Intelligence at Scale: Okta’s AI-Augmented CS Playbook

Thanks for having us here today. We are the Okta team, as Kalpano had mentioned, and we're excited to talk to you today about our Okta AI Augmented CS playbook. So by way of intros, just a quick second intro on ourselves. I'm Alana Stolzus, I lead our digital success team here at Okta.

I've been Okta almost six years, and I was one of our first digital CS hires, and we have since grown since 2020 to a group of nine digital CSMs. I want to over thank Melissa. Let's go to... Lawrence.

Hi everyone, Lawrence Leong. I'm disappointed there's no other backcountry skiers here in this group. You would think there would be some with a group this size. But I lead what we call scaled customer success.

So I have a team of about 50 pooled CSMs that we use to cover about 11,000 medium and long tail customers. We're just tripping people out with a slide. There's a ghost that's changing the slides and promising them. So I'm Melissa Allen, so I've been in Okta for seven years in March.

I lead our customer success strategy and operations teams, and I have been to 11 pulses. And if I said Vanilla Ice, if you know who was there, that was like an awesome pulse just saying Vanilla Ice. That's my intro. All right, so before we kind of get started, it's always kind of nice that Dint works yourself against who you're listening to.

I know we're always going to do that. What about my team and my org, and how does that fit to what is being presented? I wanted to give you kind of like, oh, can you go to one more? I really did not press it.

I promise you. Yeah, it is just being, there's a ghost. It's being spicy, y'all. It's being spicy.

So I just wanted to give you kind of a little bit about Okta. We've got about 20,000 plus customers, 6,000 plus employees. And this is kind of like our model. We've got kind of this digital success only base.

We've got a silver scale, which is Lawrence. We have a gold scale. We kind of have a gold covered, and then we have like a dedicated. So we definitely have our range of different customer success and how we engage with them.

And because of that, we have different areas that fit each of those needs. So we just kind of wanted to outline that for you before we got into everything. All right, and then what are we going to be talking about today? So we're going to be talking about some business challenges, our kind of our current state discovery.

Like what were we, where were we at and what were we trying to do? How we had an AI augment at CS and how that evolved and how it is evolving because it's a never in, it's, you know, an evergreen type of situation. And then ultimately partnership and collaboration. How does cross-functional partnerships really help this thing take off and are the reason that everything ends up being super successful?

All right, so I'm going to walk us through if the slides work as they should. I will walk us through a little bit about what you're today. So first we're going to talk about the nation is really that foundation for powering AI, AI augmented experiences. And what that means is you cannot equip CSMs to do more with AI unless you have a strong self-service foundation and allowing those CSMs to have more strategic engagements.

By building that self-service piece for the less strategic components. We'll also talk about what is that augmentation layer of AI on top of the CS motion. And then lastly, we're going to talk about how we build that digital customer experience and the unified customer experience across CS ops, across scale, across digital. But also how do we think about where that agentic layer fits in between all those different pieces?

So to start us off, I will talk about what are the challenges that we had that we needed to address. So first we had a cost. We have a large number of CSMs and with those CSMs, we needed to move a lot of those CSMs upmarket to help our larger customers. And we couldn't necessarily fill that gap with more CSMs.

And we wanted to make sure our customers still got that best-in-class service that they've had in the past. And so we had to think about how do we help those customers continue to be as successful as they were before, but with a little bit less human touch. Sorry, the slides are not building. We also needed to figure out how to manage risk and growth at scale.

Meaning how do we detect when a customer is at risk well before they're at risk? And how do we detect that that customer is a growth opportunity well before that the customer is able to grow? But we don't have necessarily that human pulse on the account every single day. We also had to scale digital.

So we've had a pretty robust digital motion for the past five years or so. And with that, we have hundreds of email variations and hundreds of in-app guides. And we have so much out there because we have such deep personalization. But it doesn't scale at a certain point.

And we want to be more predictive. We want to be more agentic. And today, or at least in the past, we knew that we couldn't manage all that overhead. And then we also had disconnected motions.

So we had our digital motion. We had our scale motion. We had our human motion and soon to be our agentic motion. So we had to figure out how do we bring all these things together.

So in order to figure out how to bring all together, we wanted to see what's the lay of the land. What do we have in front of us so that we can inform what we're going to build? And you'll see this journey is actually a map that we've had since about 2020 on where we wanted to go with digital. And we started very, very low on this chart where we wanted to be able to have a personalized journey.

