The AI-Powered CSM: Inside a Multi-Agent Expansion Play

Julia Nimchinski:

And we are transitioning to our next session.

Welcome to the show, Desiree Diedeker, Principal Customer Success Manager at Gainsight. How are you doing, Desiree?

Desiree Dietiker:

I’m doing very well, thanks for the intro. I was gonna do a little intro on my slide, but this is perfect.

Julia Nimchinski:

Yeah, we are super excited to host you here, and there is one traditional question I’m asking every session here, since this year was intense in experimentation, deployments, you know, agendifying everything that could be agendified. In GTM, so curious to hear more about Gainsight Agentic platform, and some early results, ROI, case studies, anything tangible that you can share.

Desiree Dietiker:

Sure, so I actually have, we had our annual, executive summit a few weeks ago, we call it CXO, where we gather, you know, chief customer officers from a bunch of different companies, and this is less, like. At Gainsight, we’re still building, you know, some of these agents, but there are… there are people who are really doing a lot with agents already. And, I had a customer share an example, which I thought was really amazing.

He basically was telling me how his organization built, like, an internal AI solution, and it’s supporting their entire customer lifecycle. So, think from onboarding all the way through.

renewal and expansion, and they’re thinking about it in terms of, like, an agentic co-pilot model. So, the agents are acting as both co-pilots for CSMs and their TAM organization.

They’re doing things like surfacing next best actions, renewal risks, expansion opportunities, all of that stuff, based on real-time customer data. And what that approach has done for them is delivering over 15 consecutive quarters above 120% NRR.

So, not many companies can, can brag about having 120% NRR these days.

So, that is how agents have been, you know, helping them support, you know, it’s been reducing support costs.

And then it’s demonstrating clear ROI from AI-driven automation.

So, this customer is very interesting because they’re actually building agents for their customers.

So, not everybody has, you know, a global AI operation dedicated team to manage the development and optimization of those agents.

But, what my key takeaway from this was, is that they started somewhere, and, they’re… they’ve been able to build this model, and fairly quickly.

So, that’s kind of my key takeaway there, and I think a huge, shiny example of how agents can help support that customer journey.

Julia Nimchinski:

What an amazing beginning.

Let’s see it in action.

Desiree Dietiker:

Alright, let me go ahead and go into presenter mode here. Just gonna share my screen. All right, well, I’m gonna, skip the, skip the intro here, since you already did a beautiful job introducing me, but just high level, Principal CSM at Gainsight.

And what I want to talk about today is how AI and agents are really transforming customer success, and not really, with the tune of, like, replacing humans, right, but by amplifying what we can do.

So, I’m gonna share how I use Agentic tools, like an insight agent and an expansion agent in my day- to-day, and then end with some takeaways, that you can apply to your day-to-day.

So with that, I’m gonna cover, kind of, why agents are becoming more and more essential for customer success today, how humans and agents can really work together inside of what we call, like, a multi- agent framework, and then share some kind of lessons learned and, including how AI became kind of a forcing function in my own workflow.

So, we all know the software industry has changed.

Products are built, they’re bought, and they’re supported completely differently now. Agentic competitors are emerging fast. I mean, I think we all see the ads on TV with Michael McConaughey with, like, Agent Force and all of that, and, you know, we really need to be rethinking how we are engaging and retaining our customers, and how agents can help support that motion.

So, think about how we used to build, right?

Everything was manual. It was very high effort. Now, anyone can be a coder.

Like, look at this example here of this guy who built a math game for his kids, right? And they’re using tools like Cursor, which is a company that went from 1 million in ARR to $100 million in ARR in 12 months.

Like, that is faster than any company ever. They didn’t hire a massive headcount, they didn’t have a massive marketing budget, but that shift has started to raise the bar for customer expectations for our customers, too. You know, if we can’t deliver faster outcomes, customers are going to start thinking about building solutions themselves that… and that is what makes retention super existential.

So… If we think about it in terms of a moat, right, the old moat is gone.

