Strategy Meets Build: Putting Agents to Work on GRR and NRR

[Speaker 1]
Hi everybody, thanks for joining us. My name is Francisco Lozano, I lead commercial and go-to-market for Atlas, which are our agents here at Gainsight. I have three exciting members of my Gainsight team here that are gonna walk you through their experiences, what they've seen as our customers go through this journey of everything being human-led to starting to deploy agents and what our customers have seen and what our teams have experienced as they go through that.

So Chuck talked about it this morning. I'm sure a number of you have already started to use agents and AI in your day-to-day work lives to make your things easier. Could be things like prepping for a QBR, summarizing meetings, or understanding risk if you're getting more advanced.

I'd love to just take a quick poll. How many of you are starting to use agents in your work lives right now? About half.

That's pretty good. That's great. So let me just sort of set the stage so you sort of understand what we've been experiencing.

Talk a little bit about where AI is right now on a macro level and how it's being applied to customer success. Where agents fit right now in terms of both strategy and execution, and then we'll get into the panels themselves, right? Like half of you in the room?

We're about 50%, maybe a little bit higher in terms of organizations that are deploying agents in their work right now. There's some companies that are obviously moving a lot faster. Some companies are a bit more cautious, right?

Sometimes you're probably getting pressure of maybe your VP or your CEO. We need agents right now. We need to run.

You can't do that until you actually have the foundations in place. So that's kind of the stuff that we'll talk about. There's a big, big shift in gain sites undergoing it right now of moving from seat base and peer usage to actually outcomes, right?

I don't want just a bunch of people sitting with hands on the keyboard. I wanna make sure I'm driving outcomes. I want you to charge me for those outcomes, right?

And then looking at how then is that affecting net revenue retention, right? Right now it's anywhere depending on where you are and what segment, 80s, kind of low 90s, right? So how do we start using agents to start driving towards those key metrics that we want moving, right?

What was our past life before, right? We used to have very manual playbooks, right? We would have CTAs, calls to actions that were all manually entered and the CSM would have to go in and look at it, right?

But now you got a lot more volume. Everybody's books are growing while the headcount's probably shrinking or staying flat, right? So how do I start using AI and agents to fill in those gaps where I don't maybe have, right?

I'm really, really stretched. Again, Chuck and Grant talked about it this morning, right? Especially in our long tail.

How do we start using AI and agents to serve those customers where I just don't have? I have a stack of to-dos this big. There's no way I'm gonna get to all those accounts myself, right?

So how do we put those context aware agents to work, right? What intelligence can it pull from, right? And then looking back, everybody's got strategy, but how do I execute that, right?

That's what we'll talk about kind of this afternoon. So we have two groups, three groups actually, right? Customer success lead, Steve Stone, that'll talk about the experience of on these accounts.

Kate Neal, who's from our advisory services that will talk about kind of how they start going from strategy into practice. And then Troy, who's from our professional services and forward deploy and engineering team that are actually hands on keyboard, building those changes and agentic workflows in the platforms themselves. So without further ado, I'll invite my three panelists up here and I'll let them introduce themselves.

I'll be going from left to right. So Troy, do you wanna introduce yourselves? Yeah, hi, everyone.

My name is Troy Woodwarne, director of CS onboarding and partners. Work really closely with our FTE teams responsible for implementing the agents once the team has defined what we're looking to do and what we're trying to achieve. Steve.

Steve Stone, I lead our enterprise customer success function at Gainsight. Work very closely with both my peers from the advisory and with the FTE and professional services team. It's been really fun to work actually across both of these teams to not just define how we're using those agents ourselves, but actually putting this in practice for our customers as well.

So thanks for coming to our session.

[Speaker 2]
Hello, Kate Neal. I lead our AI strategy on the advisory services team. I made my way to Gainsight through their acquisition of Staircase AI.

So I was employee number five at that startup and really love working with our customers to help them make this transformation in their customer success that AI is sort of demanding of us.