We wanted to give them guidance based on what they have or have not done. And the step that we're taking today is how do we not just be personalized, but also be proactive with our customers and get ahead of what they need before they even need it. And the ability to respond to those customers so that it's not just a one-way system, but a collaborative interaction. So we're continuing on this journey, but I would actually say we finally need to probably rebuild this chart and get to that next level.

In order to do that, we wanted to see what our CSMs are actually doing today. We needed to figure out what do we make sure we can provide for our customers still if we're going to take away that human motion in certain segments and bring those humans up market. So we analyzed something like 50,000 Gainsite timeline entries. Great use case for just using Claude throughout your company is to see like how can you take tons and tons of data and synthesize it in a way that you get insights as to what are the themes of my CSMs today.

And what we found is almost 87% of those interactions could be either fully or partially automatable. Now, this does not mean we're replacing our CSMs with automation. What this means is we are shifting our CSMs to more strategic engagements, and we're stopping those CSMs from doing that minutia work. So sending docs, sending recordings, sending license reporting to our customers that doesn't necessarily drive a better conversation.

So the way that we thought about all of these motions is a little bit different depending on the segment of customers. So at Okta, we have different tiers of how we serve our customers. We have basic, silver, what we call gold scale. It's a different operating model for gold.

Gold, and now we have this new segment called key accounts, meaning it is a very, very high touch model. We're actually adding CSMs to these key accounts so that these customers feel like they have that true white glove treatment. But the way we serve those customers with digital varies. We know that one size does not fit all across all these different segments.

So we do have different ways of serving them with onboarding and adoption programs through email and in-app guides. All customers get those. Some people call it a digital segment. They call it digital for all.

It's just different depending on the segment. We have what we call SBR and SIP. Those are Strategic Business Reviews and Security Identity Progression Plans. These are the core deliverables of our CSMs and our technical account managers.

And we automate those for our customer-facing teams so that they are the ones that are presenting it. But as we get further and further down market, we automate more and we provide more into self-service. We have a self-service success hub that helps our customers help themselves, no matter their segment. But it might be more personalized and guided and collaborative with their CSM as they go upmarket.

And now we look to add that agentic layer on top of it all. So what that looks like is we are providing that self-service experience through our success hub. We're giving them next best action recommendations. We're giving them automated maturity reporting, recommendations of what they need to do next, in an area that is more tactical in what they need to do but based on their business goals so that our CSMs can really focus on what matters.

So what does that mean? Is we need to augment that CSM layer with AI. So I'm going to introduce some things that we're thinking about and then I'm going to hand it over to Lawrence to walk through what exactly we've done. So this is our vision for how we think about this AI augmented CS motion.

And this is specific to how we think about it for our scale teams and our scale segments, meaning they have a pooled model of CSMs on the account. So today we have a digital outreach. What we're working towards is an agentic outreach, meaning the customer can get a welcome email that says, "Welcome to Okta. Here's what to expect.

Here are your tasks that you need to do. Is there anything else that would be helpful for you?" And then they should be able to respond to that and ask for their onboarding plan, ask for resources on how they can adopt their latest authentication methods. We should be able to give that guidance to them back and forth but also know when a customer... Can you all hear me?

I don't know for a second. We should also be able to know when that customer is at risk and still needs a human. There's still a time and a place for that human motion. So we have these risk signals to tell us, "Hey, our scale CSMs need to engage with this customer.

We need to help these customers based on a point in time." And then there's also the proactive motion. So not just customers that are at risk, but customers that we want to make sure that they are on the right path so that they can become advocates for us, they can grow with us, and we can ensure that we can work with our AE partners so that they can engage those customers for growth opportunities. So as this risk mitigation motion comes into play, this is not easy to do when you have overtaken those customers that we're serving in a segment. And we need to have that AI layer on top so that not only do we have these right point in time engagements, but when those CSMs engage, they are equipped to do it in the most effective manner.

So I'm going to hand it over to Lawrence who's going to walk through how we've done that. Thanks, Alana. So yeah, the first thing I would say is when you think about it from a human CSM perspective, one of the first things, and it's been said multiple times by other folks, but a lot of it is about change management, right? So one of the things that Okta has done really company-wide is kind of had all of these initiatives to make sure that there's AI fluency and AI comfort across all of the employees at Okta.

