CELab - Ep 180 - Lila Krutel - How MCP + AI Are Shifting Power Back to Customer Education Leaders

Speaker A: Hello and welcome to C Lab, the customer education laboratory, where we explore how to build customer education programs, experiment with new approaches, and exterminate, as always, the myths and the bad advice that stops growth of our companies. Dead in his tracks. Um, and today I'm excited to have Lila Kurtel on the show again. Lila, say hi to everybody and hey, everybody. How are things going? I'm glad to have you on the show. It's been a while.

Speaker B: Things are good. I'm excited to talk with you, Dave, and hopefully, you know, share some of my learnings out loud with the ce, uh, audience.

Speaker A: I, uh, like how you said learnings out loud because that's our style. I did a show for that for a while, and now it's all C Lab. But, um, to warm up, what do we always do? We do is you always do the International Day of today. And again, we're not going to show it on the screen, but we're going to make our picks. Um, and you can always go to the link daysoftheyear.com where I pick these from. So I'll ask you, what's your favorite day of or today?

Speaker B: Yeah, um, there were quite a few odd or interesting options. Um, maybe the strangest one was shrimp scampi day. Um, but I do love shrimp scampi, and it's one of the dishes that I learned how to make, like, as a teenager at home. So, uh, there you go.

Speaker A: Yeah. Now you're in, uh, the Portland area in Oregon. You get a lot of fresh seafood like we do in Seattle area. I don't know so much about shrimp, but, um, yeah, that's cool. Okay, here's mine. Uh, I, I decided I like this one. It's World Wish Day. Never heard of that one. World Wish Day. So, I mean, if I could have a wish for the world, it would definitely be as mundane as. Can we just have some sanity for a change? Um, but I'm going to turn this to. Our discussion for today, of course, is I wish I understood AI better. And that's why, uh, this is a fun side of playing this. Uh, that's why you're on this show today. We're going to be talking about, um, some other aspects of things, and we were pregaming this discussion. So let me just get into it. Uh, you know, what's happening to us as education practitioners and leaders working in the B2B SaaS space is that I think. I know we're pressured to move faster than so many other teams. Like, if you think about time, I've Spent in like an L and D capacity or whatever. Things are moving slow. Or academics. They do teach a course on physics. There's any number of books I could pick off the shelf and say I'm going to teach to that. Um, but now, uh, I'll start with this story. A few years back, throughout my career, I was on a customer education team that was responsible for tens of millions of users. Okay, let that hit and sink in for a minute. Usually, Lyle, you know, I've been 10,000 or something like that, maybe 50. But we are seeing such. We're working in fields where there's such a tapestry of kinds of people with different things, and we have to deliver content at a velocity that we've never done before. So all of this really, that, that experience for me, scared me too. I mean, we had a lot of stuff to offer, but now it's about how do we go faster, how do we grapple with what's coming in, how do we use, you know, what's in our lms and how can I query that better? And it might be even as simple as I need to run a report to find out how many people have taken this course for this client. And today with some of the things that I know that you want to talk about, that is getting easier to do without having to have an OPS person. We have Claude, we have Gemini, we have ChatGPT, we have other God. Like, let's talk about the gainsight properties base and all the AI that you guys have got. Right? But the thing that scares me is that there's often zero visibility. And I. And again, I'm. I'm being vulnerable to the audience here is I'm learning every day. We have to share notes, we have to trade notes, and we need to be able to take all the stuff that we've got, use AI in a, in a. In a very functional way to get us to enable our teams, to enable our customers, to enable our partners. And we have to do it all at once on, um, a velocity I've never seen. So I want to flip the script a little bit and say, all right, we're going to ask how do ACE leaders grapple with and use AI today? And this is the time recording April 2026. So if you're a time traveler, you're probably going to laugh, I hope, or maybe cry. There's, you know, memories, um, but insights, impact data, the seat of the table that we want. And we've talked about this for forever, Lila, that as educators, are we pushed aside or be Part of the strategic response in the proactive nature of a company that turns everything around and says, hey, no, we're, we're working with customer every day. We're teaching, we're helping, and they're learning, they're in community, they're helping teach us things. And there's all this stuff together. So we're going to angle on, uh, one big thing. We might talk about other stuff, but let's bring the term up. And, and I'm going to, uh, uh, I'm going to preface this with your background a little bit. We're going to talk about MCP Model Context protocol. What is that? Well, Lila, I don't put words in your mouth. You just said this. You're a gainsight. You work with skilljar Brand. You and I go back at least a decade. I really appreciate your perspective. But what I want you to do is channel this discussion you and I have been having, and we've all been having. You and I are not engineers. We're teachers, we're practitioners, we're experimenting. There's new stuff coming out. And that's where I'll say to you, maybe we start with what we just talked about is, uh, when did we hear that term mcp?

Speaker B: Sure. For me, I think it's only been about two months, um, since I was introduced to this lovely new acronym mcp. And uh, even knowing what it stands for, Model Context Protocol, doesn't really help you understand what it is. Um, we wanted to start there. And you know, still a fairly technical definition is that it's an open source standard for connecting AI applications to external systems. But what does that actually mean? Um, and so, um, a little bit more in my own words, what I would say is that an MCP server gives your AI assistant, which might be Claude or ChatGPT or Gemini or whatever, um, but it gives your AI assistant arms to kind of reach out and connect to systems like your LMS or like your customer success platform, just as examples. And at this point, most MCPs are primarily about reading and analyzing data in those other tools. Um, some of them, like the Gainsight CS MCP that launched in the last month, uh, allow limited writing or doing in the system as well. But mostly at this point, they're about being able to read data, um, and analyze it. And through your AI assistant, you can ask questions conversationally to better understand your data or find information and answers about your customer data.