You know, it used to take years and millions of dollars to build a SaaS company, but today, a few people with the right agentic tools can launch, scale, and compete, even with your companies, faster than ever. So, agent-led startups, they’re flooding the market.

Buyers are rethinking everything. Build versus buy has always been sort of this, like, scary thing in the background, but it’s a real conversation happening at most companies right now. So, what maybe was your moat last year?

That may already be obsolete. Retention is no longer just a CS challenge, it’s more of, like, a survival strategy. And so, how do we move forward in this new market?

Every customer matters more than ever. That’s the answer, right? Retention is now existential.

We all know budgets are tight, switching costs is lower, expectations are higher than ever. So, as Sam Altman said, we are in the fast fashion era of SaaS, and as a CSM, I see this every day.

You know, you have to be able to prove the quantifiable value of your tool, and not just to your biggest customers anymore. So that leads me into thinking about, you know, how we used to do CS, right?

It used to be very high touch at the top, this 80-20 model, focusing on strategic accounts, and then assuming that the long tail would kind of take care of itself. That approach worked in a high growth years, but it’s not working anymore. Budgets are flat, expectations are higher, and teams are leaner.

So, agents are how we can gain leverage, especially as you’re thinking about how you scale, and you move into that down-market segment, because that 80-20 model of support, it’s just not going to work as we’re thinking about growing and, growing our own businesses.

So, we evolve, right?

So, from humans executing workflows to humans and agents working hand-in-hand and side-by-side to deliver outcomes to our customers.

Agents are not necessarily replacing the human work, they are supporting it. So, and how agents do that is by taking on, kind of, the heavy, tactical, day-to-day tasks that are super time-consuming.

They’re sucking up your energy. It’s all the stuff that you don’t want to do on the day-to-day, or that you physically cannot do with a large book of business. So, agents are freeing CSMs to focus on the more strategic and human parts of the job, the things that only humans can do.

That’s building relationships, it’s driving outcomes with our customers, and it’s leading those customers to success. So here’s the opportunity, right?

So you think about it in terms of… human-led, right?

So these are the things that only humans are good at, right?

So, empathy, right?

Have you ever seen ChatGPT be super empathetic without coaching it how to do so? Not really. You know, strategy, executive alignment.

People still want, and customers still want, that human component of the relationship, and that’s never gonna go away, right? So, that distinctly human work that AI can’t really replace.

That’s where we should be spending our time, and that’s where we should be putting our people. And then you have digital-led, right?

So you’ve got your scalable programs, you’ve got email, you’ve got in-app communications, community.

But remember, I like to think about digital CS not as, like, a segment, right?

So, you know, you think about how you’re supporting that 80-20 model, that long tail.

Digital CS shouldn’t just be that segment, right? It’s a strategy. It’s something that you can leverage across that entire customer lifecycle to really drive engagement, consistent engagement with your customers.

And so that’s even including your strategic customers.

And then you have agent-led, right?

So things like automating renewals, you know, serving up insights, and then maybe some engagement for, for, you know, the lower end of the long tail, you know, for those customers who might not be getting any human interaction, right?

So an agent is always going to be a net positive for that group.

So, you’re really freeing people to focus on only what they can do, and you’ve probably heard us talking more and more about, like, RAS, so that’s the idea of retention as a service.

So… That’s kind of about orchestrating humans, digital programs, and agents to really work in sync so that every one of your customers feels supported.

So humans are driving the strategy, digital driving consistency and reach, and then agents handling the tactical work that frees people to focus up, focus on the really impactful things that only they can do.

So, if this is the hybrid strategy, right, like humans over here, AI, digital, I want to share a little bit about how I use agents to expand my reach as a CSM.

So, I think this slide is a really good illustration.

Because, you know, we all know one person or one CSM cannot effectively manage 300 or 500 customers, right?

It’s just not scalable. And even with 50 customers.

it becomes really hard to stay strategic with every single customer. How do you manage all of that? And that’s where agents come in.

So.