So mandatory AI learning days, right? AI hackathons, we just had one about two weeks ago. AI academies where we tested out doing cloud code and things like that. So I would say if your company isn't already doing that and you're a CS leader, that's something that you should think about doing for your own team.

But to go a little bit further in terms of what Alana mentioned, so she described this idea of, you know, I don't think it's AI or human CSM, or digital or human CSM. It's the case of how can you use all of those together at the right place at the right time. And so this is one way that we think about kind of the maturity of use cases for a CSM. And one of the things you'll notice is each of these stages, it's intentionally titled to almost be like a human title, right?

An assistant, an account analyst, a portfolio advisor. Because something else you've probably already heard is you can think about AI agents reporting to you. If you were designing the CS team of the future, you might, that team might have some human CSMs, but also some agents. And so if I step through those really quickly from left to right, you know, task assistant helping with simple productivity tasks, right?

Cleaning up your email. Hopefully everybody's already doing that. Otherwise, I think that's basic table stakes. The next one, account analyst, right?

Again, we're probably already hopefully doing this. At Okta, we certainly are helping get ready for a QBR, you know, looking at the data around an account and synthesizing that data for an account. So I call that account analyst. And then the next one really goes broader, which is this idea of looking not just at a single account, but across a portfolio.

So can you analyze hundreds or even thousands of accounts or activities? Alana already gave an example of looking at 50,000 timeline entries. And to me, that would be an example of analyzing a portfolio to identify risk, to cohort those customers, that kind of thing. But it's broader in scope than the things on the bottom left.

And then the last two, and, you know, I'm super excited by what we heard about this morning and kind of just in the industry. I think it's a really exciting time because you're starting to see agentic activities, right? Which I would say the last two columns, I would kind of put in that agentic. So instead of a productivity aid, building a gem where you have to go ask at things, can you actually hand off a process to an agent?

Or this morning in the keynote, and our boss, Ed Daly, talked about this idea of you even outsource a whole segment of maybe low dollar renewals to a renewal agent. And actually, you're not getting it to do work, but you're actually getting it to deliver better renewal outcomes, better retention. So that would be the idea of, you know, as you start to go up the chain. And one of the things we're starting to do now, I would say in Okta, RCS teams are pretty good at the things on the left.

And we're starting to experiment with, you know, staircase in the middle, and some of these Atlas renewal and adoption agents on the right. So I'll give you a couple examples. Portfolio Advisor, we actually have a play that we're running across digital and human CS that we call FastPass, right? Which is our phishing resistant authentication.

And this is something that we have thousands of customers we're engaging in around FastPass. And so as a CS leader, like I was very interested in, how effective are we being in those thousands of engagements? And in the old days, right, the way we would do this is we would look at reason codes. Everybody do your engagement, code the, you know, we have a set of dropdown reason codes.

But instead, what we were able to do, we have gone calls. We have almost all the emails with the customer. And we could actually run an AI sentiment analysis across, you know, almost 1,200 engagements, lots and lots of emails and calls. And it could actually tell us things like, what are the top three reasons that were successful or were unsuccessful?

Now that we have staircase, we can do this hopefully a lot more in a much more automated fashion. But this was something that, you know, we had to kind of stitch together with elbow grease and over weekends and evenings. And interestingly enough, what we found by doing this is we actually found that the top challenge to us getting more customers to use FastPass, it wasn't so much technical questions about how to implement FastPass. It was change management, right?

And so we actually took the results of that to our product teams, to the digital team, and we ended up creating assets, which we then give to our customers around, here's a sample template you can use if you're, say, an Okta admin. And you need to convince your users why we're doing this. We actually have, here's a sample rollout plan, right? We actually have a customized digital asset that lets each of our customers, even long tail customers, create their own personalized change management tool that they can then use across their company.

So I kind of like that story because I think it's a nice example of using AI analysis, not just on a single account, but across a fairly wide segment of the customer base or activities to produce some real outcomes. And these tools have significantly improved our traction with this FastPass play. The next example I'll give is, OK, so I also use this, remember I talked about reason codes, right? And the analogy I use here is when I go to the dentist, and my dentists always ask me, you know, they have their little checklist, has your health changed, any issues, blah, blah.

And then they ask me, how many times do you floss? And they write that down in their little clipboard, right? And I tell them every day, which isn't always true, but, you know, that's what I tell them. I think of that as that's kind of like the old way, the reason code.

