Field Notes·Bulletin·28 Oct

Pranay Yadav on the Identity at the Center podcast

Pranay Yadav of Iden joined Jim McDonald on Identity at the Center on how deep identity governance has to go, where SCIM stops and why AI agents are hard to govern.

Pranay Yadav
Pranay Yadav·Hosted by Jim McDonald·57 min

About this episode

Identity at the Center released episode 457, a Sponsor Spotlight on Iden, on October 28, 2026. Pranay Yadav, Iden's co-founder and CEO, spent 57 minutes with host Jim McDonald on why governance at so many companies still runs on tickets and spreadsheets once single sign-on is in.

Jim ran IGA programs as a practitioner before he started hosting, so much of the hour is about depth and order: which apps need governing down to the permission, which only need a clean offboarding, and which to connect first. The last third turns to AI, both using it to run governance and governing the agents that now act with a person's identity.

About the show

Identity at the Center is a weekly podcast about identity security and IAM, hosted by Jim McDonald and Jeff Steadman. Sponsor Spotlight episodes are made possible by companies that support the show, and go deeper into one company's view of IAM.

About the guest

Pranay came to identity from venture capital at Accel, where he kept hearing the same story from IT teams at companies of every size. He started Iden after seeing it again at his own company. He also hosts Iden's own podcast, Field Notes recordings, with people who run identity inside real companies.

What comes up

  • Why "Pranay has a GitHub account" is a fact and "Pranay can approve his own pull request" is a risk, and why an access review that stops at the app isn't what auditors ask for.
  • The 2 cases where knowing only that someone has an account is enough: offboarding, and the CFO's count of licenses.
  • Where SCIM stops. It creates and removes users, and nothing in it maps the group in your identity provider to the role inside the app. Iden calls its connectors SCIM++ because they carry on past that point.
  • Depth in practice: invoice roles in a procurement tool, teams and repositories in GitHub, and an internal ERP whose RBAC goes 3 or 4 levels deep.
  • Why a governance program is only as complete as the weakest app in the stack, and how to sort apps into P0, P1 and P2 by audit scope, the data they hold and what a license costs.
  • When to automate an access request, and when to keep a human in the loop until the pattern repeats.
  • AI for governance: building connectors faster, and giving LLMs deterministic tools to call so every action leaves an audit log.
  • Why AI agents are hard to govern: they borrow a person's identity, share one credential across many tools and run on long-lived tokens.
  • Why nobody funds a multi-year identity program anymore, and Jim's line on phase ones that took a year and connected 5 apps.

Chapters

  1. Welcome and Sponsor Spotlight introduction
  2. From venture capital to identity governance
  3. Where the friction starts for small IT teams
  4. Starting with the apps that lack built-in provisioning
  5. The floor and the ceiling of who feels the pain
  6. How deep governance has to go
  7. When knowing "has access" is enough
  8. Where SCIM fits and where it stops
  9. Depth in practice: SaaS roles, GitHub and internal apps
  10. Breadth and the long tail of apps
  11. Prioritizing apps into P0, P1 and P2
  12. What governance means functionally
  13. When to automate access requests and when to keep a human in the loop
  14. The experience for the IGA administrator
  15. AI for governance
  16. Why governing AI agents is so hard
  17. No appetite for multi-year identity projects
  18. Lighter note: advice for a firstborn
  19. Wrap-up

Mentioned in this episode

Transcript

Read the full transcript, about 9300 words

Machine-transcribed from the episode and lightly corrected, so a word here and there may be off.

Jim

Welcome to the Identity at the Center podcast. I'm your host, Jim McDonald. Jeff Steadman was unable to join us for today's discussion. Today's actually a sponsor spotlight, and sponsor spotlight episodes are made possible from organizations that believe in what we're doing here. We take a deeper dive into IAM trends and perspectives straight from the source.

So today's sponsor spotlight is on Iden. They can be found at idenhq.com/idac. And joining me today I have Pranay Yadav. He is the CEO of Iden. Pranay, welcome to the podcast.

Pranay

Hey Jim. I'm glad to be here.

Jim

We're really glad to have you. And this is your first time on the podcast. Surprisingly, you've been in the industry for quite a while. And I was actually wondering, and every time we have a new guest on the podcast, we want to start with the question: how did you get into identity? Did it choose you or did you choose it?

Pranay

Okay. Well, first of all, Jim, I've heard a lot of your podcast already. So even though I am new to the show, I'm not new to the show.

And just to answer your question: I didn't choose identity. Identity chose me, I'd like to put it this way. I actually came to it from the other side of the table.

Before starting Iden, I used to work for a venture capital firm called Accel, and my job was to talk to portfolio companies and invest in new cybersecurity companies, AI companies, data companies. As a part of that process I ended up speaking to many of the IT and security practitioners at all of those companies, all the way from 50-employee companies that we invest in to pre-IPO stage, multi-thousand-employee companies.

One thing that kept popping up every time you speak to them is they usually bring in single sign-on pretty early in their journey, because that's something that you need when you hit a certain critical mass of employees. And then quickly after that, things start becoming tickets and spreadsheets. That's how the industry works.

And that's exactly what I saw even when I started my company, where we went from a handful of engineers to now 20, 25 employees, and it's still largely manual for us as well. That's where the idea was born: there has to be a better way to solve identity governance. So yeah, that's how we started Iden, actually.