[Speaker 1]
Great, thanks. All right, so we're gonna get into the questions. We'll do about I think 20 or 25 minutes of questions and for the last 10 or 15 minutes we'll open it up to your questions and make sure you put those in the sliders section within the Gainsight Pulse app.

All right, Troy, I'm gonna put you on the spot here first. So when agents join a CS team's retention and growth work today, what does that actually look like in a customer environment? What's different a year ago from today?

Yeah, it's a really interesting question. As you mentioned earlier, customers are going through a route of really different places in their maturity journey. But if I look at the trends of what was a year ago and kind of what is today, a year ago there was a lot of experimentation.

Basically everyone tended to get a kind of a license, go and play, go and figure it out. A lot of that is very much call and response. So it's I need this information, I'm gonna prompt this and then I'm gonna get the information back.

Whereas now it's moved into agents are always on. So they're always in the background working on them from that standpoint. And it's also started replacing work rather than just enhancing work.

So previously it helped me with my email, build this deck for me from that standpoint. Now some of those items are starting to happen automatically and then building in those human in the loop cycles when they're not 100% sure which direction to move that forward. So they're kind of the big changes.

And I say every customer is going from, I've seen customers who are still getting access to these things all the way through to talking to a, kind of you had SAP earlier and I was talking to other customers who have built 50 plus agents in the last couple of weeks, just really, really cycling up and speeding up that agent, a gentrification of what they're doing. Steve or Kate, anything you wanna add to that? I think the maturity is the biggest call for me, right?

It could be 50% of you guys, raise your hands, the other 50%. I was kind of making a joke earlier. I pulse over the last couple of years, we've been talking about agentic AI.

Some of the conversations were what is a cloud, right? And that's a really hard hurdle to overcome, right? But while we have varying levels of maturity and people are starting to experiment, don't feel like you're left out if you are in that camp, right?

That's how I would say to start, is to start experimenting. That's the best way to learn. You'll get it really quickly.

Become a builder yourselves, a lot of you in the room, which is really great. Good. So good segue for Kate.

So you probably have a lot of customers that come to you and say, I wanna do all this kind of stuff with agents, right? So when they wanna do that, what would you say is really the business problem that lies underneath that, right? Where do you see people get stuck and where do you see them being able to accelerate?

[Speaker 2]
Yeah, so I think we're all experiencing leadership teams and boards kind of pushing, do AI or adopt AI, and often without a strategy behind that. And so I think that one of the things that you can do as you're kind of embarking on that journey is to make sure that you have a solid strategy from the business owners of what you're applying and why. And I think also to start with the end in mind when you're designing an agent, what is the outcome that you're driving?

What are the guardrails that you need that agent to kind of operate within? And then you get to things like the tooling, the features, the vendor demos, right? So just making sure that you've got that order of operations.

And then the other pitfall that I see is customers who just slap AI over the top of our existing workflows, right? We've got a lot of things that we've done in CS that are from 2009 or beyond, and it doesn't work to put 2026 technology over a 2015 workflow. And so we also need to be mindful and innovative about how are we redesigning the motions and the processes that we run to deliver and execute on customer success?

How does AI change that? What can we take away? What can we add?

What new signals do we have? And start from that point rather than just trying to make an existing process that isn't really working faster, right? You're just gonna get to the same results quicker and what we wanna do is have a different result.

[Speaker 1]
Just wanna say that, that's a really good point. Just what we're going through with our kind of renewal agents a lot of the times the processes that are put in place right now are due to human constraints. We can only get to a customer 180 days out because this is the resources we have.

As you move forward in this world, you don't have to have those constraints. You can say, actually, when do I want to first contact a customer? When do I want to first do these actions?

Without those manual efforts in place. It's almost highlighting the inefficiencies that you have. Exactly.

And making those processes better. So how do you, okay, feel free to ask. How do you bring that back to GRR, right?

To the key metrics that maybe, to that's important, right? Not necessarily workflow driven but making sure you're looking at that key metric at the top still.