The new way with AI, AI could just look at your sentiment, and regardless of the reason code, it actually gives us, in my view, better data than the manually entered reason codes that people put in, right? And so the dentist analogy would be, you don't have to ask me, you look in my mouth, you can tell by poking and prodding, you probably know how many, how often I floss. So whether I tell you I floss, you know, once a day or whatever, you know, the AI analysis kind of trumps the reason code, if that analogy makes sense. All right?

OK, so the next one I'll go through real quick is we actually had to do a bunch of CSM onboarding. And so one of the things that we worried about in that onboarding, obviously, you do all the training, and at the end of it, before your CSMs are ready to go to customers, is you have them do mock customer calls. And typically, we would do this with an existing manager, and we would set that up, and it takes time, right? You have to set up the call, you have to kind of scaffold it, and then you'd have to set aside time to do these calls.

And so what we did is we actually brought in an AI tutor that was interactive. We fed it all of our playbooks. It was not hard to set up. And then we actually had the CSMs conduct calls, and these weren't just one way.

These were interactive, right? So I had to do it. My whole team did it. I had to do it.

And, you know, a funny thing is that when you set these up, you can actually have a slider that says how agreeable or disagreeable the customer is. So we set it very disagreeable at first, and we were all a little bit horrified at how hard that personality was. But in reality, once my team got through this, and they were all fairly stressed out about it, but once my team got through this, they would joke the AI person was called Paulette. So they were all, "Oh, we don't want to have to face Paulette." But when they got onto real customer calls, they would say, "You know what?

Those real calls were actually easier than Paulette." And we would say, "Well, you know, you were stressed about meeting Paulette, but that was kind of the example." So that's, I think, another nice example. And then the last one I give before I turn it over to Melissa is, on more to the right, up and to the right, the autonomous processes. This is a little bit of what Ed Daly was talking about this morning in the keynote. But how do we actually hand off a whole process to an agent?

So in this case, we're looking at some of that scale CSM outreach. It's the part of the outreach that's not very productive, right? A lot of back and forth. Are you the right person?

Is this the right time to meet? Similarly, in our renewals motion, I think Ed mentioned this this morning, there's a lot of that back and forth. And so can you actually agentically have an agent, you know, handle that, to offload that, as Alana said, from the CSM so they can focus on higher value work? So that's something we're looking at.

We're now starting to work with Atlas to do this, both on the renewal side and the adoption side. And I think, you know, this has a lot of possibilities for letting CSMs do higher value work. So with that, I'll pass it over to Melissa. Awesome.

All right, partnership and collaboration. So raise your hands, how many, where's my admins at? Admins? All right, I got some cool gems in these next few slides for you guys.

So as it goes with AI, obviously, it's such the largest buzzword ever, but you can't use AI to fix everything. You really have to use it to fix a problem that you already had a pathway for it and you want to make that pathway better. You want to make that process better, more efficient. You can't just go, I have this huge renewal problem, a huge scale problem.

I'm just going to throw AI at it. If you don't actually already have a process that you want to use to optimize it, you're kind of just at a no state. Right. So everything like Lawrence was going over, it's we're figuring out these use cases, we're training stuff, we're figuring out how we want it to work so then we can work more efficiently and effectively.

And then also help support that kind of higher, like shorter ratio for those dedicated CSMs for our higher error customers. All right. And this did decide to just like randomly. I told you, you thought I was messing around with you.

Definitely a ghost. It's an AI clicker. Oh my goodness gracious. It is and it does not like me talking about AI like this.

All right. So here we go. So I want to show you this slide earlier and good gracious. Y'all my thumb was right here.

I don't know what happened. All right. So you've seen this slide earlier. The reason I have re-put it up here is because I want you guys to see all the little things at the top where I've got digital success, covered success, success, CS operations.

And I did put a thing called TDI. At Ofta, TDI is our business systems team and data team. I was like really long so I just shortened it to our abbreviation, which is TDI. They are integral.

All of these partners are integral to make things successful and make things work. So what I wanted to show you here is talking through Alana's pyramid and showing you where we all work together, where we have to pull in ops with digital, where we have to pull in CS with digital and ops, where we need to pull in the data team and the business systems team, because collaboration is how we actually are successful not working in silos. So ultimately, one customer experience with a unified goal, right? That's how having the right cross-functional partners, having that shared vision of success and being able to make that really go together.

So obviously we want increased NRR, increased DRR. How are we going to get there? Through a combination of digital success, cross-functional partners like CSOps and with our CSMs and Tams. All right, so this is kind of some cool stuff.