Jim

And one of the things we talked about was that you really have a passion about removing friction from the process and enabling IT teams to be more efficient. My question is, when you started talking to teams, especially smaller IT identity teams that rely heavily on support tickets, where were you seeing the friction?

Pranay

I mean, the friction starts all the way from 50 employees. That's when we feel the friction actually starts, where an IT manager or an IT practitioner starts doing a lot of those things by hand or tracking them on spreadsheets. So it actually starts from there.

And they usually have just brought in an SSO tool, because SSO can't be done by hand, right? That's why you bring in a tool. But everything else along with it, like provisioning that user, deprovisioning that user, tracking what permissions they should have and actually ensuring that they have the actual permission they need, all of that starts becoming a support ticket and something that you track on a spreadsheet.

I can give you an example. Let's say you are a 50-employee company and you're an IT manager at that company. You usually would have brought in one of the SSO tools out there. Then as soon as somebody joins the company, you need to give them access to whatever their manager says they need access to. For example, an engineer might need access to GitHub, Vercel, different engineering tools, right? And you want to give them access to a certain repository, certain modules, certain resource groups.

All of those things are usually done either on Slack or on a support ticket, or on good days maybe somebody is even tracking it on a spreadsheet. That's the universe I'm talking about, and that's how the friction actually starts. And this just keeps getting amplified as a company grows into more employees, more contractors, more applications in your stack. So that's the problem.

Jim

How did that problem shape what you wanted to have Iden become?

Pranay

We really started by talking to people and going to the root of the problem. We started by looking at, okay, what are the apps that you want to automate first, right? For example, some of these apps may already have SCIM enterprise plans, for which you've already gotten some sort of provisioning out of the box. But everything else that you didn't have provisioning out of the box for, that's what we started solving from day one.

So our first milestone was having, let's say, 10 of those apps go live first with our first early customers. And then we scaled that to now around 220 and north of that. So that's how we have shaped our own product from the beginning itself.

Jim

Is there a typical profile of a company that tends to experience these problems the most? Could you give an example, without shaming anybody of course, or calling out anybody by name? But do you have examples of the types of clients that fall into the exact description that you're talking about earlier?

Pranay

Sure. The way I look at it, there's a floor and a ceiling to the kind of customers who face this the most. The floor is people who have outgrown their SSO and they just can't deal with spreadsheets and manual tracking anymore. That's the floor for a lot of our customers who come to us, and this could be anywhere between 50-employee companies and 30-employee companies, depending on the volume they have, right, and the scale at which they're growing.

So that's usually the floor of the customers where we get in. And the ceiling is usually where people are like, I'm not okay with a two-year enterprise bloated IGA program. So that's the kind of sweet spot that we operate in: people who have outgrown their SSO and spreadsheets, all the way to people who are like, okay, it still doesn't make sense for me to go for a two-year enterprise IGA. So that's the sweet spot we operate in.

Jim

Yeah. When I think about identity governance, right, it's kind of baked right into your name. But identity governance has always been around who has access to what. I think there's two sliders when it comes to that. When you say access to what, it's like, how broad do you go? And then who has access, like what does that access mean, how deep do you go?

Let's focus on the deep first. I mean, how deep do you have to go to be effective at governance?

Pranay

Almost immediately, right? So, "Pranay has a GitHub account" is a fact. Like, I have a GitHub account, so I can do something. But "Pranay can push into the repo and then approve my own pull request", that's a risk.

So authentication only takes you so far. Think of authentication as getting in through the door. Just ensuring that Pranay is able to log into something, or Pranay has an account in GitHub, is probably just a starting point for you. But what can that person actually do once they're inside the app? That's where the risk lies. That's what the attackers and auditors and IT teams are actually concerned about on a daily basis, right? So that's how deep you have to go.

For example, an access review. You could have an access review that typically says, okay, yeah, Pranay has access to Salesforce. But that's not what the auditors want. The auditors want to see: has that person been reviewed to have access to, let's say, download all the records of their CRM, or even make changes to an invoice or something like that, right?

So that's the depth we're talking about, and like I said, it's important from day one to answer that question. You can't be like, okay, one day I'm going to just wake up and now I'm going to track everything else, because it doesn't make sense without...

And that's actually the beginning of, like I'm sure many of the listeners might already know this, the beginning of identity was governance. It's not authentication. For example, if you go back to the early 1990s with the whole Enron and all of those things, right, the first thing that came out in the market was, hey, we need to ensure that everybody had the right permissions, and there shouldn't be any segregation of duty or policy violation or something like that, right?

So that's what I'm talking about. You have to ensure that the permissions that somebody has inside the software, or inside whatever you want to manage, are accurate and reviewed by someone. So that's actually the genesis of governance as well.

Jim

Yeah, I think you're right on. All those points are buzzing around in my head, but let me ask the contrarian question, which is: is there ever a time where just figuring out what apps somebody has, at just the top level, yes or no, is that ever useful?

Pranay

I'm sure. I mean, you could just need to know that if somebody has asked you to calculate your SaaS spend, for example. If I have a Salesforce account, I have a seat in Salesforce that you're paying for. So if you want to track down your spend and your SaaS licenses and stuff like that, maybe that's what you need. You only need to know whether that person has a license or not.

But then very quickly you start asking questions like, okay, is the person actually using that license or not? So that's where the usage again comes in. So yeah, just answering whether someone has access to a certain app, I would say it's a very simple question to ask that's often too simple for a real-world scenario.