I am still managing my strategic accounts directly, but I’m supported by a layer of multiple agents that are helping me to kind of automate and augment the work that used to be fully manual. So I’m saving a ton of time, and I’m able to spend more time on, you know, the strategic activities, like an EBR, or a conversation, or in-person meetings.

That are actually driving value for my customers. And, just to give you a couple examples, we have our Insight Agent, which really helps me surface, like, customer risk, analyze customer sentiment, and then we have an expansion analyst, which is helping me identify growth opportunities within my accounts.

And then beyond these, I am also using AI tools, and like the standard LLMs, like ChatGPT is my preferred, but I know some of my colleagues are using Claude, they’re using Notebook LM, for drafting and analysis.

And then I use even another tool, an AI tool called Gamma for creating decks for my customers.

And all of those things feed into and support my day-to-day as a CSM.

And together, they are helping me scale, they’re helping me personalize, and they’re helping me focus on strategy.

And then that last part here, we have renewal agents, which are fully autonomous.

So, those are managing low-touch renewals for long-tail customers who, as I mentioned, they might only be receiving digital interaction.

And so I think of these agents as more of, like, an extension of the CSM team, so they’re able to cover areas that we just don’t have the human bandwidth to reach.

I always… I like this slide, too, because I think it’s important to differentiate between the types of agents. I’ve talked about a few. There are different agents that really augment work, and then there are those that are autonomous.

So, augmented agents.

I love the analogy of, like, an Iron Man suit, right? So, these are agents that are working alongside me to enhance and scale my impact.

Staircase, one of our products, is an example of an agent that augments my workflows. And then we have the autonomous agent, so that renewal agent that I just talked about.

These are agents, but they’re acting independently based on specific parameters. So they’re executing specific workflows end-to-end.

Like our renewal agent, which might manage renewals for those long-tail accounts that, as I mentioned, don’t get direct human intervention. And both of those play significant roles in that multi-agent kind of layer and framework that’s supporting our entire, you know, customer success and go-to-market motion.

So one is really amplifying everything that I can do, and the other is extending my reach through automation.

And we… one thing that I always like to point out, too, to customers, right, like, and I think we’re all kind of seeing this play out in the market, is we need more data.

We need more data to do our jobs, and there is just no way for us to analyze on an individual basis all of that data. And so we used to see only a few core kind of signals of customer health.

So we, you know, we had things like product usage.

Like, that’s a semi-good indicator. We have NPS, maybe once or twice a year, right? So we get sort of a pulse on the customer.

And then we have our good old CSM sentiment, right?

Like, how does the CSM feel about the health of the account? These are… limited data points that gave us clues, but really not the full picture. And so, where Staircase supports me is that it’s… it’s changing It’s providing me depth beneath the surface, so it’s connecting all of the signals that we’re getting from emails, from meetings, from support interactions, to reveal risks and trends that I really couldn’t see before.

And so, with AI, I’m getting my visibility is expanding beyond what I could track manually.

And it’s been a huge game changer for me.

So, you know, a couple of things that I wanted to share, too, on what AI can surface, right?

So, for example, Staircase can flag a churn risk hidden in an email.

So, look at this example.

You know, it’s a procurement message saying, our team has started an RFP process. Even though the notes sound super positive. it’s signaling that they’re evaluating competitors, so if I’m just going through on my day-to-day and just, like, scanning emails, I could miss this.

And this is huge.

It can also detect, expansion signals. So, in support tickets.

So, let’s say, you know, I don’t know about you, but, as a CSM, I am not spending my day-to-day, living in Zendesk.

I just don’t have the capacity to do that, so I want AI to do that for me. And so, it’s grabbed an expansion signal here and a support ticket, we’d like to add more licenses.

You know, it’s highlighting an upsell opportunity that I probably would never have seen unless the support agent pinged me on Slack. And then, the other thing that it also does is eliminating bias from those interactions that we’re having with our customers on the day-to-day.

So, we’ve all had calls that we thought went really, really great.