Case study. We call it the HandRaiser program. Maybe a branding change shortly. But basically what was our problem statement?

What was happening was we had these great digital success programs and then we had Lawrence's scale team. And what was happening is we would send an email. They wouldn't take the action. We wanted them to take so then we'd provide a live human to go in and check in.

But those were cold calls, right? They never had a CSM. So they're like, the scale CSMs are like searching through activities and trying to see who the right contact is. And Lord knows if that person even wanted to be approached.

We didn't know why they didn't make the motion. Was it because they didn't open the email? Did we have the wrong contact? So we knew the process of digital first and then human, right?

But we were like still stumbling upon the best efficiency for that. So what we did is we decided to figure out a bridge and a solution for that. So the really cool part is we leveraged the digital success team. We leveraged those emails.

And instead of just saying, okay, after so many days, 30 days, 90 days or whatever of non adoption, give it to Lawrence's team, we actually built in a link, like a little button. It's like, hey, do you want help? You want someone to reach out to you and talk to you about this? My gosh, looking back, how do we not think of this sooner?

So if somebody clicks on it, not only do you have an engaged person that actually wants to be reached out to you, you actually have the right contact. And they're not a cold contact. They're saying, yes, please, I would like some help. So when Lawrence's team gets this, they're like, sweet, I know Bob wants contact.

I know Bob is going to actually answer my call or my email. He's actually going to want to meet with me because he said, hey, actually, I would like help with this. So it really reduced that cold call piece, probably the stress of his CLCSM is doing all these cold calls to people. They don't know if it's even the right contact.

And it increased actual open rates and stuff for the digital success program because people were actually opening him and we were actually then engaging. And so it helped with all of that. So I kind of go back to that whole collaboration and partnership thing. So I'm sure you guys have heard of RAPIDS.

So this was kind of the RAPID on that. Like the digital success actually recommended, hey, we think there is a better way to do this. And so we had the whole agree, the perform, the input and the decide. Now, for smaller things, we're not going to do a RAPID for everything, right?

But honestly, big initiatives, you want people bought in, especially if you need data from a data team. You want them bought into the story, to the use case, to the KPIs and the success of it. So by doing this and saying, hey, you're so important, you are on our RAPID, it gives them accountability and ownership and they feel valued and then they can actually go in here and they're actually going to prioritize their work with you. So ultimately, super successful, granted, it's about a month old.

So we're still dealing with first results, but we're pretty excited about it because this really opens up a whole new world for us. How many other things can we do? And the whole, telling you this, a spicy AI ghost. So the great thing is we're actually learning all this stuff.

So as we start doing the adoption agent for Atlas, we're going to glean all these learnings from this kind of stuff first. So then we kind of know the gotchas. We know some stuff that we can already apply once we get all that rolled out. And for my admins, some nitty gritty details is we actually use the email raw events log and that's how we could tell who was clicking and what link they were clicking and all of that.

Because I'm sure you know in the normal log files, you can just tell that they clicked. Well, what did they click, right? But if they actually clicked the link that said, yes, please, I do want to help, we were able to pull that contact and the fact that they clicked that, and that's what we were able to do to create CTAs. Which leads me to my other cool admin spotlight.

Okay, so Lawrence's team is rather large and they're all very unique with regional languages or specializations. So how are we going to get these CTAs? We didn't just want to round-robin to everybody, right? So we actually create a very robust system for this.

We have 51 user groups, guys, 51. And what we've done is we've said, if you speak German, if you are an expert on this particular product line, you're in a set user group, okay? And inside that user group, we actually also cap you at 40, at this point, 40 CTAs. Because we also don't want to overwhelm you.

You're a scale CSM, you're having a cold reach out on some of this stuff, some of the others you're not. But we don't want to overwhelm you with 100 CTAs if people are just cranking these out, right? What we do is in these user groups, you have a cap of 40. Once you hit 40 CTAs, you get popped out of the user group.

So you cannot be assigned a CTA until you are less than 40. And we also do this to where we have user groups for, like I said, just specific regions. That way, the German folks only get the German-speaking CSMs. And then they're capped at their limit as well.

So this actually really creates a much better experience for the customer because things are tailored towards exactly what their needs are and their language and their products. And it's also a much better experience for the scale CSMs. They're not overloaded, they're not having to do something that they're not really an expert on or that they feel is comfortable on. So by building this out, it really did a lot of cool stuff.

