you are onboarding new applications, being able to evaluate those entitlements and help you understand what is the privilege associated with each of those entitlements. Is it high privilege? Is it medium privilege? And giving you the ability to make adjustments to that. We also talked earlier about this idea of effective privilege. I can see, for a particular entitlement, how that privilege severity was defined, and it can include the effective privilege that is associated with the embedded entitlements that are a part of that role. That is the power of the Identity Graph. Now we are going to transition to an end user who needs to request access. They need access to go in for themselves.
We have generated a new user experience where instead of having to go in and search for particular entitlements or know the name of something, we have a new agentic experience where you can go in and interact with search and say, I need access to run payroll, for example. We are going to do an analysis of who the user is, what roles they have within the organization, and make a recommendation for the role or the entitlement that they should request. This is a high privileged role, and it is also a Just-in-Time role. I am going to add this to my request and submit it. What traditionally would happen is that entitlement or that role would be provisioned to the end user, and they would have standing access to go perform that task.
Well, since this was a just-in-time privileged entitlement, we now have a launch pad that the user can go to, and when they are ready to use that entitlement that they are pre-approved for, they can click on Start Session. That will be when we automatically provision the access for that particular item within the boundaries of policy associated with that access. We may require the user to provide some justification or a ticket number to be able to activate this access. But once the activation happens, now we provision that, the user can perform that task. They are time-bound based on the policy associated with that particular entitlement. When they are done, they can simply end the session, and that de-provisions that access. So now we have moved from standing privilege or standing access for privileged entitlements to Just-in-Time access when I need to use it.
I have got pre-approved permission, but it is not a standing privilege. Last but not least, we talked earlier about certifications. This is one of the toughest things that our customers have to deal with. You are asking day-to-day professionals to go in and certify and review the access of all of the employees that they have got. I know that when I talk with Levent about this, there are things that I have access to or people on my team have access to.
We are not quite sure whether they should have that access or not. We do not have all of the details about the entitlement or how they are using those entitlements. With this new mechanism that we have developed, we can go in and automatically generate and review the access associated with all of those users and make recommended decisions on the access for all of those entitlements.
It is going to be based on a variety of guardrails that are defined as the certification policy is put into place. I can create guardrails that say, for example, if this is privileged access and it has not been used in the last 30 days, recommend revoking that access. If it is access that has been reviewed multiple times and have been approved, let us just go ahead and automatically recommend keeping that access. But what we see as a reviewer now is all of the recommended actions that are being taken. I could override these if I choose to, but I can see all of the details associated with every single one of the decisions that the system has made, and I can choose to adjust or confirm those decisions.
We believe this is going to really reduce the amount of time that people spend reviewing the access that their employees have and really create more value. All right, Chandra, I am going to turn it back over to you for any closing thoughts.
Sorry. Yeah. No, look, I don't have any particular closing remarks. Scott, I'll turn it over to you because I think we should really move to taking Q&A now.
Yep. Perfect. Yeah, there's several questions that were submitted. So if you have others, please type them in and I'll go through a few here. So the first one I wanted to relate to the team, there's a couple of different versions of this, so I'll just read one of them that kind of encompasses it. But a lot of identity security players are trying to become the runtime enforcement layer for Agentic Identity. What gives SailPoint the right to win here compared to other identity vendors which are already operating in line for human access management?
Sure. I'll start. I'll answer that in two parts. One is, look, the only way to really enforce runtime controls in a very, very effective way is to have deep identity context. And the best way to think about this is the proverbial kill switch now. The kill switch for agents. A lot of runtime platforms claim to have a kill switch. Our view is an agent kill switch is supposed to kill a bad agent. Now, a kill switch should know or what's the basis on which a kill switch will know what is a good agent versus a bad agent? It's like giving a security guard a gun and say, "Well, shoot the bad players." They've got to know how to distinguish the good and the bad players. So that requires knowing deep identity context.
You've got to be able to say, this agent is a bad agent because it violated all the policies of access. There is a pattern of access behavior here where it's tried doing bad things. We have analyzed the intent for that agent, and it doesn't have the right intent in the way it's prompting. And with that depth of context, you have to combine that with the runtime control, which is what we are doing, and that's a real differentiator. And the context comes from discover and govern, and protect is where we enforce that. So that's sort of part one. Part two is in most cases, we actually coexist. It is not a one-size-fits-all. It's not a one winner takes all. So let me give a couple of examples. So you could say Okta, Microsoft Entra, they are in line with the traffic.
They are looking at the token. They are doing token exchange and so on. They are really the front door. Once you are inside, at a very fine, they will tell you have access to the Snowflake database, okay? Go. You are validated. Now, it could be a human or an agent, but a specific row or a column of data inside Snowflake and whether the data is very sensitive, and at the very moment that human or agent, can they have access to it? That context, they do not have. That context we have, and that is why we are the policy decision point, and we are the real-time authorization engine. The IDPs do authentication, and we do authorization, fine-grained authorization. Hopefully that makes it clear on why on the differentiation and the need to coexist.