[Speaker 2]
Yeah, I mean, I think that's a struggle that we've had in CS for a long time is that we measure ourselves and our role and our success in terms of activity. And like you heard in the keynote today, things are shifting to outcomes. And so when you start at the outcome, you can work your way back.

And so if we know that our outcome is to retain and grow customers, what are the activities that drive that? And then I can start working backward from there on, okay, renewal or expansion is the outcome. What has to be true for that to happen?

What has to be true in the step before that to make that true, right? And so I like to do that backward planning to make sure that what we're designing is getting to that outcome rather than the activity, right? The activity is how we deliver the outcome but it isn't the outcome itself.

And so an EBR or a QBR is the activity but the outcome from that is strengthening the partnership, growing that expansion or retention motion. And so I think that ensuring that before you design an agent you know the outcome that it's gonna drive is gonna help you get there.

[Speaker 1]
The fact is, I'm sure everyone's going through this right now budgets are starting to get pulled back on tokens and the likes. If you're not able to tie it to those outcomes, you're not gonna get the budget to go ahead and do this in the first place. So you have to start there to be able to get that exact buy-in.

All right, thanks. Okay, Steve, we're going to you. Fun one everybody wants, agents versus people.

Yeah. So where should an agent own the work? Where should it assist?

And where should human judgment still be the leader in this, right? And how have you seen customers draw those lines? Yeah, so I kind of want to start with the last piece, the human judgment.

We've heard human in the loop throughout the keynote. I think it's really important to understand even in the design process before you even get started, humans are going to have to use their judgment to understand, like to Kate's point, which of these activities really need to be even agent-led in the first place, right? So what is the highest value of activity that I can focus my CSMs on?

Keep them very much forefront, but then what can we automate through either agentic or even sometimes a fully autonomous workflow? What's that first part of that question again, Francisco? Yeah, so where should agents lead in terms of the actual work?

This is a huge opportunity. So if you've been coming to Pulse for a long time, we've been talking about digital as not a segment, it is a strategy, right? It was still really hard to do.

It was still really hard to execute on. We talked about the rules-based path that we were in. It took humans to build those rules.

It took a lot of steps in the process to actually execute on a digital strategy. I think we're right now, a couple of the baseline key pieces are the same, right? So let's start with what the agents can do.

Where is your data the cleanest? Where can we have those signals that we can pull from that may not even be structured data, but unstructured data to pull in, to have the AI do the work? And then we're gonna start there, obviously.

Yeah. So as we're implementing it, I think it's an interesting hurdle that I see a lot of, is the word trust and trusting the agent. And what I'm seeing is once they start trusting, there's less human cycles in that loop as they know that what is actually being delivered and executed is valuable and able to do that.

So that human in the loop cycle is what we're seeing is we start with this and then you start removing some of those aspects, as it gets better and more predictable.

[Speaker 2]
Yeah, I think another consideration is accountability. An agent can draft a strategy for you, but often the human needs to be accountable for that. I think also humans need to be accountable for the output and the outcome of an agent and to be monitoring that and iterating on it.

And so I think that's another layer where, we need to pull humans back in to make sure that either they're accountable for the actions that they're taking, or they're accountable for the outcomes of the agent.

[Speaker 1]
Yeah, we like to use the term here on the Atlas team, you need to earn the right to autonomy, right? That means you wanna start with more and probably heavier humans in the loop, reviewing everything. As you start to trust it and more comfortable, it starts to get smarter, right?

The context, what it's bringing back, just because it's done it a bunch of times, does not mean it's gonna do it right every single time. I have to run, I found that out and I've run my weekly summaries for months and months and months, and it still comes back with errors, right? So just because it's doing it repetitively, you can't assume it's still doing it perfectly, so you gotta maintain that human in the loop.

Yeah, I do think something internal is a good place to start, something low risk. If you're gonna go with full autonomous workflow, what are some of the things as we as CSMs, leaders, RevOps, what are we doing that's repeatable, that we have the clean data? Let's see if we can build an autonomous workflow there.

