From Intelligence to Action: Putting Customer Context at the Center of AI Workflows

Julia Nimchinski:
We are transitioning to our next session. Welcome to the show, Elliot Halverson, Senior Product Manager from Gainsight. How are you doing, Elliot?

Elliot Hullverson:
I’m good, thank you for having me.

Julia Nimchinski:
Our pleasure, and we’ll be showcasing how you can put context at the center of workflows and boost retention.

Elliot Hullverson:
Yeah, absolutely. Well, for all those, my name is Elliot Halverson, I’m a Senior Product Manager of recently, been with Gainsight for 8 years, but majority of those 8 years, up until about 2 months ago, I was a CSM.
So… for the context line here, I’ve started as a TAM within our SMB, worked all the way to our enterprise, so I bring, hopefully a lot of… experience covering a wide swath of clientele to the role, and predominantly, it’s going to help kind of explain why, how we apply more context and intelligence to kind of an MCP future world, overall.
And as it’s, you know, as it is, like, our world is changing every day. Work is being done, how work gets done is changing, and so this is kind of just a visual of.
this new, fast-paced world of leveraging MCPs, whether it be OpenAI, Claude, Gemini, Microsoft Copilot, of the different ways companies are leveraging this capability to help break down the silos and get as much of the intelligence into the hands of their boots on the ground, if you will, whether that be sales, account management, product management, CSMs.
There’s so many great use cases here overall. And so, I think really one of the biggest pieces to take away from this is, just because you got the answer faster does not mean you know how to use it more intelligently, and that’s where a lot of the domain expertise and context comes overall.
And so, from that aspect, I kind of wanted to help highlight some reasons why, you still really need humans in the loop, to really help leverage this MCP capability to really elevate how it’s leveraged overall.
So, kind of, if you’ve been playing with it at this point in the game, everyone’s kind of had an experience where, like, they prompt something, and the output is not that great. And actually, it’s even worse, and they don’t want to do anything with it.
And so, when the human’s in the loop, or the expertise of the individual’s loop, you’re able to really drive a much better, input-to-output capability with MCP, with any AI tool, really. It’s about the understanding of the context that really helps you guys leverage the tool more effectively.
But more importantly, this actually has kind of, like, a building aspect to it, so… The more information that you can help put into the prompt, if you will, the input. the better output you get, this then allows you to then do more from an automation level, whether it be through a CSP or your CRM, wherever that is.
It allows you to elevate how that information’s leveraged in a more functional way, other than just, like, a copy and paste motion, right?
And then lastly, In kind of a… a, evolution of our use of AI, from a holistic standpoint, you know, a lot of it was you do a prompt in return, and then you have to go copy and paste and put that somewhere, or you’re in Claude or OpenAI, and it’s just there.
Like, it’s not getting into any other key repository that helps drive other key aspects of the business, and it’s kind of developing in its own little silo. And so.
the great thing now is, with the MCP access, you’re really able to help drive this information back into the areas where it needs to be captured, and then picked up by another agent, or picked up by another automation tool.
And so I think that’s a really key thing to double down on, is The ability to really accelerate how it’s used, and then deploy it across the rest of the business to reduce those data silos overall.
And one of the cool things recently is, you know, it used to be that copy-paste exercise back into whatever tool you’re using, but now, through MCP, Gainsight as an example, you can do that all from Claude or OpenAI. Or co-pilot, and actually push that back into Gainsight, so you can push things to Timeline as just an example, etc.
So again, this is really kind of one of the key areas that, the context of the individual helps elevate how AI is leveraged internally, and then drive the right going forward motions overall.
And again, something that we’ve kind of learned from this is that The initial part, where there wasn’t enough hyper, like, relevance or, key context layered into the input actually was one of the reasons why we are experiencing, barriers of adoption to our own complementary AI tooling built into Gainsight is because people were not satisfied with the outputs, the inaccuracy of it, etc.
So, again, it does go hand in hand, and again, the more context from the human perspective you can play, layer in there, the better these outputs get, and then the more, effective And streamline, everything goes from there overall. Any initial questions from anyone on the line?

Julia Nimchinski:
Question for me, Elliot. How much context is enough context? Because…

Elliot Hullverson:
Yeah. It’s actually kind of funny. As I was going into this, so I use Cloud for Work, ChatGPT for my own personal use, and I was like. define mastery for me again. Like, when do you become a master, right? And so, the definition is usually, like, around 10,000 hours of practice on a specific task.
So the reality is, it’s really hard to put, like, what does that mean? Like, how much context is good? I think the point to take away from it is, you learn how to apply more context by practicing the motions of what good looks like, right? It’s not just, do it once, it stinks, stop, and then hopefully, magically, something happens.
So I think that’s a key piece of it. The other one is, what I’ve learned just from my own learning experiences and I am no AI expert. I’m just curious, right?
Like, I’m a… I wouldn’t say I’m, like, off the bleeding edge, but, you know, it takes me a while to gain hold, but it’s about iteratively improving what you’re experiencing as you’re doing it, and thinking about your own processes that you go through every day. From a context person to lay… to lay into that skill.
that GBT, whatever that is, that’s what brings, actually, more context and helps you develop more effectively, is thinking through little things like that. Not a perfect answer, but hopefully that helps people kind of understand how to unpack what they do and put it into the AI to help them be more streamlined. Any other questions?