Great. Thank you. I think a good follow-up to that is that we have several questions on just the differentiation versus the platform vendors. What telemetry do we rely on from those vendors, and what are we incorporating into our platform from the platform players?
Yeah. Look, Levent, do you want to talk about how we have our own telemetry, but we also get all the logs from the—
That is right. Exactly, yeah—
—SIEMs and so on.
—I can drop in, yeah. Look, we are hooked into as many places as we can, and we never turn down data. We, first of all, have our own sensors and telemetry directly that we source. Again, from Agentic perspective, we are on the endpoint. We have an endpoint sensor, as I mentioned earlier. We are in the web browser, so we can get direct interactions in a web browser with agentic pieces. We have thousands of connectors, so we connect to all types of different sources, and both for the humans and the agents combined. So we can see their access, what they are entitled to. We actually manage that access. We control that access, and we know exactly what they are provisioned and not provisioned for. On top of that, of course, we also plug into all the SIEM and XDR tools.
We can get real-time log data from many of the providers out there, from network providers and IDPs, both at the log layer but also at the direct integration with their platform. So we can see all the activity through both those platform players as well very easily. Literally, I can go into details here, but we are aggregating and putting all that information together. It is our unique differentiated data that we have through our thousands of connectors, our presence through endpoint and the web browser and the direct plugins to all these platforms. Plus, we also suck all the data in, as the customers, of course, allow us to do, to overlay it, combine it, and add human and data context to it. That is the power of it.
We make intelligent risk assessments and creations and create policies out of it that then people can deploy on their platforms.
Levent, it's not just access to applications, it's also access to data. We look at it holistically —
Good point—
—across the board.
That's right. We plug into all the DSPM, DLP providers. We have our own classification, but we also get all the data classifications for customers that have existing classifications from their Microsoft Purview or Cyeras of the world or Proofpoints of the world. So we can easily get that and also make that a policy decision, a parameter for our policies, our customers on our platform.
Got it. I think you probably answered this in some of that response. What makes SailPoint the right kill switch relative to some of those other players in the market? Where they have a lot of visibility as well, why is SailPoint the right kill switch?
Yeah. We use the word context, and I recognize it's actually an overloaded word. The plain English translation of that is, the kill switch needs to know, particularly for agents. What is the risk associated with that agent? If that agent has gone rogue because it has drifted from policy, the policies that were assigned to it, how bad is the drift? Is there a pattern of behavior over the last few weeks where it has continuously drifted? Analyze the intent, because a lot of times, what people don't realize is agents are prompting. They will go to a command line interface, and they will prompt something to get access to a tool, let's say. Analyzing those prompts and seeing whether the intent behind those prompts is right, this is what we do.
When we say the lineage, the depth of context, we take 10 parameters into account to actually determine risk. This is why this risk and posture is very important. Then we use that to do the enforcement. Actually doing the enforcement is actually the last mile. 80% of the traffic is in actually doing all the analysis, the policy drift, the risk, the intent, all that analysis to make sure that you are 100% convinced that you have spotted the real bad actor, and now is the time to stop them.
Yeah.
Chandra, the additional piece of that is the human context. Many of those agents that you're talking about are acting on behalf of somebody. If the human context provides insight that there's risk, that can flow through as a part of that.
Fantastic. Imagine an agent shows up on your front door, and they have a pass that says, "Hey, I want to access this resource. I want to get into this house." They have a visitor pass. It looks legit. That is what an access layer does. They let them in based on that pass, which is a token, with a claim on it. SailPoint does is hold on. That agent is trying to access, or that person is trying to access that house. They have bad intentions. They were trying to break into deep financial information, maybe. They are owned or they are associated with a network of bad players. They have your behavior in the past, that have malicious intent and malicious behavior in the past. They are trying to obviously get access to all the critical things in your house. Would you let them in?
That is the SailPoint power. An access there might just see the token and say, "Oh, you have a visitor pass, go in." We see the full context of data, what they have done, what they did, what they want to do, and basically stop it right there. I think that is the differentiation that we bring to the table.
Very helpful. Next question is really, again, all these are kind of related. How do we enable the provisioning certification review of agents that are ephemeral in nature? How do we manage the ephemeral—
Yeah—
—agents? Really simple—
Yeah—
—questions with that.
I can take that. We capture ephemeral agents just easily because, again, we have near real-time capabilities of detecting these agents in real time with our sensors. So we can see them, and there are multiple ways. First, as I showed, we can create policies to say, "Hey, look, I do not allow unsanctioned AI platforms to do certain things, and I also don't want them to access to certain resources." For instance, a marketing agent, however long-lived or ephemeral it is, cannot access maybe GitHub repositories or production systems for coding. So those policies get applied no matter what, ephemeral or not. We detect them, we apply our policies. And we of course also have a record of those ephemeral agents, how to detect them as well. So if you need to certify their access or audit them, we'll have all that in our evidence trails.
Great. Okay, switching gears a little bit. A couple of questions came in on IdentityIQ. So can we just kind of quickly recap what are the importance of the announcements that we made at Navigate on IdentityIQ and how does the features, or how do you manage agents if you're on IdentityIQ today?
Great question.