Speaker A: Let me ask you a question and make sure I'm connecting in same wavelength. So, uh, the platform that I've I was, I've been using, um, has some of this capability. So the other day I get a call from one of, well, I call it slack from somebody who goes, oh my God, oh my God, I need to know how many people from this customer have consume the training and blah, blah, blah, blah. And then immediately had that hard swallow because our LMS administrator was not there.

Speaker B: Sure.

Speaker A: Oh, I don't know. I didn't know this platform very well. Um, so I, I like, oh, what's this? And I, I did what I would do. I'd queried, you know, with. I, uh, prompted build a prompt, got it back in seconds, exactly what I needed, sent to the guy. Oh my God. I was, I expected a couple days. What the heck. And I go, I did too. I had no idea. So exciting to me in that moment because, oh my God, lila, you remember 10 years ago, we were working together and we were trying to generate some reports and stuff for what we did and.

Speaker B: Right.

Speaker A: Pain. The time I. We had to go to like a, uh, you know, solutions architect or something and go, can you slow down and explain this to me in layman's term? I go over to like Chris Anderson or, or Edgar Ramirez or some of, you know, viltas, and they'd be like, oh, it's so cool, but it took so much. Yeah, it's just crazy.

Speaker B: Yeah. Um, I think we're, you know, getting into kind of the, um, how this empowers folks in CE potentially in a new way. Right. It democratizes access to data and analysis so you're not reliant on your CSOPs, um, or product ops, you know, team or business analysts to do this analysis for you necessarily. Um, it becomes much more accessible, much more quickly. And um, because it's more accessible, you, there's, I think many more sort of examples or use cases or, you know, um, practical applications that become possible that previously you would have, you know, you wouldn't even have pursued because it would have taken too much effort to get to that information or that data.

Speaker A: Okay, let's. Let's return to you and then come back to this because you're. This is exciting to me because you said something in there which led me to come to the term the art of the possible. And when you're talking about, you know, I would have just like me re. Paraphrasing what you said. Things that I've seen before in my role. I would just stop cold like, okay, I'm gonna have a fight with this person. I'm gonna have to go that it's gonna take a week. Okay. Eff it. I'm not gonna be able to get this because it's too hard now.

Speaker B: Right?

Speaker A: You're. You're bringing the energy and excitement in because a light went on. You got a new tool. It's approachable, it's human, even though it's AI. Tell me more about. Okay, where are you in your role today? And then from that, what are the kind of things you've been grappling with? And now how are you? How did all this stuff come to you? What are the things that you're kind of seeing that help you from an AI perspective in the role that you have? That's a lot on the role first, I think.

Speaker B: Yeah. So, um, I'm continuing to lead education services at Gainsight, and I also oversee our technical communications team as well. So doing more work in the last six, eight months or so, um, to improve and increase kind of the collaboration between our documentation and education. But also, uh, the organization is invested further in digital customer success. So we have a new leader, um, that's overseeing all of our digital programs and that's really helping us get, um, better aligned and work on the common customer journey and deciding what are the touch points and the milestones and resources that we want to or need to provide users with, um, at, uh, each kind of step along their journey and what's the right tool or delivery method, um, you know, for serving that content, how do we distinguish, you know, when to use the different channels and so on. Um, so that's like a little bit about where we are in maybe kind of my role or. And team. But also I think it makes sense to mention that I feel pretty fortunate as well to be at a company that, um, where the leadership are like, very, um, invested in helping us all become more skillful and familiar with AI tools. And so when you asked, like, how long has it been since I had even heard of an mcp? And I said it was really just two months ago and the person who happened to be explaining it to us was our CEO at our, um, revenue kickoff. So, you know, uh, like we are hearing about and being supported and encouraged and actually it was right there that they decided, let's get everybody an enterprise Claude license. Oh, wow.

Speaker A: That was a gift.

Speaker B: Yes. Yes.

Speaker A: Um, are you a fan now of Claude as well? Like, I.

Speaker B: Sure.

Speaker A: Oh my God, I'm getting so much done.

Speaker B: Yeah. Um, I do, I do like Claude. Um, and, um, most, I think teams in our organization anyway, have transitioned kind of away from chat GPT and more toward, uh, leveraging Claude. Um, and so that's how, like, I, I, for example, am accessing our CS MCP and, you know, experimenting with other MCPs as well.

Speaker A: I am really digging this because now you're relating at a level that I think everybody in our audience can resonates with. The last podcast I just did was with Jess Katz, who's at Forder, who laid out her, uh, you know, swarm of AI agents that are doing some amazing. It literally blew my mind. And I wrote and I wrote and I wrote and I listened and read stuff. Can you, can you tell me a little bit more now? So here you are at Gainsight. Chuck. Chuck, right. Um, yeah. CEO said, okay, you all get an enterprise, ah, cloud license. Oh my gosh. And then some people are probably like, what is that all about? But you're probably like, okay, now I can actually not have those quotas or the admins are going to have to like, I over. I already ran out of my. I had to upgrade my license and I have a personal account and. But I'm, um, like, I get so much out of it. I'm. Here's my money. Take my money. Right. I, um, don't. No questions. And now I'm exploring. Well, take me. Can you take me a little bit through before we get the next topic? Like, your journey is.

Speaker B: Yeah.