For example, if your CFO asks you, hey, can you give me a list of all the people who have access to Salesforce so I can calculate how much are we spending? That's probably the only time you'll only need to know whether the person has access to Salesforce. That's it. Every other use case that actually comes to my head, you actually need to know what they can do inside, or how long they have been using it, or do they even use it in the first place. So something like that.

Jim

So Pranay, I'm just thinking back to my days running IGA systems as a practitioner, and we have thousands of applications. I think there are certain applications that we have to go a mile deep, like I want to know every table that an account has access to. But then there's the travel system, and I just want to make sure that when this person leaves the organization, I cut off their access.

So in your opinion, is there a time where doing that simple yes or no, someone has access, does that make sense?

Pranay

I mean, of course it does make sense. It depends on the use case you are trying to solve, right? If the use case was something as simple as, like you said, maybe just offboarding, the only thing you need to know at that point is, at any point of time, what does the person have access to, so you can actually go ahead and clean it up. So like you said, offboarding is a very good example of something that could just be solved if you know in real time what the person has access to.

So I think that's a great use case. The other use case that comes to my head, where you are okay knowing if a person just has access to certain things, is maybe to track down your licenses and spend. For example, if your CFO asks you, hey, can you tell me what are the users in my Salesforce, so I can track down how much are we spending on Salesforce? That's probably the other time where the team is just okay knowing that, okay, in Salesforce I have 10 users, and these are the users who have access to Salesforce. And that's all you need to know, because that's what the CFOs ask for.

So I think there are a few use cases where just knowing that someone has access to a certain thing is enough. But very quickly, the follow-up question to that needs you to know what's beyond that. So what does the person have access to, how long they have been using it for, or are they even actively using that license or not? Things like that.

Jim

Yeah, it seems like the way most systems get integrated today is if they have a SCIM connector. But that doesn't always get you the depth of information that you want to make intelligent decisions about access either, right? So where does SCIM fit into the picture, and how do you go beyond that?

Pranay

So SCIM, like you know, was built to manage users, ideally, right? To create and remove users in a third-party software, right? And just by the definition of that, the original intent of SCIM was to create and remove users and maybe a group attached to it.

But that's where the problem is, like you said. It's very hard to map, let's say, a group in your identity product to an actual permission in the app, because the app has built a very fine-grained RBAC in itself. To establish a link between, let's say, your IGA group and that entitlement, or that entire fine-grained RBAC, that's a challenge, right?

And that's why SCIM is probably not enough, and SCIM doesn't go deep enough. So we borrow this technology from SCIM, and we actually call our connectors SCIM++. Think of our connectors as SCIM, but we can actually go deep into every entitlement, all the way to the last mile of entitlements. So think resources, modules, repositories, channels and stuff like that. And that's how we can go deep into each app as well.

Jim

The SCIM++ idea, that's kind of fascinating, right? To me, that's where this depth of coverage really starts to get interesting. So can you talk a little bit about what that means in the real world and give us an example?

Pranay

Absolutely. Maybe I'll give you two examples. Let's take an example of a very classic SaaS app that most companies use. This could be your vendor procurement or vendor management platform, right, like your procurement software.

Now within that app you have different roles for each user. Your user could be in the invoice team, they could be on the review team, they could be on a bunch of other teams. Within each of those roles they have certain permission sets. For example, this person in the invoice team can raise an invoice, but they cannot approve that invoice. Now the same person in the invoice team can also be a reviewer, an approver, and, let's say, they can even have permission to delete that invoice in the first place.

So that's the depth I'm talking about. A typical SCIM would have just let you create a user in that app, saying, hey, I want to create Pranay and I want to add them to this vendor procurement platform or whatever, right? But what you actually needed in the first place is, when an employee joins your company, you want to add them to, let's say, an invoicing team within this app, and you want to give them access to only read or update that invoice, but not to approve it or delete it. So that's the depth I'm talking about.

And this could also be extended to, let's say, GitHub, the most common engineering tool, right? Within GitHub, you need to ensure that the person is added to the right team, and within the team they have access to the right repository, and within the repository, are they the maintainers of that repo, or are they just a committer to the repo? You need to know all of these things to actually do governance, right? For example, when you're provisioning someone in GitHub, you need to know to what depth you want to go to give that access to that user. So that's the depth I'm talking about.

Now let's switch scenarios and talk about an internal app. Internal apps are where it gets really messy, because an internal tool is a lot more deep into your business. There's a very fine-grained RBAC usually built for internal tools, from our knowledge. For example, you could have an internal ERP that you have built that serves 10 of your teams, and each team has different access to different tables within that system itself. And then within each table, you could have a reader, writer or whatever, right?

So look at that depth. If you were to build an RBAC for this system, that probably will go three, four levels deep. But will that system talk to an IGA tool that only speaks SCIM? I don't think so. You need to go deep into each app, all the way to the last entitlements, and that's what you want to manage. And that's the hard part, really.

That's why I always say to people, when they ask me this question, authorization is a much tougher problem to solve than authentication, just because of the sheer size of it. RBAC differs app to app. It can go so deep in some apps, it can just be very surface level in many apps, right? So it really has to match the actual schema and the entitlements that you have. And that's why integration is so difficult to build, and that's when we started solving that, actually.