It saved efficiencies and, honestly, kind of a higher satisfaction score for the customer and the CSM. And the really neat part is that screenshot on the right, what we did is after we built everything, we're like, as an admin, you don't want to have to maintain, "Hey, Lawrence is going to email me every single time, "Hey, I need this person added, I have a new hire. Hey, I want this region changed. Hey, I want this cap changed." So we built an inline dashboard so whenever one of his managers or him want to go in and actually change and add, he can just do it in the dashboards and then everything syncs automatically.

So he does not have to reach out to us. It's immediate to whatever his changes are necessary that he's wanting to make. And we have it all available on the Gainsite inline dashboard. So kind of a cool spotlight for my admins, right?

All right. So what is our ultimate impact? Honestly, it's an impact of better together with intelligence at scale. Now, a lot of this stuff is we're getting more used into AI.

We have all these great AI visions, but until we can completely get all the way over to all the agentic wonderfulness, this is going to help us train and learn and get a good vision for the gotchas and what we want to do in those great use cases. So automation is a foundation. AI augments. I don't know if you guys can hear me as I'm telling you it muted me because I'm trying to tell you that AI augments, not replaces, and then a unified customer experience.

So that leads to success, good partnerships and good satisfaction for our customers and for our employees. So all right. That is it. Thanks, guys.

[ Applause ] Thank you so much. That was amazing. And this is actually real world talk. So it's not just, yeah, I can do this and I can do that, but these people have actually achieved it.

So another huge round of applause for our speakers. Thank you so much. [ Applause ] I'm not sure if you are already familiar with this. If you attended any sessions before, you probably know this.

But there is a slide that you can put in your questions. How do you find that? You go to the Pulse app and right on the main screen, go to room 211, and that's this is 211, and that's where you will find it. So with that said, I'm going to start reading off some of these questions.

The first question is for the AI use cases, what's the tech stack? Buying versus building in what use cases? Anybody want to take this question? Yeah, I can take it.

So, yeah, buy versus build is very huge right now for us. But we also know we can't build everything, and also we shouldn't build everything, right? So what we're doing is we're taking tools that we already have, like Matic and GainSight and EverAfter, and we're utilizing the AI that we can from them while also supplying our TDI team, our data team, with different things that we eventually want to build or want to do. Now, that said, if there are new things like Atlas agents, they are definitely promoting, hey, definitely do pilots.

Definitely do stuff for like a year. Let us know if we can't build anything that's going to even compare to that. We want you to keep moving forward. We don't want to hinder that.

But they also do, I think, want the chance to see if it matters for all of Okta or if it's just our use case, and that kind of changes it, too. If it's for all of Okta, they're most likely going to want to pair, a build, and buy. But for our specific use cases, they are kind of letting us, after much security review, trust me, they are allowing us to buy on a certain front. Great answer.

Thank you so much. The next question we have here is, how are you measuring the impact of your digital success motion? Really important. I can take that one.

So we try to align our metrics as much as possible with our actual CSMs. So North Star Metrics, NRRR, GRRR, just like our CSMs are. But then we have leading indicators and lagging indicators, depending on what we're trying to achieve. So for my team specifically, we have three different motions across the team.

Each are measured differently. We have our digital journeys motion. So think, how do I reach out to customers at scale through email, in app guides? And we might look at, for example, are we having high value CTAs for Lawrence's team?

Because if we do our job well, then CTAs that Lawrence's team has to use to reach out to customers are going to be much more effective and things that actually require a human. We also have a self-service experience aspect, where we look at how many customers are coming to our success hub and what percentage of our customers have adoption breadth, meaning they've adopted core features across what we offer. And then we also have an automated field insights motion, which is automating all the insights that our field presents to our customers. And that's somewhat of an efficiency play, meaning we're saving hours for our CSMs.

But we also want to make sure that the data that we're using is more effective with our customers, like benchmarking and automated maturity scores. And with that, we want to make sure that our customers have stronger renewal rates, stronger growth rates, just to name a few of our metrics. Thank you. That was awesome as well.

So this is a question after my heart. So it goes, as you shift to automated self-service actions, how do you define strategic engagements for your CSM? So it's basically, you're taking the CSMs from your, for the scaled customers. And so what are you doing for them?

Like what are the high value activities that they would be? Yeah, I can take a first crack at that. It kind of goes back to that, the pyramid that we used a couple times. So at the upper end of the pyramid, the higher ARR value customers, that's when it's much more of a human centric, you know, white glove concierge approach, right?