Only to kind of realize later that frustration was brewing. Staircase will analyze the words and the sentiment across those emails, across tickets, across calls, and it will pull out things, like, we’ve had an ongoing issue with our data sync, which may indicate, you know, some rising dissatisfaction. And then comparing these signals to broader trends across accounts.

AI can help us get ahead of those potential risks, and, you know, potentially pull out, opportunities sooner, and, really create a complete and objective customer picture. So now I am going to jump into a slide-led demo, emphasis on slide-led.

I have learned the hard way that live demos are like my toddler, always acting up when you least expect it. So, instead of, tempting fate, I’m gonna walk you through a few screenshots that will show you kind of the same flow, and not risk showing any sensitive customer data or anything like that.

So, this first view is my staircase account overview.

This is my go-to for call prep. So, I can get a really quick glimpse into the overall customer journey.

I can see that this account is trending really positively. The summary view shows me, kind of, high- level engagement metrics. And sentiment scores at a glance, and if I want to dig deeper, I can expand that customer journey section to see every engagement that’s happened over this time period.

it gives me a really fast read on the account, in just a few minutes before any meeting. So this is what I do before, any call to prep.

And so.

If I dive into that customer sentiment report, I can get a quick view of how my customer’s sentiment is trending over time.

So I can see if things are improving, if they’re declining.

This is huge, especially heading into renewal conversations or even an EBR. It helps me walk into those meetings really understanding the pulse of the customer and being… You know, fully prepared. I can also click over to see, the stakeholder tab, and where I can see the relationship heat map, which gives me a really quick view of how connected we are across an account.

So, I can see who I know, where I might be single-threaded, and who else on my team has relationships with people at that customer.

So, I know if I only have one strong contact.

you know, I know I need to work on building more relationships, or if I need to strengthen ties, I can lean on a teammate who may already have a good relationship, and I can spot gaps and opportunities before those key moments, like renewals, or expansions, or even an EBR. And my expansion analyst in Staircase, helps me spot expansion opportunities I might have missed before.

So, that kind of thing used to be totally manual.

I’d be pinging teammates on Slack, I’d be trying to figure out if a customer was ready to expand by digging through old notes, through emails. you know, now AI just does that for me. It’s constantly scanning all of those different customer communications and picking up on patterns that I’d never be able to track myself.

It’s pulling all of that together so I can see where a customer might be ready to grow, and it’s serving that to me. I love this example here, because what it’s showing is Staircase has identified that a customer’s Gainsight admin is exiting, which might be a great opportunity for services, right? Maybe we can sell them some services.

This might have been a conversation I wasn’t part of, or maybe even invited to. But now I can see all of the opportunity details, what the ask is.

It’s also summarized, right, like, how we can help, and what’s already been discussed. And it even recommends an action plan based on similar opportunities we’ve seen before.

So it’s pulling in data that it’s learned from what worked in past situations to suggest next best steps for me.

And we just looked at a few examples of augmented agents.

So these are the ones that help me do my job better and faster.

So I’m going to shift gears a little bit and talk about autonomous agents.

Which, take it a little bit further. So these are agents, as I mentioned before, they actually take action on their own without me having to jump in.

So, one example is our renewal agent.

It is fully autonomous. It’s handling renewals for unmanaged accounts, so those are accounts without a dedicated CSM. And at a high level, it’s enrolling the right customers into that renewal motion.

So it’s taking all of that customer data, and it’s making a judgment on who should be entering this program.

It’s going to set up an engagement plan. it can send communications, so whether that’s an email, or maybe generate a phone call, or maybe it sends an email first, and the customer doesn’t respond, so then it follows up with a call. And it’s doing this autonomously.

And then it might even generate quotes, or escalate to a human if needed.

So, say we don’t get a response on any of those outreaches.

And then once the renewal wraps up, it’s going to record the outcome.

it’s going to update our systems, and it’s going to learn from that process so that it gets smarter and better over time. So, this is an area that I think, just really leans into how we can help scale, especially for those unmanaged accounts that we just don’t have the human bandwidth for, and agents are perfect for that segment.