So the whole idea behind Iden was, hey, we know that identity governance has been there for three, four decades now, right, almost. And yet coverage is a single point of failure, because your governance program is only as good as the depth you serve and the breadth you cover. So that's the actual problem we are trying to solve.

Jim

And I love how you went into the depth. So let's talk about the breadth a little bit. If I'm an organization, they have at least hundreds of apps, if not thousands of apps. So what do I need to govern out of that? And then how does Iden solve that problem? Do you guys have a lot of connectors, or what's the approach to reaching that broad number of apps?

Pranay

Right. So like I said, if a customer or a prospect is thinking of a governance roadmap, my only advice to them is: your governance roadmap is only as complete as the weakest app in your stack, right? Because it doesn't matter that you got the top 20% right. It always matters if you got the remaining 80% of it right.

You could have very strong governance workflows, LCM workflows, set up for your Google Workspace, Microsoft, Slack and a bunch of the apps. But from an audit perspective, or even from an IT operations perspective, until you solve the long tail of usage in your workspace, for example all of the apps that your sales team might be using on a daily basis, unless you cover all of those things, your governance probably is not good enough. So that's what it comes down to.

For example, when you're onboarding an employee, is the employee able to get access to everything they need on day one, or they have to wait for a few days to request access tickets and get them approved? If you're offboarding someone, are you cleanly removing them from every app that they had access to, or you're hoping that by turning off their Google they'll not be able to log into something else that they might have had access to? And from an auditor perspective, your audit scope might include ERP and a bunch of those apps, and maybe the governance today for all those apps is driven by spreadsheets.

So yeah, breadth is as important for a governance roadmap as depth is.

Jim

Yeah. And so when I think about the Iden approach, when I have a spreadsheet of apps, is it like a bingo chart of: here's which ones we have connectors for, these are the ones we need connectors for, Iden will supply connectors for? How does that work?

And then how long am I typically looking at, from the time that I start connecting an app to the time that it's actually connected?

Pranay

Right. So usually when customers come to us, we start off by looking at the real estate that they want to manage. That could span maybe hundreds of apps in their stack. Now, not all of those apps are equally important from a governance perspective. Some of them may be in scope for audit, so they probably make the P0s, because you have to govern them first.

Some of the apps may be crucial for you to ensure that data is secure, so no ex-employees, or everyone who has access to that has the right least-privileged access that they should have in that app. So there are multiple scenarios which make a certain set of apps P0, and the same way you can extend it to P1s and P2s.

So when we take a customer on board, the idea is we go through the list of apps and we give them a few heuristics to identify which of these apps are P0s for them. And they mark in the list of apps: hey, these maybe 30, 40 apps are my P0s, because I'm concerned about the security or controls, or just access to those apps, or even cost as well. Sometimes some of these licenses are very expensive, so you need to ensure that someone who doesn't need access to that app doesn't have a license in that app. So it could just be cost as well, right? So that's how we constitute the P0s and then P1s and P2s.

And then on our roadmap, the way we go live, our first milestone is to get all the P0 apps live on the platform. Now, the way our connectors work: I think I mentioned earlier in the episode that we already have a catalog of around 220, and maybe a little more, connectors already built right now. These are all connectors over and beyond the SCIM plans. If you have a SCIM plan, we can onboard them in just a matter of seconds, because all we need is a SCIM token and a SCIM endpoint. That's it.

But the real puzzle is, okay, how are we able to do apps which are not on a SCIM plan, right? Those are the apps I'm talking about. Our catalog is around 220-plus for all the non-SCIM integrations that we have built. So when the customer gives us that list of P0 apps that they want to take live, we map it against the ones that we already have, and then our time to hit those apps and go live with them is a few minutes maximum. If that app is SCIM or API or something else, it could just be a setup that takes us a minute per app. And if it's a non-API app, then it probably takes us a couple of minutes to go live.

One of the things that I would love to share here, and that's some sort of badge that we wear with honor, is that for most of our onboardings that we have done in the last, let's say, one or two years, in the first hour of going live we are able to hit around 10 to 15 apps for every customer. So that's the speed I'm talking about.

Now let's come to the other part of the question that you asked: if we don't have a certain app in our catalog already. That's where, again, our claim is that we can build those connectors for you in less than 48 hours. And I picked that number because I want to really stump people when I say that sentence. I want them to feel like, are you really going to be able to do it in less than 48 hours? And the answer is yes.

So that's the other claim. When we onboard a customer, we take them live with the set that we already have, and for everything else that's missing in our catalog, we can build those connectors for you in less than 48 hours, no matter what the app is. It could be a SaaS application that you want to manage, it could be an internal tool, or it could be a legacy mainframe or anything else as well.

Jim

So this is what has excited me about the Iden platform as we've gotten to talk: you've got the depth, like you can really drill into each app you can connect to. You've got a lot of out-of-the-box connectors, you've got platform connectors like the SCIM connector, and then that 48-hour turnaround on any other connector. You really can integrate your entire enterprise with relative ease. That's an exciting future.

Let's go a step beyond the breadth and depth conversation about connectors and talk about identity governance. I'm wondering, from your perspective, what does governance mean from a functional perspective? Is it access provisioning? Is it identity correlation? What are the parts and pieces of the functionality?

Pranay

I think governance, for me personally, is just the same question being asked at multiple different times in the journey of an employee, right? Should this person have access to this thing right now? That's the question. Or, who has access to what? You can ask the same question on day one, and that's your joiner problem. When a person joins the company, what should they have access to?