We can catch the errors that's lower stakes than something that's actually going to be surfaced in front of the customer, and it's a great place to start. And then the second point on what you just mentioned there, Francisco, is it needs refinement, right? You're not gonna set this agent once.

Your needs of the business are gonna change. Your initiatives, key products are obviously changing much quicker than they were previously, so you're gonna need to continuously refine these skills. It's not just a set it and forget it kind of thing.

It's something that you're gonna have to kind of work and mold as you all grow and your companies grow. Yeah, and a skill doesn't mean perfection. A skill has been easily refined over time, but it still needs optimization.

So maybe you're building them yourselves. Maybe you have an AI team, like we have internally at Gainsight that builds these for us. They sometimes take months of refining these skills before just releasing them internal for our internal use.

So don't just assume, again, that because you got a skill and you built it, it's good to go and you can put your feet back on the desk and the agent's just gonna run. You're gonna get in trouble. You're gonna get your hand slapped, as we have experienced probably internally.

Any time that you guys can, an example you can use of where maybe you had to pull back use of an agent because it wasn't quite ready for general release or a customer's usage? Yeah, so we've seen this once or twice, and it's less around the quality of the output and more identifying things that we didn't even know what was gonna happen. So all of us build processes in place and we try to make it as predictable as possible with humans in there.

And then when you actually go down and revise that and put that on paper and then try and identify it, sometimes when you start doing that, your customers will respond with insights that you just didn't expect. For instance, we had a customer from that time who came back and just asked a whole host of questions that just wasn't expected and realized, do you know what, we need to actually go and build a community page or something like that that's gonna drive this to help the renewal cycles from that perspective. So that's a good segue for you.

So what has to be true in someone's environment for an agent to be useful, right? So things like data, workflow, technical debt. What do you look at and where do you start in terms of fixing or optimizing the current state?

So I was gonna add to the previous question. So the interesting thing, especially over the last year, the models are improving at such a rate that how many times has people talked around here about hallucinations? That conversation was a lot less than it was this time last year.

So the intelligence isn't often the problem anymore. It's the context layer. And so the easy answer is data at the end of the day.

The big problem with data is that you go to any IT company inside your organization, you say we need to clean up our data and that could be a year, two year program from that perspective. So where we really like to start from a data point of view is actually what is the data that is going to add high value context to that agent to drive the outcomes that you're looking to do? It's very easy to be saying it's nice to have this, nice to have this.

But if you can pin down and go no, these are the things which are integral for it to be able to do that, it just helps that. And as you mentioned, starting with that outcome then helps you back that from that perspective in. The other point is very much don't identify the existing process.

Rethink the process that you're trying to identify and that will really build that and set you up a lot better because otherwise you're going to end up having to, you'll get to the end of it and you won't get the outcomes you're looking to really achieve.

[Speaker 2]
Yeah, I'll just add one thing. The data and the context is super important. I think we all know that the output of any AI, whether it's an agent or a skill or something generative, is only as good as the data and context that goes behind it.

And so I mentioned earlier doing that redesign, right? What should this motion look like with that AI first mentality? And a lot of us are in a boat of like our data's messy and that becomes very overwhelming to look at something, a tool like Salesforce and tell your IT or your ops, clean up the data.

When you do that redesign, we can start to point to what are the specific data points that feed this agent and its ability to do the job that we've set it out to. And that can really help narrow the burden of that hygiene and I think get to that result a little bit quicker than just saying, clean up the data.

[Speaker 1]
There's something you, sorry, something just came to head. One of the things that we are seeing that wasn't possible before is continuous improving data. So for instance, one of our challenges we have is contact mapping.

And so therefore, we know that that's really, really hard but being able to use AI as part of the responses to then improve that data is kind of a self-fulfilling kind of improvement loop that we're seeing. Okay, you mentioned something around context. I think this is really important as we're redesigning these workflows for agentic and when human was in the loop and when humans were, you know, and are executing, they're usually feeding context somewhere.