Julia Nimchinski:
Not yet.

Elliot Hullverson:
Cool. And so right here, again, these are just a couple skills I’ve developed for myself. And again, we, you know, we launched Gainsight MCP Access, for Staircase, early March, is that early? And we got access to it internally. So these are things that I was already thinking about that are parts of my day that I’ve been doing for 8 years manually.
And so they’re hardwired into me, but I was like, man, I would love to really have this kind of lifted off my plate. And so these are some of the initial skills I developed just for my role as a CSM. And then when I got the job as a product manager, I was like, now I have to think differently. I have to think about skills I don’t have.
I have 8 years, 10,000 plus calls of talking about CS strategy. I can bring that part to the job, but I don’t know how to be a product manager. I know, like, tidbits of it, but now I need to up-level my knowledge to fit a senior product level, title, right? And so this is the great thing.
it allows you to accelerate career transitions, it allows you to accelerate capabilities while you’re still learning, but you can do all that in a much faster approach. And so I think that’s a really important thing to think about. And so, again, the most important thing from my perspective is You don’t want to have everyone doing their own thing.
You want collective input on what that thing should be. have a centralized location where it’s developed and then rolled back out to the team, so it is reusable. It can be triggered in plain language, because everyone talks differently, everyone says things differently, and so that’s important for that aspect, too.
But more importantly, it does go back into some system of record, because if you can’t measure it, you don’t know what’s working or not working, and we’re at the square one of you know, Excel or spreadsheet proliferation, chaos everywhere. So again, these are just really tangible things to take back, think through.
I don’t want that to be, I mean, in the essence of, like, kill people’s desire to learn more and test and play. But understand that if everyone’s doing it radically different, or it’s not standardized, it does come back to kind of just mass chaos overall. And also. we don’t want it to be difficult for the individuals to gain something out of it.
I’m sure everyone here has had an experience where they go down a rabbit hole with something, via AI, ChatGPT, and, like, you get super frustrated because it keeps on breaking on you, and you just don’t know how to fix it.
And so, again, that can actually detract from the day-to-day work of an individual, and they don’t want to get stuck And that loop overall. So… Those are just some tidbits. Again, I’ll pause here if anyone else has any questions online before I go into maybe a little bit of a potentially live demo of sorts.

Julia Nimchinski:
Yeah, so we had a demobook.co, and our community is asking, how does this compare to that CSM agent, if you’re aware, Elliot.

Elliot Hullverson:
That, I do not know. I haven’t had a chance to play with Hook’s agent. So I can’t give a direct answer. Maybe if I show you, maybe you’ll get a perspective on how it works. Yeah, perfect. Awesome. Any other questions, from the audience?

Julia Nimchinski:
Not yet.