On, let's say, any good day when the employee is working their job, the person says, hey, I need access to this, right? Should this person have access to this? Has the manager approved it? Should we provision the access for this person? That's the mover problem, or the day-to-day access ticket problem. When someone leaves the company, again the question is: should this person have access? No. So remove it.

So I think it's the same question being asked multiple times through the lifecycle of the employee, or the governance journey of a company. And that also flows down to the auditor question. The auditor needs to know whether you have actually reviewed every person and their access to everything else. So like I said, governance is just answering the same question every single time: who has what level of access to what? And then everything else around that is also part of governance.

Answering that question is one thing, but actually doing it is also an important part of governance. That's why I put governance in two things. One is, of course, the actual governance, where you decide if every access looks right. And the second part is orchestrating that access: for example, provisioning it, deprovisioning it, granting it, reviewing it, all of those things, right? So these are the two parts of governance, the way I look at it.

Jim

So you rightfully pushed back on my last point, right? Because I was talking about functionality in terms of the traditional silos. I think what you're saying is it's all part of one big process. And I encourage our listeners to go out, it's idenhq.com/idac, and see what this kind of integrated approach to access governance looks like.

But I think you're right. All these are different flows leading to the same goal, which is ensuring that the right people have the right access. So I think that's really good.

But I still want to pick on some individual traditional workflows, like the service desk ticket. So a service desk ticket comes in. Talk to me a little bit about when it makes sense to automate it versus go ahead and create another ticket to go ahead and provision the access. It's harder, I think, to automate in traditional IT, but man, it can really save a lot of time.

Pranay

Absolutely. I think that's the thing, right? Automation is not the solution for everything, because automation only works for scenarios where you can predictably and deterministically identify what the outcome should be. Everything else that you don't know of, or that you haven't encountered before, probably doesn't need automation. It needs a human decision first.

So the way I would look at it is, when you're seeing something for the first time or the first few times, you should ideally just keep that human decision in the loop. You should always have a human in the loop, because somebody has to review it, somebody has to decide if this thing looks right or doesn't look right. And that's where the tickets actually come into play.

So you're absolutely right that when an access ticket first comes in, hey, I want to give this person this access to this application for this duration, somebody has to review it and be like, okay, this sounds correct, because I know the context, I know why they need their access, right? So that's why the human decision is still relevant. Automation should kick in only when you have seen repeatable patterns over and over again.

For example, if the same kind of tickets start showing through the door every single time. If multiple people are requesting actions with the same permission style or template, or if the same action has to be repeated, let's say when you're onboarding someone you need to migrate their Google Workspace data to their reporting manager, or you want to migrate their Adobe data to their peer in the group, or their team lead or whatever, right? When you have seen those repeatable scenarios and workflows in action, in practice, that's where I think things should start being automated. And that's usually what happens.

At a smaller company, when they're 50 employees or 30 employees, they would know by now what an onboarding looks like. For example, when you onboard an engineer, what do they need access to? They know it by now. When someone leaves the company, they know already what it means, what's our checklist that we want to automate. They already have that checklist in place as to what that offboarding looks like. They already know what an on-call scenario means for AWS. For example, when somebody requests access for on-call, I need to give them access to the S3 bucket, I need to give them the right access to the logs and everything else, right?

So by the time they have figured out what these templates look like, that's when the automation should start kicking in. And that's where tickets stop being manual and they start being automated.

Jim

Yeah. Talk to me a little bit about the experience, like the end-user experience. And I'm thinking of the IGA team, because a lot of our listeners are the IGA team, and they're going to be using this tool. They may look at Iden when they are considering replacing a legacy IGA. They need to bring a lot of their processes, modernize a lot of their processes.

It sounds like you've got a lot of that capability in place. I mean, all these great things that you do have got to be technology-enabled, but put that aside, just talk about the experience a little bit. What is the flow like for an IGA administrator on Iden?

Pranay

Right. So like I said in the previous answers, it all starts with connecting your real estate to Iden. For example, if your tool today looks like 20% or 30% of your total stack that you want to manage, switch to Iden, because we can take that coverage from 30% to, let's say, 100%, right? So that's the first thing they do. They bring their real estate, all of their apps that they want to manage.

That might take anywhere between, let's say, a few hours to maybe a few weeks, because we might be dealing with some internal tools, documentation for which may be missing, or maybe it's present, or maybe it's even some mainframes or legacy tools, right? So the first few weeks go into just bringing your real estate on board to Iden. So that's the first thing that we do.

The second thing we start doing is looking at the most critical workflows that the IGA team is concerned about today. For example, for some teams audit and offboarding is more important because of the regulation or whatever industry they serve, right? For those people, the first thing that we start solving for is: hey, how do we ensure that the offboarding workflows are intact, are airtight? So we are actually offboarding across all of the apps. We're actually producing evidence that the actual task was done in place. We need evidence of that, because auditors don't want evidence of you initiating a task. They want evidence of whether the task actually finished and completed as expected.

So we look at the most critical workflows. It could be offboarding. It could be a user access review. For example, they could have a SOC 2 or an ISO or a HIPAA or any of the compliance frameworks coming up. And then, hey, we have been reviewing things via spreadsheets, or via CSVs flowing around the company, and people replying to email threads saying, hey, I approved all of that. They want to streamline all of that. They want to be more good-looking in front of the auditor: hey, look at our process, it's all under control. So maybe for those teams that's the most important workflow, so that's what we do first.