Speaker A: How have you, as a customer education leader and practitioner.

Speaker B: Yeah.

Speaker A: Grappling with that experience of using this in your day to day. Sure.

Speaker B: So I think this is also a good time to say, like, if folks who are listening, if you haven't heard of or you haven't used an MCP yet, um, you're not behind, like, it is. Okay. Uh, to be more in the learning and discovery phase. Um, I mean, we all are somewhere on that journey still. And, and, um, I said this to Dave before we got going. Um, and then I totally lost my train of thought. Um, about

Speaker A: all good. My train derails once a day at least.

Speaker B: Um, I totally lost what I wanted to say there. Um, but I guess what I wanted to share is that I had some skepticism on that day at the RKO where Chuck was sharing, um, about what MCPs are and what we're building for our own products and customers to leverage. Um, but I think what's really valuable is if you can, within your own organization, cultivate a culture of sharing and, um, you know, learning together, that that is going to be really valuable. Um, we, you know, very actively proactively encourage our teammates, um, to, you know, again, sort of like learn out loud or you know, show what your experience experimenting or share what successes you've had. Um, you know, tips and tricks for working with AI. Um, and it's those, you know, those sort of show and tell moments that we're doing in standups or team meetings or you know, sort of all hands or what have you and on Slack and things like that that um, are what get us all thinking about. Oh. Or you know, recognizing like this is how this could be useful for me. I see how you did something here that is actually really similar to something I need to solve or that we're doing manually right now and it's very tedious and you know, and so on. Um, so if there's you know, a tip there, it's about like encouraging that kind of knowledge sharing and experimentation and it being safe to make mistakes and try things out and then try something else and you know, and all of that. Um, and so I come, you know, to talk with you today, Dave, and the audience really as a fellow like learner on this journey, not an expert by any means. Um, and uh, so, you know, day to day I'm trying things out with Claude and in connecting to our MCP and asking questions, um, to see like can it help me with this analysis or with this problem. Um, and sometimes I get stuck and sometimes I get frustrated or you know, sometimes it doesn't tell me the perfect clean, uh, word that I would like it to. Right. Um, but that's all part of the process, I think.

Speaker A: Yeah, I really like how you're. I like how you responded to that because that heartfelt, uh, look, I'm a fellow learner practitioner. At the same time something happened to us recently and again I talk a lot out loud about my own personal journey. Why am not, I'm not all about going all the way up into a leadership path because I'm at where the learning is happening and the change is being made. And this is why I've gone from company to company to kind of like cross pollinate and understand all of the problem sets. Now with the uh, entrance of AI and things like mtp, which now let's start really breaking into that and talk about how this is going to shift the power a little bit. I love the fact that my enablement team, my manager is Kristen Thompson at Vast Data Global. She's the head of a global, um, education enablement and probably metal or title, but she inspired a discussion with all of our team last week when we were at offsite and we did a show and tell and oh my God, that was so fun. And mine took a little bit different twist on some of the others because what I was doing is showing some of the tools that are both. Oh, here's what I did with Claude for you to do this. But there's platforms that I know that, you know, skilljar Insight as a brand interact with things like Pluso or um, Leah by, you know, Learn experts and Sarah Sedgman. And there's all these I call complementary type AI AI products. And in addition to the thing. Look, what I love about MTP is that for me it's kind of like an API but on steroids. And it's, it's, it's. Now we're getting to the point where I can ask the Oracle questions. Oh dear Oracle, blah blah, like, tell me this tech stuff so I don't have to bust my brain talking to a million people. And now we're getting the kind of stuff out that we've always wanted and we've been, and it's just been so hard to do. Yeah, let's, let's get into, into some of this more because I really liked how you brought the emotion into it. The, uh, there is this fear element and you know, fud of what they call it fear, uncertainty and doubt. And it's good to be a skeptic. But I'm going to say this, and this is part of a book I'm trying to write is the ship has sailed. If you're against AI, be skeptical all you want. That's a good scientific point of view. Experiment, try things. We're not going back. There's no way. So if the point I want to make to our audience here is that the differentiator now is for all of us to group together, work out loud and learn from each other and grow, solutions are going to come. Now because of the art of the possible, you and I can go off and make some really cool query. I don't know about you, but I went through the Google Prompting Essentials course on Coursera. I learned so much about that. I put my ego to the side saying, I know all this crap. I'm a tech nerd. No, I'm not. I learned so much. And now what I do is archive all of my prompts, not just in Claude or OpenAI or whatever, but I also have a running log about what it is, what the outputs. And I spend a lot more time crafting really good prompts using the TCI framework, which is, you know, your task, the context references, and then you, you know, experiment and iterate.

Speaker B: Right.

Speaker A: I'm a scientist so I have to do that. So okay, now let's get into it. Um, so before M mtps, before we started this, let's go back to well, how did we try to get, and this is a great question that for the gainside ecosystem is we're all involved in trying to show impact. You tell me a little bit more about that. Like, okay, let's go back and again, square up that feeling and the experience we had and then let's move on from there.

Speaker B: Sure, sure. Um, I think this will sound familiar to the audience that in the past um, getting your training impact analysis, um, was definitely about begging, borrowing resources from CSOPs or product ops, trying to get into their queue, uh, often being told actually we need to prioritize something else. Why don't you check back with us in a month. Um, so getting deprioritized a pretty frustrating but also one off analysis. So uh, the data would age and become stale. Right. Um, and what are you gonna do? You gotta go back and start that whole process over again and in the meantime it's not the same people anymore and they don't have the context and very much like you're back at uh, square one.