So that’s kind of it for me, but I am going to just… just run through some key takeaways, that I’ve learned on using agents and customer success.

I think the first is that data… absolutely impacts the quality of your agents.

AI is only as strong as the data you feed it, so your foundation has to be solid. You know, you don’t want to be serving your customers, what’s the term I heard last week? AI slop.

So every company I work with, you know, brings data challenges.

So if that’s you, you’re not alone.

But as we are moving, I think we’ve been hearing, right, like, this is where things are going.

It’s not going away. You know, we have to start to learn how to embrace these tools, and the first step is getting our data to a point where it’s usable by the AI. So, whether that’s cleaning up data, or getting it to the right place where you can actually start leveraging these tools, start thinking about that.

And then second, you know, use forcing functions to drive change.

For me. Adopting AI wasn’t optional, it was absolutely necessary. I came back from maternity leave last year, or 2 years ago now, and inherited a bunch of accounts overnight when another teammate departed the organization, and I needed a way to Scale, to learn about these accounts quickly, and still do right by my customers.

And that’s where AI came in.

Like, it helped me ramp quickly on these accounts and stay on top of everything. So, when big changes happen in your org, use those moments, right?

Use those to push some transfer… transformation, even if they feel uncomfortable. It’s important. And the third is.

That example that I gave earlier of that customer that has had, you know, 15 quarters of a 120% NRR, right? They started somewhere. So start experimenting.

You know, try one workflow, one play, see what happens, and then iterate. You know, I see a lot of people, even my customers, they’ll try something once, or even internally, right? Like, CSM’s here.

You know, we have a culture of experimentation, but I have seen people, you know, try something and say. this didn’t work for me, right? So iteration is so important.

It takes some finessing to really get it right. When I first started using ChatGPT, for example, my prompts weren’t great, and I got some of that AI slop. But the more I practiced and refined, the better the results got.

So you kind of learn with the AI and how to guide it, to get what you need, and now I save so much time on things that used to be super manual.

So keep experimenting.

The more you use the tools, the more you’ll get from them. And then finally, know the ROI, right?

Be super clear on what success looks like. Are you trying to save time? Are you protecting your GRR or your NRR?

Are you trying to uncover new expansion opportunities? start with those outcomes, and then measure against them. You know, there’s a lot of noise out there about the AI bubble, right?

So, the truth is, if you go in with clear goals, and you track those results, and you iterate, I cannot, you know, put enough emphasis on that, on iteration, you will start to see the results.

Julia Nimchinski:

Thank you so much for the amazing presentation, Desiree. And, we have a lot of questions, actually.

Desiree Dietiker:

Okay.

Julia Nimchinski:

community, so let’s try to hit a couple. One of them is, how do… how do agents decide when to act versus when to involve the CSM directly?

Desiree Dietiker:

So, I think that, you know, so that’s where I meant when I said, you know, giving some specific parameters.

So you build the guidelines, right?

So you build the guidelines.

So maybe it’s exception-based.

So maybe you want the AI to do the outreach first.

And try to get an answer or a response from that way.

You know, and maybe it’s a situation, like I mentioned earlier, where you’re first sending an email. Maybe the customer doesn’t respond to the email, so you generate a phone call. Maybe you don’t get a response from the phone call.

And then maybe, you know, you try again.

You, like, run that loop again, and then if you still get nothing. then maybe there’s, like, a time frame around it. You know, say, like, you know, 10 days, or 30 days, or whatever, or maybe it’s a certain amount of time before the renewal is due, and then you generate, like, a CTA or something for someone to go ahead and reach out.

So you generate, and you specify what you want that process to look like, and the AI will follow it.

Julia Nimchinski:

And in the case of Gainside, for, you know, if our, like, audiences decide to go with you as a partner, as a vendor.

How do you support, people experimenting and, you know, like, in terms of, like, the CSM side of things? Yeah, just curious to hear more. How does it work?