The digital tools are still useful there. A, they can be a productivity aid for the CSMs. For example, QBR prep. Our CSMs can kind of push a button and they get a QBR prep deck, you know, prepared for them that they can then customize.

Second thing is there are still things at that high end where customers can self-service. So certain things like, I think Alana mentioned this when she talked about the analysis of activity. Our CSMs were doing a lot of repetitive work just to show usage reports, right? So a big customer, they have a lot of times, even though they're in that concierge model, they don't mind and probably prefer can they self-service and get their own usage reports.

And then our CSMs there, they still do the big, you know, the high value QBR and value piece of it. And then I think lower down the pyramid, so kind of once you get into mid and long tail, that's where we always try to lead with digital, right? And ideally in an ideal world, we know this would never happen, but in an ideal world, say a low adopted customer, we have digital programs for them and that customer would get the digital materials and then they would deploy. And you might never need a CSM.

So the more we do of that, the better. And then we still have CSMs go in at the tail end, right? If that, if a set of customers has gotten those digital outreaches and is still in a risky state, i.e. well under deployed, then we will assign a CSM to them.

So that's kind of our focus at the mid and long tail, more of a risk-based focus, but led with digital. That's awesome. Thank you. This is a loaded question and a two-part question for Okta, per se.

It goes, as a security-focused company, how does Okta set internal policies around the use of AI with customer data? Big question everybody has on their mind. And then since many customers may be hesitant about AI adoption, what strategies are you using to address those concerns? You want to take that one?

We'll break it up into two parts. I'll take that one. So honestly, I feel like we've been, we have a really good system in place. We have a VP of AI strategy and underneath that, that is an Okta strategy-wide.

And we have a governance committee. So whenever we want to do a new AI tool or we think there's a different use case that we haven't used before, you actually have to submit it to this governance committee. And they, just a minute, I think I went a little bit back there. And we also have a Slack channel, but long story short is we actually do have governance.

We want to make sure that the use case is strong. And if we say I'm sorry, like honestly, our MCP, we've been really wanting to do that for Gainsight. But Okta has kind of a plan to do this kind of Okta 1 agent that they want to pipe all of it in. So they're like, it's delayed for now.

So we had to submit it. They monitored it. They said it's delayed for now until we can get this other tool ready to ingest. And then they're going to release it.

So it's pretty governed. The other thing is we actually have a ton of training. We have AI training that we are required to take that says what we can and cannot put into, say, Gemini or Claude or chat GPT. And they're obviously able to see everything we do put in there.

So definitely don't want to get into trouble on that. So I think there's there's policies, there's governance, and then there's an overall strategy that is all outlined. And Okta is very AI first. Like they want us to be able to use these tools to make things more efficient.

But also Okta is building to secure agents. So they also want to make sure that whatever we're using, we're also doing it in a secure fashion. What a great answer. All right.

Probably the last question of the session. And this one is what platform do you use for Paulette and mock meeting training? Yeah. So something we didn't mention in our presentation, but we're also Okta is also a big user of Skilljar.

Right. We were using Skilljar for our education. This is more like customer facing education before again site acquired Skilljar. So excited about kind of the integrations to come and the synergy between Skilljar.

Now that's part of Gainsight because I do think from a digital perspective, education community is all an important facet of that as well. For the letter AI to answer or for the Paulette piece, we actually used a tool called Letter AI, just because it was a tool that we already had that had this capability. And we use that strictly for internal enablement. And I'll go ahead and answer the user group thing just because it's kind of an ops quick question.

So how we cap the user groups is there's actually we created a field on the user itself. And we put like the number 40 in there. So we have rules that can once we hit that 40 count that we can pull it out of that user group. And also the really great thing is one of the questions was like, well, what if you, how do you know 40 is too many or 40 is not enough.

So after a while, we basically realized that 40 was kind of like that sweet spot. However, Lawrence has a lot of early career and new hires and interns. So they're not going to obviously be able to take 40. So the beautiful thing about this being a field on the users, we can say, okay, these new folks are only going to get 10 or maybe even five.

And then as they progress, then we're able to change it through that dashboard and he can increase that to that level. So that's kind of how it's actually quite easy. Once the rules were built to do this, this whole workflow, it was easy to go in and out and really kind of admin it and customize it as he sees fits. So thank you.

Thank you, everybody. And huge round of applause, please.