Speaker A: That the feeling, the felt experience of that Lila was. I think it was kind of damaging to me in a way if I were to hold the space and carry it forward. I moved on from different, between different companies because I felt like I knew what I needed to come. I knew what I needed to be able to show leadership, that our uh, programs work with evidence. Right. And I like what you said before with this kind of technology we're having. So we go back to what uh, MCP is about. It's more like that really intelligent AI assisted API that now I can ask in plain language. Hey, and this is what I do. I go up to, let's say Jane is my ops person. Jane, I really need to get, I need to get this information. I know the data is here and then there's some in Salesforce and some other stuff here. Then I've got this other whatever data lake or lake house or whatever. No, I don't want to get into data crap anymore. It's. I. Now you said democratized and that is well, crap. I can, I mean I can query Claude and I have authority and it's, you know, it's, it is in on it, everybody's cool, there's no security flaws. But I can answer that, ask that question in plain language and Log and come back and say, oh yeah, Dave, you know, the so and so customer did this kind of training. Anybody could do this. Then I could make a little agent or something like that. So. But I think what you're tapping up against too is that this political, there's a political nature to the job we do. Can you tell me more about, like, is this shifting that power dynamic that we have as education practitioners in high school, high velocity companies?

Speaker B: Yeah, absolutely. Um, you know, just having access to the data and analysis. Um, you know, it's, you are, you're now more empowered, um, you know, to do some of that work yourself or to explore hypotheses and hunches, you know, quickly. Um, fail fast, I guess. Yeah, big time popular expression, right? And especially in software. Um, and, um, like the time element, you know, not waiting around for other people to tell you about your data or, you know, come back with a set of conclusions, um, or findings and then so often being like, well, but you didn't factor this in or what about this detail? Because they don't know the data and your context as intimately as you do. So, um, yeah, I think this is. MCPs are, you know, are really good at like giving that control back to the CE leaders and practitioners. Um, and so that's, that's part of what makes this very exciting.

Speaker A: That is really exciting. And I mean, I've been part of the journey that you've been on at Gainsight. You've had the good fortune to be there throughout, you know, your tenure started before mine and you're still there. So you've seen all the changes and I'm seeing the shift even externally now. Like when I'm within a company, I have these kind of tool sets in my disposal. Now. I am an ops person, or at least in a limited capacity. I could go, right? Oh yeah, Like, I loved how you, you articulated that. You know, like, hey, I got, I'm just, I'm gonna fail fast, but I'm gonna. I got a hunch.

Speaker B: Yeah.

Speaker A: 100. Something's happening here. Hey, Claude, can you go off and give me like, um, the last three months of data? I want to look at this, this and this. If you have it, then if that's set up properly and actually go to Salesforce or some other thing, or your usage data, you can knit together a report and it's gone up and figured it out. You leave that to the engineers, right? But now you come back and have the prescient data, you have the current data and you could actually package that up and have that run every week or two weeks. Now, you're not, you're not the middle person. Where I've always felt, I think this, again, we're talking a lot about emotions here. I have always felt as a education practitioner and leader, uh, at different points in my journey, that I've been hamstrung by this. This has opened a door, that and this being AI. But the prospect of using MCP now cracks it further.

Speaker B: Right.

Speaker A: Oh, all the stuff I've been trying, I've been on the circuit. When you went to Azukwa after I was at Gainsight and I was talking about all the, the interstitial work that I had to do to connect this to this, to this, to this. And that was months of work. But when I got it, it was automated. Now you don't have to do that stuff. And now we can make those. We could look at somebody could come up with a really cool idea and deploy that and we can all share that and benefit from it. And now it's internal in our own secured shell around Claude. So that stuff's not going to escape out. And then we've got. Yeah, okay. All right, wait. I'm getting excited here, but let's, let's make this practical.

Speaker B: Sure.

Speaker A: What I'd like you to get you talking about now is the reality. The use cases meet where, um. And I'm not going to ask you to share your screen now, but if you do have anything stage up, that'd be cool. But even later, screenshots or whatever would be cool. Let's talk about, um. You had a win recently.

Speaker B: Yeah, yeah, I'd like you to talk

Speaker A: about that because I think that was such a cool thing.

Speaker B: Sure, sure. Um, so this was maybe just a month ago, you know, we were, um, promoting some new instructor LED courses. And they're, they're paid offerings.

Speaker A: Um, no, I'm just curious because I'll go back. I take the training too.

Speaker B: Um, uh, sorry, Dave, I didn't catch. Were you asking a question?

Speaker A: Yeah. What are the new classes?

Speaker B: Oh, sorry. So we were launching, uh, some new gainsight administrator, what we're calling like experiential, um, courses. So this is kind of going beyond like the basics or introductory admin training and, um, getting a little bit more advanced or deeper into like working with your data or, um, uh, setting up a risk and churn process which would touch multiple parts of the, uh, CS application. Right. Not just Rules Engine or just Journey Orchestrator or something like that.

Speaker A: So you're deeper and unrealized. Value of the platform. And now you're getting the administrators to really understand. Now you can do this and this and this. And here's how it works out in practicality.