Desiree Dietiker:

So, you mean, like, with our product, or just in general?

Julia Nimchinski:

Yeah, just if… I mean, I know that we’re talking a lot about all of the futuristic concepts, but in the case of enterprise, we’re seeing a lot of folks just trying to, you know, integrate and starting integrating a Gentex CSM type of workflows. So, how does it work if, you know, just you… Right, just at the beginning of the curve.

Desiree Dietiker:

Yeah. And how do you support them?

Yeah, so we do have a couple customers, that are kind of co-building our agents with us right now, so for us, we’re considering this as sort of, like, think of our, our platform as, like, our Gainsight CS, right? Like, that’s our… you know, our bread and butter, you know, this is, like, what we’re known for, this is what we do, and then we’ve got a couple other products that are sort of, like, emerging products. Then we have this, like, Y Combinator type thing that we’re running with our renewal agents, and, we are co-building that with our customers.

So we are internally doing the process of experimentation with our customers, and they’re helping guide us, and we’re helping guide them.

Julia Nimchinski:

Another question here, what early metrics show that AI-driven expansion plays are working?

Desiree Dietiker:

Yeah, this is another good one. So, we are just, like, we are just right now, that expansion, analyst is in beta.

So we’re just starting to use it internally, so I hope to have some metrics around that over the next couple quarters.

But I think so far, what we can tell, like, leading indicator- wise, is it’s generating more opportunities for us, and calling out things that maybe we didn’t… we weren’t already aware of.

And so, you know, maybe some leading indicators that we’ll track are, like, CSQLs or, you know, leads generated or something like that.

But hopefully we’ll have some more solid metrics in a couple quarters.

Julia Nimchinski:

And… One technical question here.

How does Gainsight multi-agent system actually identify and qualify expansion opportunities?

Desiree Dietiker:

Yeah, so it’s… what it’s looking at… so if you… if you saw that slide earlier where I was saying… where I was showing you how, like, staircase looks under the surface, so what it’s doing is it’s… and we’re training it, right? So we’re training the model as… as we go, and we’re training it to learn as well.

And so we’re giving it specific parameters on what to look for.

And so we’re using, like, some prompting, and then we’re also designing and And putting in, right, like, specific keywords that we’re looking for.

And then the AI is basically scanning that whole layer of customer-facing communications to pull those things out.

And, what I’m finding is, as we’re continuing to build this, and as I mentioned, it’s in beta, so we’re still kind of tweaking it, and we’re providing feedback internally on, like, hey, these are opportunities that are existing, we want it to, you know, and it’s grabbing information still on, like, you know, current opportunities in flight.

That’s still good, because it’s giving me, you know, a higher level overview of conversations that might be happening outside of my purview. But we’re training it.

We’re training it as we go, and we’re continuing to tweak it, so… That’s how we’re doing it.

Julia Nimchinski:

Thank you so much, Desiree. And lastly, what’s your prediction for GTM and AI?

Or 2026.

Desiree Dietiker:

That’s a good one. I think, you know, and maybe this isn’t like a… Everyone’s probably saying something like this, but I think… it’s just gonna keep growing.

I mean, it’s just amazing what I’ve seen, even with my customers over the last year, right? When we first started releasing some different features in our CS platform. I had a lot of customers who were, like, kind of hesitant.

Like, you know, we don’t know if we want to leverage AI yet, like, what are your security parameters? And, you know, we’ve got a lot of really good, like, we’re, you know, ISO certified and all these things, but, you know, there’s still a lot of customers that had hesitation.

In the last 6 months, I have seen that completely switch. Where customers are so eager to try AI, and I think it’s because people are starting to recognize that this is the future, and, you know, there’s no stopping it at this point. And so, you may as well embrace it, and, you know, try to take advantage of the tools that we have.

Julia Nimchinski:

Fingers crossed.

Desiree Dietiker:

Yeah, we’ll see how it goes.

Julia Nimchinski:

Thank you so much again. And we are transitioning to our next session.