One thing I think is often overlooked is how do we keep that context rich? And so when we're redesigning these plays, we have to make sure that we're feeding everything that the agent is doing back into that context layer because if we just feed, feed, feed, feed, feed agents without replenishing that context, that context becomes really stale really quick. So that's definitely a different piece that needs to be solved for when we do redesign those workflows for the agents.

Yeah, so let's say you do run before you can walk, maybe to Kate, right? So what would you say breaks trust with CSMs the quickest?

[Speaker 2]
I think it is the data and the context, right? If you've got messy data and you turn AI on and your team gets the output, they're the ones closest to the customer. And so if you give them a signal that isn't accurate, they're immediately gonna shut off, right?

And that first impression, that first moment with AI is really imperative to help your team with that change management and adoption process, right? AI is totally different. It demands a totally different day from us than pre-AI.

And so I think, again, making sure that you've got the right data and context feeding into whatever you're delivering to the team before you deliver it is gonna help. And again, I think also educating them, right? No human is ever 100% accurate 100% of the time.

And the same is true for AI. It's improving all of the time, but we have to maintain critical thinking, which is something that I worry about the younger generation coming up with AI and being able to just generate an entire essay, right? Or even in our line of work, I get emails or documents or slide decks from customers that are pages and pages long, and you can tell that they just threw that into their LLM and didn't iterate, didn't go back and forth and deliver that.

And so I think maintaining that aspect of critical thinking, really working with your LLM, right? Of, hey, let's get a first draft, but then I'm gonna go through it and say, these are things that are on point. This is where I want you to adjust and kind of have that back and forth.

So teaching your team, enabling them on, how to work with AI, how to take an AI output and turn that into action is gonna help a long way with that.

[Speaker 1]
No, go ahead. To allay some of your concerns. No, I do think, and what we're starting to do internally is actually reviewing the role of an individual.

What tasks are they doing and what tasks is it no longer, previously we went through a world of SAS, and it's like, oh, the software does that. Now the vernacular is the agent does that. That's not actually part of my role.

And I think we're getting to this hurdle right now from, okay, it's giving me really rich information to, okay, that information is just being put there and this is now what I'm trying to do on my day-to-day actions. And I think it's gonna take a little bit of time, but I think we need to be purposeful around those JDs and what is expected of the individual. And it's likely, job title, FDE has come about, wasn't around 18 months ago, well, it was, but not to the same way as it is right now.

Yeah. Right. So now we're gonna talk a little bit about operating model and change.

I'll go to Steve and then to Kate, right? So things are changing, right? So how does the CSM role itself, right, the team design, the governments change as you start and CS teams start to incorporate agents in their daily life?

Yeah, I think the biggest change you're gonna see is we're gonna be able to, CS teams are gonna be able to move a lot faster and they're gonna be able to focus on the things that really matter with their customers. So how this might show up, we talked a little bit about, hey, budgets are, headcount's not increasing, work is gonna continue to increase, right, but we're not gonna necessarily be able to get more humans. We talked, I forget who it was on the stage, mentioned something around how, that exact point around we're not gonna be growing our teams, but we're gonna be taking on more work.

So I think what you're gonna see is the high-value activities that we know, CS first principles are going to still be very important to the CS role. I mean, CSMs kind of have become a catch-all. What of that catch-all can we kind of outsource to AI and keep them back and get the CSMs back focused on what's really important to drive progress with the customer?

How this shows up is really, they're orchestrating a group of agents, right, so every single CSM essentially becomes a manager of their agents. And some of that actual work that, you know, building a business review that used to take 12, 14 hours that is now done in minutes through a skill or a prompt, you know, that's a massive improvement. Governance, I mean, this is something that's really important.