Speaker B: Right? Right. So, you know, we were talking about it or sharing about it like in the usual places where we, you know, admins hang out in a slack community or you know, in newsletters that go to admins and things like that. But we still weren't really seeing the registration numbers that we needed to actually host and offer the courses. Um, and we knew that there were lots of existing customers with unused training passes. Um, so, you know, for anyone who might not be familiar, a, uh, training pass is basically like a prepaid voucher. Um, customers bought a number of, um, vouchers as part of their initial contract that entitles them to send a user to, let's say, 10, 10 courses over the next 12 months. Um, and what often happens, of course, is that customers buy them or they're included in the contract, they use some of them, but then they kind of forget about them and they have unused passes and as you said, like unrealized value kind of sitting on the shelf.

Speaker A: Right.

Speaker B: Um, so we wanted to figure out like, who, who are those customers with those unused passes because they are, uh, the perfect target audience potentially for these new courses and they're not even gonna have to spend anything. Like, they already have these entitlements. It's basically free. Um, and you know, the benefit for us, of course, is that um, they like, the more they know and the more they can do with Gainsight, the more workflow, workflows and processes that they build out. And Gainsight, obviously the stickier the product is, the more embedded we are in that, um, company. So this seems kind of silly, but before the MCP was available, I had no way to directly go to my LMS and query for which customers have unused passes. Um, which, yeah, again, like, seems kind of, you know, like an obvious ah, gap or what have you. But, um, what we did, uh, or what I did is I went to Claude and I asked it, you know, which of my customers still have unused training passes. Um, and I also looked at some, had it look at some other things because you can connect Claude, for example, to your Gmail in addition to other tools. And we have um, like emails that go out to every new customer that share their codes with them, uh, for the training passes. So there was sort of a combined thing ultimately here where I was like, tell me the customers that have unused passes. Okay, great. Now I know the 43 customers. Um, but I need to know who to email at those customers to share this, you know, new um, course announcement and information about their unused passes with. So I was also able to find out like, well, who did we email originally to share the training, um, pass and the sort of the promo code with and like, you know, put that data together effectively to send an email campaign to that, you know, subset of customers.

Speaker A: Wow, okay. And this is, this is actionable. Like anybody can do this if, you know, particularly when they have some kind of a setup like this because, uh, you've done what would have been before very hard.

Speaker B: And I just, yeah, very tedious, like going in and you know, looking at every customer to see who does this code still have any uses available, you

Speaker A: know, terribly tedious if we think about it from another layer of abstraction. One of the things that I benefited in my career was from um, was a stint as a database architect, dba. And if I were to solve that problem, I would say, okay, I'm going to open up the LMS's backend and I'm going to issue SQL commands. And that was a, I mean now you're talking code basically select from ah, star where customers field is, didn't use and blah, blah. Now you're, you're human, right? It's Right. I didn't know I could ask this question.

Speaker B: I probably could have gone to support and said, can you, you know, get a technical, uh, you know, person near or something to query the database and give me this list? Um, and then waited, you know, to

Speaker A: week, two week, three weeks. Hey, you're bumping again.

Speaker B: Right, right. But instead we were able to do this like, you know, in the space of a day. We had the list of customers, we had the contacts that, that we should email. Um, and we were able to get it out there quickly. And I know you would. I uh, think you had put in your notes like, so was it successful? And be honest, like was one of your questions. Um, and I'm happy to be honest or candid with you about it. So we didn't get that many people to register for those particular experiential courses that we were, you know, like trying to promote specifically. But we did get um, a lot of people responding just to use their passes for other instructor led. Uh, so it was still a win, right? Uh, you know, much better than like not either not doing it at all or you know, waiting um, for, for another team to provide us with that info.

Speaker A: That is phenomenal. And I, uh, mean this is what's coming for us. And it's so important Lily, because you're able to take the initiative and walk it through the steps and basically go here, here's the email, here's the draft email, here's all the stuff, here's all the people to contact. Anybody in marketing that was in charge of any of that kind of distribution, just go, oh, bless you. You've given me literally everything including the draft email. So in. But that's empowered you to be a lot more proactive and effective down the road. I was going to say efficient but effective because you solve the problem and you learn something new out of it. Rinse, wash, repeat. Now I think you have a couple other cases like um, what about like let's get into retention.

Speaker B: Sure.

Speaker A: Certification.

Speaker B: Certification, yeah. Um, so I started maybe with a little bit more open ended questions for Claude and our CSMCP with this example. But I wanted to understand what are the correlations for accounts that have a certified admin? Uh, are they healthier? Uh, do they have higher adoption? Uh, what, like what are some of the positive correlations basically of um, having one or more certified admins at an account? Um, and so I did start kind of broad like with my prompt and in Claude. Um, and this, over the course of my conversation with Claude I ended up asking about 23 questions.

Speaker A: Holy cow.