—y ou can install it, right?
Yeah.
Yeah.
Do you want to take the IIQ, and I'll take the agent side?
Sure. Look, the reality is lots of customers are still running on top of IdentityIQ, and we're continuing to invest there. There's a lot of new features that we delivered as a part of this latest version of IdentityIQ. Part of that is really modernizing the platform on which IdentityIQ runs. As a piece of software that runs within the customer's environment, they're responsible for the web servers, et cetera, that are running on top of it, and those are maturing and changing. Part of the investment that we made was making sure that we could help customers run on a secure environment while we continue to invest in the core capabilities that are a part of that application.
Yeah. We understand and obviously, while we're investing in IIQ, a lot of customers still have agentic needs, and our Agentic Fabric does not require, it can run standalone. It doesn't require you to be on our Human Fabric. It has direct connectors to IIQ. It can bring in the human data context from those connectors easily. So you can have IIQ and our Agentic Fabric run side by side, without doing a migration of any kind of sort or anything like that. That's the beauty of our setup here.
That's right.
Okay, great. I'm going to go back. Another question came in. I just want to make sure we clarify this point, so I want to ask this question as well. It said, in line with previous replies, what happens when agentic activity happens on the endpoint, not necessarily via MCP? Do we have visibility into that, or how do we manage that endpoint that we may not, do we see all that traffic?
Yeah, look, we— Yes?
Yeah. I'll handle that one.
Sorry, Chandra.
Oh, no, please.
Okay.
Go on.
Look, we meet our customers where they are, and what we see is while there's a lot of convergence on MCP, we realize a lot of people have different setups that they have to accommodate for. We have layers of protection that basically come in and both detect and protect and govern those agents. Obviously, if the customers have MCP, we plug into their MCP gateway, or we have our own MCP gateway. We also have an MCP server that we can use on locally. But if people don't want to do MCPs or if it's not maybe a shadow AI, we have an endpoint sensor that people push out to their endpoint devices, and we can basically monitor all the processes and figure out what the agents are doing and apply our policies that way as well. We don't have to have an MCP.
Yeah—
We basically protect and guard against non-MCP agents as well.
The only thing I would add is we also have plugins. For instance, for all the big endpoints, we have a plugin. So be it Claude Code, Cursor, Codex. Right? Let's say there is a new endpoint agent that gets launched tomorrow. If they don't support MCP, it'll take us a week or something to actually build a plugin. Because we have the full context, and we view the endpoint as both a detection mechanism as well as an enforcement mechanism. As long as they have some published API, we can do enforcement through them. It doesn't have to be MCP-compliant API.
Yeah.
Yeah. It has to be both, right? Because sometimes agents will bypass MCPs and go directly to the endpoint, so we have to be able to do both.
Right.
Good point. Okay, another question here, again, switching gears a little bit. Just in terms of what's available today, what's coming. I know, Chandra, you had a slide, but maybe just, is there a way to kind of, for investors to think about what we announced at Investor Day, what's GA today, what's coming?
Sure.
Any kind of high level—
Yeah. Think of them in sort of three tranches. A good chunk, a good majority of them are general or will become generally a vailable November 1, which is our start of Q4, right? I guess in about 20 days from now, that is sort of tranche one. Tranche two is by end of year, sometime in December before our Q4 ends. Then there are a few things that will be in Q1 of next year. So think of a February, March timeframe of next year.
Yeah.
Those are the three tranches, but a good majority, I would say 80% of what we talked about is November 1, end of year.
Great. I think one last question here, and I know we are running up on time, so I think this is a good one to end on. Can you just help us understand where customers are on their journey to help govern and secure agents? There is a lot of different noise and tools in the market. Just help us understand what we are seeing from customers in terms of where they are on their journey.
Yeah, I will take that. Look, we had, I think, I want to say close to 120, 125 conversations last week. Here is sort of the theme, the big takeaway, which is it was even up until two months ago, discover. It is still that, but there is a big change in terms of they are like, at some level, we are overwhelmed with discovery. We might not be discovering everything, good and bad, but we have enough. What do we do with it? There is one big shift in the market, which is, take action on it. You found 300 exposed credentials, rotate them or take them away right away, right now, right? Theme number one is take action on it, runtime controls.
Theme number two, particularly in the last two weeks after the White House meeting with the tech giants, and the President signing into an executive order of the Super Intelligence Accord, audit and compliance, right? Just last week, the volume of that conversation, the fidelity that just went up by orders of magnitude because if you go read the executive order and a lot of it is all about accountability, and accountability comes from audit and compliance. With fine grain controls, right? Those are the two big things in terms of market or customer adoption from discovery to runtime controls, audit and compliance.
Perfect. I think that is a great spot to end. If we did not get to your questions, I will follow up. If you have other things that came up or things that you were thinking about, please reach out to me. We do go quiet on Friday, so hopefully we can touch base either later today or tomorrow for any questions that are still outstanding. Otherwise, we will look forward to talking to you again during earnings. Chandra, Levent, Dan, thank you very much. I think that was very helpful. So appreciate everybody's time.
Thanks, Scott. Bye.
Thank you.