I think I definitely wanted to touch on this. Where everything kind of had to go through, I won't call it a bottleneck, but it had to go through a single point and then be built out, now can be, you know, done by anybody. I mean, I am not a very technical person at all, and I built my own agents.

I have, I rebuilt our entire risk process with the Cloud Artifact as an output that automatically sends emails and I use natural language to do this. That's really cool, right? But what it does is it allows it to get a little wild, wild west out there and everyone starts to, everything starts to break down.

And we have different pockets of the business running their business differently. So it's something we definitely need to watch out for as we go into this agentic world where anybody can develop anything. Yep, Kate.

[Speaker 2]
Yeah, I think AI exposed CS. In a way that, you know, we always have people that are so busy, they're doing so many things, they're doing so many activities. How many meetings did you have?

How many emails did you send? How many QBRs did you prepare for and execute? And when you automate those tasks through AI, it sort of exposes that a lot of that was busy work that doesn't necessarily lead to the outcome that we want.

And, you know, when we start to replace that work, we're asking, okay, well, what's left for the human? And I think one of the things that we're seeing is the CSM role is bifurcating into two distinct roles. One is more commercial, right?

Where they own the renewal, they own the expansion motion. And the other is technical. They're product experts.

They can take, you know, a customer outcome or a customer request and vet that and help them achieve that result. And so in my mind, one of the most important things any of our human team can do is continue to talk to the customer, have conversations, have continuous discovery with your customer, because all of that becomes context for your AI to create your success plan, to create your EBR. And so when we take away that administrative burden, all of that sort of analyzing and synthesizing of data that we asked our team to do, the things that are left are the important things that require that human judgment and that human finesse.

And that's really where we should be helping our team in addition to Steve's point of how do they manage their team of agents? And I think that's another skill set that's new to CS.

[Speaker 1]
This is the opportunity that I spoke about earlier around digital as a strategy, right? It's showing up now in a way that we didn't really could have ever thought about when we first started talking about this at Pulse two, three, four years ago. It's a massive opportunity to really get the CSMs back to the important work.

All right, so we're on to our last question before we get to the Q&A section. All right, so I have a new CSVP that just stepped in the role. They want to look like a rockstar and deploy AI and agents, right?

They want to start right away. What's the one thing you would tell them to do, one thing to avoid? And then what is like a 90 day sort of crawl, walk, run look like?

Big question. Yeah, big question.

[Speaker 2]
Anybody in the room starting that? No.

[Speaker 1]
So I think you mentioned we need to, the first thing you need to understand is what is the big problems that we're trying to achieve here and how are we tying those to the outcomes that the organization's being able to achieve. So actually understanding that, I do agree with the statement in terms of playing around from that standpoint, but at that level, you need to be able to be solving big problems and being able to actually tie that to improve GRR and the likes. Where I would say is start is identify a cohort.

Now, a lot of our customers are working to the places that they can't get to. That's why we're starting at the long tail because they don't have the people to get to it. So it's, I can actually make an improvement in that area, then identifying that cohort and then building out that agentic process off the back of that.

So that's how I would attack it. There's a few reasons for that. One, it's greenfield a lot of the time.

Two, from a change management point of view, I'm sure there's a number of people here who are still struggling inside of their organizations as to, we're not replacing individuals, we're enhancing what we're doing. So by moving and going into that area, it's kind of, people just recognize that, okay, yeah, it is an area that we can go and attack. The new 90-day VP spin is interesting.

Right? Because my answer might change if we're talking about how do we just get started with AI. But in terms of getting up to speed in any new role, obviously there's gonna be a huge opportunity to have AI expose some of the breakdowns currently within the organization.

You're gonna be able to validate that data. You're gonna be able to go in and immediately understand, do we have the right data to even build this agentic workflow that we wanna build? And rallying that effort as a first initiative that can be massive for the entire organization is really gonna be where you're gonna earn your stripes.

In terms of just getting started with AI, I'm gonna change the question a little bit and just say, hey, how do I get started? I mentioned even what I've built on my own, right, as a non-technical person that started on an airplane. I was just playing around.