Speaker B: Um, you know, so refining as we went and you know, iterating, um, and some a trick that, that actually my manager shared recently I think is a good one. Um, when you're working with AI assistants to say like show me the math. So um, it's, you know, AI is quite quick to come back and say oh well, 56% of your customers are, you know, blah blah, blah. And it's like okay, what is that actually based on? Where like what are they including or filtering out? Um, you know, how, how is this, how are they arriving at this number? Um, so show me the math or show me, you know, sort of the numbers, how you got to. This can be a helpful way to verify like the accuracy and assumptions that AI might be using because I'm the first one to tell you that um, you can't just uh, assume everything of course is accurate, um, right off the bat and it is going to make some assumptions that are faulty and it's going to sort of act. Uh, another metaphor that I heard recently is that Claude specifically can be kind of like a stubborn teenager. And I think this works because sometimes you have to remind Claude of like what it knows how to do in order to get it to do it. Um, and sometimes you have to like point things out. Like when I say I want all of the active, you know, subscribers, that means I need you to look at the plan expiration date and it should be in the future. Like you should know this, right Claude? But um, anyway, you do need to um, check its work. Uh, you do still need to have some understanding and context around your data or like the underlying kind of data in your systems. Um, otherwise you will definitely get back from AI things that are not quite accurate. Um, so, you know, circling back to the, the certification, um, conversation and trying to kind of understand some of the correlations what I, you know, ended up finding, um, through that conversation and kind of back and forth, uh, where you know, along the way I told it things like, oh, don't include anybody with a um, gainsite email address. You know, I don't want to like, uh, you know, include employees basically as an example or you know, don't include accounts that are um, at risk or have churned, you know, or something like that. Like you, you know, you, you want it to have a clean data set. And so throughout, through some back and forth we figured out like this is who you should be including and excluding and so on. Um, and so I found that um, like for example about 41% of our CS customers have at least one certified uh, admin. Um, um, and then a little detail I was able to discover too is that there's about 250 people who have earned a certification tied to a personal email address. So we don't actually have them associated with an account. So that's like important to know about where there may be gaps or you know, things might be underrepresented actually because of, you know, those, those sort of nuances about your data. Um, and there, there were some other interesting details. Like um, our four most certified customers had 10 or more um, certified admins at those accounts and their ARR was like um, 750,000. So there's a sort of somewhat obvious correlation there that the more the company is invested in your products and services, the more they're likely they're also to be invested in training and certification. But, you know, sort of back the data backs that up, which might have been, ah, you know, a hunch at the outset. Um, and that uh, in general, accounts with at least one certified admin have a higher ARR on average compared to those that don't, um, have at least one certified administrative.

Speaker A: That's wonderful. All that stuff is so hard to measure if you don't yeah. Now you can kind of spar with Claude and go, give me a little of this. What do you think about that? And you know, it's connected. You're checking its work, which is anyway it should do. But now you're able to demonstrably pull out or tease out things that help the company, you know, both yours and your customer. Right to go. Look, uh, I'm not gonna, uh, you know, I'm doing, I'm developing admin training right now for Vast. And that question is always on everybody's mind. What's the value of this? Customers, mine, everybody. But you're correlating that. We can't quite get to causation perfectly, but we got a strong correlation with a high end, you know, your number of data points.

Speaker B: Yeah.

Speaker A: And now you're going, look, now you can promote that. You're also in the gain tight skill jar ecosystem. A company that thinks a lot about EBITDA roi. And now this goes back to the board to say, oh, well, there's a number there that leads to increased ARR. It's always hard to build a certification program. Again, certification, high stakes. There's a proctoring, it's value. It takes a lot of time to develop. You want ROI back on that? Now you're actually sewing the other way. This is not, uh, every company is like, well, I want to make money off of that. No, don't do that. The money comes as a second order derivative of the work that you have done and that doesn't show up anywhere. Now you're actually getting to that kind of data without having to be the OPS person spending months developing that. So that's amazing work and I like the output. That's great output.

Speaker B: Yeah, yeah, this was encouraging. Um, you know, when I shared it as ah, an example of like here's something that I tried with Claude and our csmcp. Um, you know, my manager's comment was like, this is good data to support, you know, the case for pushing for more certification. Right. Um, so another nice thing is that like after I've, you know, I've done this one time, I, I, um, can repeat it, uh, or you know, I can take those, the same approach or similar approach and what worked and apply it to other programs of ours, you know, to better understand their impact. Um, as opposed to if I was, you know, working with ops, for example, again in the old, you know, world, like it would be probably a year before I could go back to, to them and say, hey, can you refresh the data? You know, we Want to see how, you know, things have changed or, or what have you. But instead, um, you know, I can do this anytime.

Speaker A: You could craft an agent that runs it automatically too, couldn't you?

Speaker B: Right, right.

Speaker A: You go, okay, I've done this work and what I try to do is take a page out of the playbook of Jess and others and go, okay, we got this prompt, I got everything working. Now let's wrap this up into something that runs and gives me a report. Let's add another one. Let's add another one. And now you have a suite of these tools. Um, okay, I want to, because for time's sake, move into the next topic area. But other use cases that we could talk about are like flagging at risk accounts that have that low training engagement, meaning we got a lot of support tickets. They aren't engaged. There's a problem there, and that is that call to action that would show up in maybe Gainsight for the CSM to go. I need to prioritize these folks and get them to participate and express the value of that. The others are, um, now we'll go with the incomplete learning. Um, you might have people that have kind of dropped off at some place in a learning path or a sequence of learnings and now you want to like, try to show those folks with a drip campaign and digital customer success. Okay, hey, I see that you've started this program. Here is the next step in your journey, or this is why. And then bumps and bumps and bumps. And now you're helping hold somebody's hand a little bit along the path, keeping them from getting distracted, but now getting the back on product. So these are really great use cases. Is there anything more in your head that you want to comment about with those things? And these were big wins. These are kind of the things that we were trying to do in Gainsight proper. And now we have all of this community. We have Skill Jar, we have, you know, px, we have all this stuff and we have a suite of tools that can come together to give us information.