So once you bring on board the real estate, the next thing we do over the next few weeks and months is ensure that their most critical workflows are set up. It could be offboarding, it could be your audit governance, it could just be your onboarding. For some people, day one onboarding is a great experiential thing that they want to offer to their employees. Maybe the CIO's top of mind right now is: hey, I want to ensure that when employees join my company, on day one they should have a smooth-sailing first-day experience, right? So maybe that's the foremost thing for them at that point.

So usually we just mirror whatever their requirement is, whatever their roadmap looks like, and we try to mimic and set up that thing within Iden. And of course, over time the goal is we keep adding more apps to your real estate as you bring on board your P1s and P2s, and also keep adding more workflows on top of it.

So your entire setup evolves over time. What you start with is, let's say, 10 apps and the most critical workflows. That's your P0 apps and your most critical workflows. Then that evolves to P1s plus some workflows, and then P2s plus more workflows. So that's how it evolves over time, essentially.

Jim

It sounds really interesting. I know personally I want to see a demo. We have to talk about AI, right? I mean, sometimes we think this is the podcast called AI at the Center. And I want to talk about the impact of AI in terms of what does it mean from a product standpoint: how are you using AI to build the tool, or how do you make AI-enabled tools available for your administrators, but also how does AI insert into the governance processes?

So in other words, the end user, which in most cases, unless you have self-service, the end user is your identity practitioner, right? But how are you inserting AI into the governance process today? And where is that not effective? Or put better, where does the human element still make sense in governance?

Because I think where AI really becomes valuable, from my perspective, in the conversations I'm having, is you've got to get those critical decisions that a human needs to make to the human, but you can't overload them with a bunch of stuff that they just rubber stamp. What do you think?

Pranay

Absolutely. I think spot on. The way to look at AI across governance is, again, two ways: AI for governance, and then governance for AI. What you're talking about is AI for governance, so let's talk about that a bit.

Interesting thing, right: you said that the end user is a practitioner. For an IGA team, the end user is always the end employee, right? So when we talk about AI for governance, what we should really look at is: how should we use AI? Of course to make the lives of the practitioner easier, because that's the goal of AI, right? And also to ensure that the end-user experience improves with AI. That's the whole point. Otherwise, if your IGA tool still needs you to go to a very clunky-looking dashboard and raise a ticket and then wait for two days to be approved, I think you lost the whole plot here, right? So that's the way to look at it.

Now let's talk about those two scenarios. If you talk about using AI for making the lives of a practitioner easier, we solve it in two ways. Number one, of course, the whole idea of getting your apps onboarded to your IGA platform and expanding your real estate is the first thing we use AI for. This industry has typically been very manual and very sluggish, in a way that it takes weeks and months to build those connectors. You go through the entire hassle of scripting it yourself, and then stitching it together, and then running through tests and QA and everything else. So we shrink all of that, and that's how we're able to achieve the less-than-48-hour timeline for building those connectors, right?

So that's the first thing: you need AI to build those connectors and go live with your apps with your IGA tool. That's number one. That's, say, half the problem solved, because that's where you were spending a lot of time, and that's where a lot of practitioner time was going in.

The second part of that was: okay, should a practitioner still be required to go into the dashboard every time, look at everything and remember what it needs to do and then do it? Or can we use AI to make that a little simpler? So that's what we started doing in the last few months. We started exposing the whole functionality to your LLMs directly, because LLMs are very good at tool calling. They might not be very good at deterministically doing things.

For example, you can expect to go to Claude and say, hey, can you remove this person from this? But you're not fully sure, and it's not auditable. It's not deterministic whether it's actually able to go and delete only this person and not someone else with a similar name. So you can't know that, right? So while an LLM is great for orchestrating a lot of things, if you have deterministic tools built for an LLM to actually call, instead of doing things by hand, that's a problem half solved.

So that's what we started doing, right? Bringing a lot of our IGA functionality, exposed to an LLM via MCPs or tool calls, whatever you call it. And the idea is, if you say to an LLM, hey, this is what I want to do, it can actually go ahead and do it deterministically this time, with an audit log attached to it. So that's the two ways in which we are helping the practitioner improve their IGA experience.

The second part of that was, of course: how do we use AI for improving the end-user experience? And that's why I think many, many companies are trying to attempt that. People are trying to say, hey, why don't I let the end user just describe to me what they want, and then I'm going to create a ticket on your behalf, or I'm going to call the workflow for you. So that's the other part.

Of course, you have a lot of tools, including ours, where you can just go to Slack or Teams and say, hey, I need access to AWS as I'm on call, right? And you should automatically infer what that means. You should be able to check my PagerDuty, you should be able to check my on-call tool to see if I'm actually on call and how long I'm on call for, and then give me all the relevant access attached to it, for example S3 access, audit log access and all of those things, right? So that's what we are doing as well.

So to an end employee, we just look like a very nice-looking tool that is available via Slack, or you can come to our dashboard and ask the ticket in plain natural language, and we do it for you. So that's the way we look at AI for governance, at least till now. And that changes every week now.

Jim

It seems like you could have an unlimited number of use cases. But if AI can help from a governance standpoint, like, hey, I noticed this thing over here, like Pranay's gotten access to 100 new applications in the last week, hey, look at that. In other words, it's picking up on the signal and then letting the team know, like, hey, this is something potentially fishy.