I've been told by our CCO, we have got to start being an AI, like start using AI and be AI first ourselves. I just made myself start going into a conversation, right? It doesn't take much.

Build something very simple. Then you can add the layers of complexity on top of it. I tell my team all the time, right, don't come to me with a simple question or problem.

Have you spent an hour, have you spent two hours in a meeting with Claude, right? Have you put the time in there? Because what we really need is we really need everybody to see the benefits for themselves, for them to really be believers and kind of overcome that skeptic hurdle that some of you might be facing.

[Speaker 2]
Yeah.

[Speaker 1]
Cool.

[Speaker 2]
No, I mean, not to belabor the point, but I think don't start with tools. Start with the strategy and the redesign and let that sort of flow into the rest of the process. And then you can figure out what tools and what parts of your tech stack support what you're doing.

I think from a crawl, walk, run perspective, I think crawl is getting your data in your context layer. Walk is AI that's augmenting your humans and run is autonomous.

[Speaker 1]
Yes, great. All right. So now we're onto our fun Q&A section of the panel.

I'm gonna start with the first one and anybody jump in. So for a CS organization that's just starting to experiment with agents, right? Where would you recommend starting and seeing the most impact?

I think Kate kind of has mentioned how we want to design these agents with the outcome first, but I want to take the word agent and deconstruct it a little bit and kind of make that a smaller mountain to climb, right? When we think about agents, we think about immediately our mind goes to this fully autonomous workflow. You don't have to have a fully autonomous workflow to build an agent.

Some of the simple things is how can we use a scheduled skill to run, to serve me up information that's really important, right? Something that's generating just a prep document for our leaders. These are these, I don't want to call them low value activities, but these are these annoying pieces of work that we have to do that take a lot of our time, right?

Like how can we start automating some of that? So starting small with some of the internal pieces is really important and don't make agent be, in your mind, this huge end to end autonomous workflow. Start with something small and then just start chipping away.

To build on that, right? Without plugging, have a look at our software. Staircase has got agents inside of it.

And so they're always on agents that are driving a lot of those kinds of things. So it doesn't have to be a custom made agent that you're starting with. A lot of the software that you're already dealing with have agents in there that you can use that have already had learning and knowledge from multiple other customers fed back into that to be able to then drive those behaviors inside.

It doesn't have to be inside your LLM. It can be not just game type, other software as well.

[Speaker 2]
Let me just take a different direction with this. So the question is the most meaningful impact. One thing that I see all the time is when we talk about AI, whatever form that's taking, we're using it on an account by account basis.

And I think the most powerful thing about AI is what it can tell us across our portfolio by looking at all of the context, all of the customer communications, all of the data in our customer success platform and surfacing for us, what are the signals across my customer base that lead to renewal or expansion or risk or churn or what are the common areas of the product that lead to the most customer friction and then upstream making optimizations to those things so that downstream you're actually impacting a broader swath of your customer base rather than staying in that one account at a time.

[Speaker 1]
Great. All right, on to the next question. What's the biggest mistake you see companies make as they start introducing agents and AI in their CS work?

[Speaker 2]
Yeah, I mean, I think I kind of already answered that which is just kind of slapping it on top of what we've always done, the way we've always done it. We're just gonna get to the same result faster and more efficiently rather than changing the behavior of our team, changing the behavior of how we think and deliver and execute customer success and so going right into tools or letting a vendor demo sort of drive your direction and how you implement AI in your company is probably one of the biggest mistakes I see.

[Speaker 1]
To take it just a step before that, the biggest mistake I see, especially in CS org, is decision paralysis. We could do this, we could do that, we could do this and they kind of go, okay, we're gonna do a load of these different things. I think from an exec point of view, if you can go, right, okay, this is what we're going to do, this is what we're gonna tie it to and actually go and really be purposeful with what you're doing there, then that then shows confidence in the ability to execute and then you can expand from there.