Speaker B: Yeah, well, one thing I would say is, like you mentioned earlier, um, you know, in, maybe in the past you would have used APIs perhaps to, you know, get at the data. Data. Um, well, like, that's not something that I would have been comfortable with or, you know, knowledgeable, uh, uh, with. So again, I like, would have been dependent on somebody else. Um, but I again, like, as with so many things related, um, to AI, like the fact that these AI assistants, you can interface with them in a conversational way with you know, natural language, um, is really ah, an asset to those of us who are not super technical or developers or what have you. Um, um. And again you still do need to know your data and understand the context. M. We've been talking a little bit internally about Will mcps for example, um, change how or what we train our customers to do or make what we're training them on moot or unnecessary when they can just go to an mcp, um, or actually an agent and say go build this report or dashboard and Gainsight CS or something like that. Um, but what my experience with um, experimenting with the AI assistant and the MCPS and stuff tells me is that the users, our customers will still need to understand the fundamentals of how a tool like Gainsight CS works. You won't just hand over to the agent, um, all of, all of your work and assume that it's going to do things, you know, correctly or um, the way you know, you, you intended and um, give it all that power. Right, like to email your customers or you know, send them EBR decks or whatever.

Speaker A: Yeah, the wrong ones or other customer data or stuff that's incorrect. So you're actually talking towards the limitations where yeah, the CLAUDE could give you something that looks right. You've already talked about that previously and it isn't so the caveat or the cautionary tale to anybody in this with wrapping up. What you're just saying is be skeptical. Know your data doesn't mean you have to be a DBA or anything. It's just know where it is, know what's contained. And then I think we can even shortstep this topic by saying you also have the uh, proclivity now to be able to, to work with your data hygiene, which has been a problem for us. That's like this forever, Lila. I mean, yeah, how many times do we do something like oh, this is garbage or we could do a Salesforce loop or another Salesforce cleanup because people are busy and somebody might just key in something wrong or is an accident. And you don't want to cede control of all of these things such that, well now AI has taken over and we let it. So you're expressing to me the limitation of like what? Look at the data, make sure you understand it. You don't have to be a programmer or developer. Right now you're talking with your, your, your you know, Claude or ChatGPT or whatever as a human, kind of like a smart intern or you know, a new team member that's been really well Trained, but they're there for you. Um, and then, then you're talking about governance and making sure that.

Speaker B: Yeah.

Speaker A: Um, that we have a process around it such that this is where I put my enterprise security and, you know, CISO type hat on and I freak out because I'm like, okay, if I were to give, uh, like, say, for example, I. One of the things that I didn't have this scripted, but one of the things that I experienced at Atlassian was something that was very strange because we're looking at our data and we're doing a lot of analysis and we're seeing kind of a slow decline in certain learning properties. Like, and that was a bit worrisome until we understood that there is a natural. Uh, this is nobody pointing any fingers. And it wasn't that substantial, but it was noticeable that we want to be careful about. How do we build the piping with MCP type tooling from. Not from us internally. Like, I'm, I'm working in an LMS and then I'm doing those reporting. That's safe because I'm here. But what if we turned the spigot open to a customer and now they're looking at a knowledge base and, and they're looking at my lms, but they're pulling data on their own and having Claude teach them, not us. Now you've got a problem that you're introducing. Well, people are dictating their own learning journey, but we need to shape that and structure that in some way as educators, because you could get a warped experience.

Speaker B: Sure, sure.

Speaker A: Right. Does this replace a BI tool, though? That's, uh, the only other thing that. Or is it complimenting it?

Speaker B: I don't think so. Um, I do think they're more complementary. Um, in my limited experience, I think the mcps are really helpful at doing quick analysis and discovery, answering, uh, questions that you have in the moment. Like you shared earlier a very familiar, um, scenario of like, my OPS person is out or offline and, uh, somebody's asking me for a training report, basically. And I don't do that all the time or rarely. So I don't remember, like, how do I set that up or run that report directly in the LMS or whatever. Um, that's something that the mcp, for example, could help you with, you know, very quickly. But, um, but it's also limited in that, like, I'm the only one that sees the output from the, the mcp, unless I, you know, of course, like, pull charts together and put them into a presentation or something like that and share that with others, but on an ongoing basis. Like if I want leadership to have visibility to our metrics, we still need a dashboard. For example, you know, in like our CS application.

Speaker A: Ideally, yeah, but you could scaffold it and take that to one of your ops people and say, could you please now, on your timetable, put this into our system and have it updated. Now you're moving on.

Speaker B: Great point. Right. Like, you kind of can build like a proof of concept in a way. Um, you know, with, with your AI assistant and MCP access and say like, this is what I, I want to see in our BI tool or CSP or whatever. Yeah.

Speaker A: Oh, that's so cool. Okay, I'm getting more excited. I think the art of the possible is showing through here. Let's, let's start walking out the door by talking about talking again, breaking the fourth wall. Audience, this is your time to shine. Um, I'm really wanting to get to some actionable things that you can do, any of us can do. Go out there, do this, do that, do the other thing. Try experiment, be in. And I'll lead into this by saying, carry your, um, um, skepticism, but take an open mind into this. Um, okay, so let's say I'm a CE leader listening right now. I'm sold on what you've talked about. I'm really excited about this. What could I do tomorrow to start working towards this and open up the capabilities?

Speaker B: Yeah, I mean, I think m. Step one is potentially finding out if your LMS or CSP vendor, uh, has or is working on an MCP server. So, you know, that like, that's a limitation, obviously. Um, and my, you know, my two cents is if they're not even working on it or you know, they, they act like they don't know what you're talking about or then I might question, like, how long you want to stay with that vendor. Um, and I'm thinking, you know, outside of like Gainsight products, m, my documentation teammates leverage a third party platform for managing the help center documentation. And that vendor right now that we're using, they don't have an MCP and they don't have any plans to build one. And to me that's a red flag because I want to be able to ask and answer questions quickly. I don't want to wait, um, you know, to ask my colleague when they're back online, you know, whatever, Whatever. Right. Yeah.