Or during user access reviews, like, hey, these are the types of accounts you really want to look at, or the types of permissions that you really want to look at. And these ones are low risk, and maybe we can rubber stamp those because they're not as risky.

Pranay

Yeah. Yeah. Thanks for filling that up. Actually, I missed that whole thing altogether. Yeah.

Jim

So the other thing that is obviously hot right now is doing governance for AI agents and their access. I still feel like we're at the beginning of the wave for this. I mean, you've got Meta's Muse and the OpenAI bots and GrokBot. Now, I think by this time next year, every enterprise user is just going to have a personal agent that does all their work for them.

But hopefully the way that plays out is we can all get five times as much work done, right? Or we can go five times faster than we were going before. It's definitely going to add a lot of productivity, but I think it potentially adds a lot of security risk for the enterprise, where you have these agents that have so much access. So what is the answer from a governance perspective?

Pranay

I mean, there have been several attempts to solve this, and there are many NHI management tools that claim to do this. For example, they claim that, hey, we can actually catalog every agent that you use, and we can catalog what access they have and all that. But in reality and in practice, every security practitioner you speak to will tell you the same thing: it's hard. It's hard to track them, for many reasons, right? Let's talk about some of those.

From my perspective, the first reason why that's hard in the first place is the agent inherits identity, or it uses a delegated identity from the user itself. So there goes your user attribution, right? For example, you can't tell for sure whether it was me who did that action using my automated deterministic workflow that I was in control of, versus it was my agent who just took that same path and went on to do a certain thing. So there's no attribution between a human and an agent today, because they all use the same identity underneath. That's number one.

Number two, the same identity, or the same credential, is now being used across multiple surfaces and multiple tools. Just yesterday I was reading on X that there are a lot of developers who keep ranting about this: whenever they start a vibe coding project, or whenever they start an internal coding session, they have to ensure that the credentials are never in memory, they have to rotate the credentials and all those things, right? All of that is because they know that the same credential or the same identity is being shared across Codex CLI, some agent that you build, some agent that the enterprise has offered, everything else, right, across multiple surfaces. And there's a problem, right? Because if you use the same identity, to the end vendor application it looks the same. So there's no way to govern them again.

And then finally, all of the identities today are based on the same long-standing API tokens. And once they are in memory, then of course it's visible to the agent, and the agent can obviously call it and use it whenever they want to.

So I think those are the three reasons why I feel governing agents is hard as it is today, and it needs a fresh look. And that's what we're trying to attempt as well. But yeah, you're absolutely right. It's very hard to govern agents today, as the current standards and the current industry stand as of now.

Jim

Yeah, I mean, in identity it's always been the technology that creates the identity use case, and then the identity technology has to fit that use case. Right now these personal agents are coming. I'll tell you, we've known about the personal agents for a while, and every expert that I talk to, like yourself, says, well, the personal attribution issue exists, so don't use the same identity for your agent.

The reality, though, the practicality of it, is like, okay, well, that's what's happening. So I don't disagree with the cure. I just, no one's invented that cure, you know? It's like, oh, to cure cancer you just need to do this, and this isn't available yet. Okay, well, let me know when that's...

Pranay

Yeah, we're trying to change that, though. We are attempting a way to make it work with the current setup that people have, so we don't have to wait for another decade until the standards catch up and then every vendor in the world has enabled that and all of that. So yeah, great point.

Jim

So I'm going to switch topics here. I had a recent conversation, and the person was a fellow identity nerd like us, and they said one of the biggest things that's changed in the last decade or more is that you can't go into a board or to your executive management and pitch a year-long identity project anymore. Nobody has the appetite for that. They want results quicker.

So that's the lead-in. Obviously I'm setting the expectation. Talk to me about a typical Iden implementation. What are we looking at here?

Pranay

Yeah, I think the way you mentioned it is the exact same way I usually put it. To a practitioner, or the leader of this governance program, they should always be looking at the ROI of the program. And the way I define ROI, and again different people have different ways to look at it, the way I look at ROI is the coverage you get for every dollar you have invested in the program.

So for example, if you have gotten the breadth and depth of coverage for every dollar you have invested in the IGA tool, as well as the SI fee or the running cost and everything else, that's the ROI you should be looking at. And that's what we're optimizing for at Iden. Like I said in the earlier questions, the goal is to give you maximum ROI.

What I mean by that is to ensure that you get the best coverage out there. You get the best coverage for depth and breadth across your entire stack, and in terms of the cost you have spent, reduce the cost of the IGA tool itself. So that's one of the pricing levers we use. We are very cost-efficient as compared to, let's say, a legacy IGA peer, because they're also like modern companies and all that, right?

So of course the cost of running the program is lower with us, because the cost of the IGA tool itself is lower, and there's no specialized running cost attached to it. For example, if you have an SI, we're happy to work with the SI to make your rollout faster. But if you don't have an SI, you don't need an SI to do this in the first place. So that's how our roadmap looks like.

Our goal is to give you the best coverage and reduce the overall running cost, or, the way people put it, the TCO of the entire program, to reduce that cost of ownership, right? So that's what we do. And in terms of timelines, the first few months go into setting up the most critical apps with the least effort that's needed, and then expand over time.

Jim

I love that answer.