That's what I try and use, I think. All right. All right, we're gonna go on to the next question.

So how should CS leaders think about measuring success, right? If you think of NRR, GRR take a long time to measure, right? So how do you tell people that they're making actually an impact more in the short term in terms of deployment of agents in AI?

Steve. Yeah, this is a big, big, meaty one. Like anything, what I don't wanna do here is number of actions based.

How many preps did we generate? How many QBRs did we generate, right? We wanna get away from those activity-based metrics.

That's gonna obviously get you closer to what you're looking for for the GRR and NRR. You know, what I think we could do, and we could do this here at GameSight, and we probably need to, is really look at some of these exception-based, kind of what we're learning from, we do have agents deployed on these accounts, or we did not have agents deployed within this segment that's not serving up this information, and start cohorting our customers in the varying level of agentic activity, and really understanding what moved the needle there.

And that's obviously gonna tie directly back to GRR and NRR. Kate, do you have any immediate thoughts jumping out at you on this one? This is a good one.

[Speaker 2]
I mean, like a little contradictory to what you said. I think so, GRR and NRR are lagging indicators, right? We all know that.

So when we tie outcomes to GRR and NRR, those outcomes can become our leading indicators. And so you can sort of see how many QBRs are we executing, and that can help be a leading indicator to what your GRR and your NRR are gonna be, as long as you're redesigning the QBR before you start executing it, to be less of a vendor sales pitch, and more of a continuous discovery, a conversation with the customer, and a mutual success plan of how we move forward. And so I think, again, it sort of starts at the end with, hey, we wanna impact these metrics, what are the outcomes and the activities that we know drive that?

And then those can become our leading indicators to make sure that we're gonna see the results that we want, and our lagging indicators. It's more about activity, but quality of activity.

[Speaker 1]
Super important clarification, I'm glad you said, we're gonna be contrary. Activity, the activity is incredibly important. The activities of the agent is measuring the activities of the agent, is what I was referring to there, so.

Let me give you a real life kind of example. Working with customers on the adoption agent, and what they've identified is their sticky features and functionality from that standpoint. And so what we're able to do is look at that customer base, look at what features they're doing, and then use that agent to drive the adoption of those features.

And so therefore, you're able to talk to your product teams appropriately, talk to your internal teams to say, okay, we're doing this, you've told us these are your sticky features, and this is how we've driven the adoption in this, and thus, you should, in fact, drive GRR and NRR from that standpoint, so. Yeah. Yeah, you have some of our customers are using things like monthly active users.

It's more of an immediate impact, right? But in the long term, right, just because you're working on an adoption, that will affect GRR in the long run, right? More people are docking the process, more likelihood they are, obviously, to stick with you and renew and expand and grow.

Okay, we have two minutes left, I'm gonna try to get these. So, how do you think the way CSMs are measured should change, this is probably for you again, Steve, should change the way the role is changing, right? Yeah, I'll try to be really quick with this one.

I think, I mentioned this earlier about getting the CSMs back to the first principles of CS. I think it's really important. I talk to customers all day, every day, and they're all in firefighting mode, and we're not sure where to focus, and we know these things are important, we can't get around to them.

The activities that we are supposed to be driving as CSMs, we need to hold our teams accountable to those things again. And I think that's really, really, really important, which is why I was saying we're on the same page on making sure the high-value activities are being executed on. I mean, I think that's something that we've kind of lost our way in CS, and I think that we're quickly finding our way back to being the important team that's driving progress to our customers.

[Speaker 2]
Yeah, it has to be outcomes, not activity. Like, I think CS needs to stop being afraid of revenue responsibility, because that's how we get budget, that's how we get headcount, and that's how we make a difference, and so we need to own those numbers.

[Speaker 1]
Great. Great discussions. Thanks to the three of you.

They'll be available after if you want to chat. We're obviously over at the Gainsight booth in the other room if you want to chat, so appreciate your time and your questions. Thank you.

[Speaker 2]
Thank you, guys.