Speaker A: Yeah, that's cool. So for that process, would you think that you need to get it involved procurement? I mean, I know that the One of the platforms I use has this and I played with it. I actually like this better than some of the stuff that a lot of folks are starting to push out because this is one of the top needs I have that wasn't really being addressed is I need information quick. I don't have time to understand how to do all the reports and stuff. Yeah, darn. Well don't have time to make a report with all of your little drop downs. And I don't want that. I want to explain to a computer law, give me this information. If you can't, you can't tell me why. And like. Okay, so another thing, uh, okay, to answer that question, does it usually have to be involved in that? If.

Speaker B: Yeah, I mean, I'll, I'll tell you, of course, candidly, again, like, I really only used MCPS in the context of my job. So, um, it. As far as I know, if you're, you know, for, for work purposes, it needs to be involved to perform like security reviews and approve, uh, you know, your AI assistant, ideally enterprise license, um, because this, it will be accessing your customer data. So, you know, you need things to be secure and to protect your customers data, etc. Um, so I think that's, yeah, sort of, you know, table stakes. Um, to the best of my knowledge.

Speaker A: Yeah, do your homework and other. What I always say to folks is, hey, there's so much potential, potential here. But I'm, I will go to my, you know, chief security officer or something if I don't know and say, can I do this? And because here's what I'm trying to do, here's the tool I want to use. Yes or no. Have you vetted it? Most of the time. That's cool. Or I, uh, know lasting. Uh, we actually had an environment that was like a sandbox where you could play with anything, but it was very secure. And that's where we were encouraged to play and experiment. And if something, we needed to take something into production after that, that's where we'd work with our security teams and such to bring that into the fold. So again, the, um, message is just don't go and do things and break stuff because you can get yourself in big trouble right at the same time, just ask because that usually is like, yeah, we want to. All right, we've got only a couple minutes left. Let's start, um, working out. This has been a really great conversation. I appreciate the time you spent on this. Appreciate you. It's great to work with you. Um, some of the things Let me kind of run down this and then take a minute to do any other pitches or leave us with a great, uh, enlightening message. Um, but MCP service, this is not just a technical curiosity anymore. In fact, it's a power shift. We in CSR CE are now starting to. We know we're always last in line, right? Oh, I need stuff. We're not going to get it. This is something to drive towards because we'll liberate you from that, you know, that priority pipeline, which is a real thing and there's a reason for it. So I'm not complaining. You don't need to be an engineer. Neither one of us are. And we are having, we're working out loud. Um, in fact, I'll probably send another link out that we were doing a survey via C Lab about how are you using, how are you learning? And in fact, to any of you out there in the audience, I really want to learn from you. We want to learn from you. Um, step up, share what you've got. Let's get good non trivial stuff out here. Lyle, this was a great conversation because now it's opened up something that I didn't really know much about, honestly. Um, and you know, the other thing I would say is get out and learn. Just learn. Try stuff, don't be afraid of it. But then learn and share. Um, I'm really heartened to see how skilled you are by Gainsight. Gainsight brand is really building this ecosystem of AI. It's very complicated to start getting your head around. So I know you've got pulse coming up. Probably going to have a lot coming on that if you can't go. There's online stuff. Um, m. This ties into a lot of really big themes folks. So what we, you know, Lyle and I have been working together for a decade, right? We're always learning. We're not the ivory tower, but we're getting out there and this is transferable across all disciplines. If you're an enablement, you know, partner, sales, whatever, L and D, this stuff is really coming to the fore and you should, you should embrace it. So with that in mind, Lila, anything else you want to say to the audience? Thanks again for being on the show.

Speaker B: Thanks so much for having me, Dave, and uh, you know, entertaining the idea of talking about something new that, you know, neither of us are experts on. But we're excited to be learning and sharing and um, and you know, I liked one of the questions you had shared beforehand about like, what's a good first question? Maybe to Ask Claude. Once you've got your MCP connection established. Um, and so you could go about this in a couple different ways. Like if you have a very specific use case in mind, like the one perhaps that I shared about wanting to understand the impact of admin certification, you could start with that. Um, absolutely. But you can also show up with a very open ended question, um, more along the lines of um, look at my training data, Claude, um, and surface insights that you think a CE leader or my leadership might care about. So giving it more like free rein to um, go and analyze the data that it can access, access um, your training data and um, share analysis insights, um, based on what it imagines would be useful to a similar leader, to yourself or your executive team or something like that, and just use that as a jumping off point. It's going to share things that you're like, yeah, yeah, yeah, I know about completions or enrollments or whatever, but know, tell me more or what else? Um, so that's like a very safe, easy way to get started.

Speaker A: That's great, that's great advice. I appreciate that. Appreciate you and okay, so now we're walking out the door. If you want to learn more, we have a podcast website, of course, customer education. You can find all of our stuff there right on all the podcatchers out in the community. And again, just like Lila's doing, we're sharing our experiences, working out loud, trading, you know, gems and things and even successes and sometimes failures. I know I have plenty of them. I learned from them. So you can always Find us on LinkedIn. Special thanks to Alan Kota for our theme music and if it helps you out, please help us out by subscribing to our show. Get that, get out an Apple podcast, Overcast, Stitcher, Spotify or whatever and leave a review. That ah, really helps us more than anything and your audience. Thanks for joining us. Get out there, educate, experiment just like we're doing here and find your people. Thanks everybody.