Pranay

Totally with you on the point that governance should not span years, because it defeats the purpose. It's like saying, hey, I'm going to give you the ultimate security, but I'm going to give it over three years. But what about the attacks that I'm going to be exposed to in the next three years? I mean, governance has to be point in time: you get it now, and you get it as soon as possible, and then you are secure and auditable and everything over the next few years.

Jim

Yeah. I mean, I think that companies are going to have to, you know that magazine Fast Company? I kind of think that is the exact right term of what every company's going to need to become.

I remember the days where you spend two months getting the infrastructure in place, installing the software, racking the server, stuff like that, and then going through different boards, like the architecture review board, and you couldn't get on their docket for two weeks. And yeah, it would be six months to a year by the time you did your phase one, and your phase one would integrate, let's say, five applications. That dog don't hunt, to use a colloquialism. That's just not going to work these days.

So Pranay, I'm really excited about what we learned here today about Iden. Again, idenhq.com/idac is the place to go, get more information, book a demo, things like that. Pranay, thanks so much for coming on the show. I'm assuming you're open to people reaching out, connecting to you on LinkedIn.

Pranay

Of course. I mean, yeah, of course. I'm available to connect with people, happy to speak to more practitioners like yourself. I think it's very important for the identity community to come together and share their expertise with each other, and just keep evolving faster than the pace of some of our vendors. So I'm totally in for learning and growing every day. So happy to chat with more people.

And thank you so much for having me on the show. As I said, I love your show, and that's the only identity podcast I listen to. So it's always a great feeling to speak to a practitioner like yourself.

Jim

So a lot of our listeners would be disappointed if we didn't do a lighter note question. And so my question is: if you were in the situation and you had your firstborn who is finishing high school or finishing college, or maybe re-emphasize this over time, what is the piece of advice that you would give them, or the habit you would instill in them, for having a successful life?

And I thought about this a little bit earlier today, so I have a little bit of a head start. But I've always been a goal-oriented person. I have long-term goals. I know what I want to do when I retire. I have goals for the year. I even have goals for the week, and for the day, like, here's what I want to accomplish. And I make lists, and then I try to make sure I'm knocking those things off. And that's how I gauge my success in life.

Now, I know that maybe that is a little bit over the top, but I think being goal-oriented is very important in this professional world. How about you? Do you have any habits that you do that make you successful that you'd want your children to pick up?

Pranay

I mean, we can talk about it for as long as I've been... But I was just talking to someone, so it's actually fresh in my head as well. My mantra for life has always been: I never typecast myself. For example, I studied computer science and I was supposed to be a techie, but I went on to work at an investment bank, and then do consulting, and then venture capital, and now a founder, right?

So the one thing, my mantra for life has always been, and that's what I want to tell my firstborn as well: you shouldn't restrict yourself into one typecast that people have given you or you're attached to. You should totally be in for trying things that you love the most and that seem exciting for your life, because all of that compounds over time. Your experience across everything that you have done or seen just compounds and makes you a better person, in a way. For example, you know exactly what it means to be in a certain setting and be in a certain room.

And that's what I would advise them. Basically, don't restrict yourself from doing this one thing that you have been told to do, or that people around you are doing. Just feel free to explore what interests you. And the second most important thing is, don't just feel free to explore it. Try to excel in it. Put everything you have and everything you can into achieving excellence in whatever you do. Keep pushing the boundaries, show up every day, do harder things in life, keep pushing the boundaries, and I think you'll be good.

And I did the same. Like I said, I used to be a software programmer back in the day. My first job was something else, second job was something else, third job was something else, and now I'm doing completely different from what I've done in the past. So yeah.

Jim

I love it. Get a variety of experiences. It makes you a more complete person. I guess we'll leave it there. Thank you so much for doing this, and thank you for supporting the show.

Thanks to everyone who tuned in today. Again, it's at idenhq.com/idac, and then you can connect to Jeff and me on the web at idacpodcast.com. We're always open to LinkedIn connections. Download the podcast, leave us a like or subscribe, leave us a comment. All those good things are always appreciated, and we will talk to you all on the next one.

57:15 · Identity at the Center #457 · More from Identity at the Center

Frequently asked questions

Episode 457, a Sponsor Spotlight released on October 28, 2026. Pranay Yadav, Iden's co-founder and CEO, talks with host Jim McDonald for 57 minutes about identity governance, SCIM and AI agents. It's on YouTube, Spotify and Apple Podcasts.

SCIM++ is what Iden calls its connectors. SCIM creates and removes users, sometimes with a group. SCIM++ does that and goes on to the entitlements inside each app: roles, modules, repositories and channels. Access is governed at the permission, not only the account.

Twice, Pranay says on the episode. Offboarding, where you need a live list of every account to remove, and a license count for the CFO. Every other question, from access reviews to audits, needs to know what the person can do inside the app.

Sort them into P0, P1 and P2. An app is P0 if it's in audit scope, holds data where least privilege matters, or has a license expensive enough that an unused seat costs real money. Most teams land on 30 or 40 P0s and get those live first.

Three reasons, from the episode. Agents act with a person's identity, so you can't tell the human's actions from the agent's. One credential is shared across many tools, so they all look the same to the app. And they run on long-lived tokens the agent can use whenever it wants.

One email when a new issue goes out, nothing else. By subscribing you agree to our Privacy Policy.

Dealing with this now?

Talk it through
with Pranay