Elliot Hullverson:
Okay, whoop. So let’s go right into it. So… oh, sorry, I gotta XL out of this. So, I teed some of these up earlier, so this is my production cloud account, and just confirming everyone can see this. Cool. So, again, this is when I was a CSM, and again, just for context.
I own 40 accounts, so it was relatively high touch, considering for Gainsight terms. And so, what I wanted to do is create a more effective way for myself, to tee up what my next week of calls is gonna be.
And then some quick, different skill sets that, again, set you’re kind of constantly thinking about as, like, what is the next step with the client, or something internally that I have to do to move to the next step.
And so what I did was I created a skill that can look forward, specified timeframes, I can jump weeks ahead, it’s not locked down to a specific timetable. But for this example, I asked it to look at my Gmail calendar all next 5 days, pull in all my monthly strategic calls.
I have developed several skills to go into this, and so, ultimately, this is just an example of how an individual can use MCP. It’s not necessarily an example of how an agent, per se, uses it, so I just want to clarify there.
But in the essence of an agent, you still need a lot of contextual awareness to how to belt the agent to be like the CSM, so you have to know your processes backwards and forwards. That’s a really important piece.
The context is really knowing, like, what does this full workflow look like in order to do it at this level, and then even do it at a true agentic level. Like, you have to have a lot of deep domain expertise on certain workflows. And so, coming back to this, though, what I did is I created And again, this was… this was, like, my heftiest build.
This was about 6 hours worth of work, because I initially approached it as one shot. That wasn’t working, I seen a lot of roadblocks, and then it dawned on me. I have to break it down in different skills. So I had to create 6 different skill… sub-skills to go underneath this one master skill. But in this essence, what I did was.
I created an account brief that tells me a high-level clip of it. It gives me my objectives I’m working on from an account plan perspective. Some key talking points. That go into this next call. And this is all pulled from our staircase, our timeline entries, email, inbox.
So all three are coming together to provide the most high-level, accurate version of this overall. Then down here, I don’t want to just do copy-paste and have the conversation. I want to take action. So I created a loop, not a loop, excuse me, an action point where I can, if it’s not at risk.
I can simply say, update my account plan, and it’ll actually take the objectives from the call and update my plan internally on Gainsight to make sure those are up-to-date and ready to go. Now, I’ll come back to this in a second, because it’s one of the funny things to call out.
But under here, if I’m doing it in advance, I want to prep an agenda email. What are we going to be talking about in the future, send it out overall, right? So, I can send this to my draft inbox. Bid it out, end of the, you know, end of day Friday. A lot of times at Gainsight, you know, we do a lot of relationship building.
A lot of it is based on emails that we craft as an individual for our execs to get out, because everyone’s crazy busy. And so, again, I can have this draft go to my inbox to send to my you know, my manager, my exec, whoever that is. So, being able to take actions off the skill. Then down here, if I really wanted to.
I can sync this just all the way back into my timeline for this account. So again, this was a pretty hefty build, and I just… I wanted to push to see how far I could go with it, and this is great. But what I want to kind of zoom out and show you is, I did this this morning. it takes a lot of time to do a really quality skill.
Like, and you’ll actually hardline rules and other things, but it will forget. And so, as an individual, you need to be present and have deep understanding of, like, what it’s supposed to be doing in order to get it done right. And so, I just want to highlight that, like, this is the simple prompt. And then it does all this.
So again, it’s… it’s not just, like, a plug-and-play, like, you have to understand, from a human perspective, the context that goes into this to know if it’s right or wrong, or how to make… how to, like, know it’s not doing what it’s supposed to be doing from a skill perspective with claw. It’s just an example.
So just wanted to highlight that overall. And then down here, it’s just, again, there’s other parts where you have to know you have to run it silent, because if you don’t, each one is its own layer of information, and then you have a scrolling effect. It gets pretty significant, and that’s an aesthetic, unpleasant scenario.
Or sorry, an unpleasant aesthetic. The other one… So, when I moved into product management. Again, as a CSM, I do a lot of, hey, Mr. and Mrs. Product Manager, client has this request. Is this an enhancement? Is this a risk? Where is this on the roadmap, right? So I understood the context of.
product managers do a lot of research to assess and prioritize what to do on the product roadmap. And usually, historically, that’s been a very labor-intensive research process, or It is the first 5 clients that raise their hand to talk. They get prioritized, and so maybe it’s not the full scope of the picture.
So what I did is I built a skill that I called the voice of the Customer that, based on what the client’s looking for, I hit the prompt, it searches, again, Staircase, Gmail, all of Gainsight Timeline for every single person who logs something for X number of days back.
Because again, if you give it too much leeway, it’ll take forever to return that, and it’ll time out every single time. And what it does is, it’s a specified breakdown of which clients are asking for it, what’s the confidence level behind the ask, pulling bits and pieces from call transcripts into it so we have, like, good evidence.
We come down here, we break it down by segmentation. And again, this is… sorry, I’m gonna… this is a real-term ask, and we just released it this past week. So again, like, this is to show how it can be accelerated. And then down here, I give a projected value of GRR that this could impact at different segment levels.
And all of this is just a simple way, utilizing MCP access, context from the human, to get the maximum output to do a thing that we need to do. And again, how do we automate in the future? Those are other key aspects of it.
But, again, I thought this would be a really good example of me going from 8 years, a lot of conversations in CS, unpacking my day-to-day to have effective, automation of certain, internal things I do. But then, as I move into a new role. I have to up-level that skill set, and I have to be ahead of the curve, right?
You don’t… you don’t live in an AI world. And have, like, a 6-month ramp-up time anymore. Because AI can accelerate that to about 3 weeks, if you really need it to be.
So again, this is just a core aspect and a piece that I wanted to highlight in this changing role, but how the human in the picture can really help elevate how you use these tools To really get the best out of them, and help everyone else around you. Come back to the deck. Real quickly, any questions about the two skills overall?

Julia Nimchinski:
We have some questions here. One of them is, what customer contacts is most consistently missing from churn models?

Elliot Hullverson:
Customer context is missing from churn models. That’s a… I mean… I have maybe an idea, but that’s a pretty broad-based question, and each client has a different scenario. I think, you know, honestly, if I was going to be just brutally honest.
A lot of it’s based on… there’s just… we’ve identified it, it happened, we don’t know anything that happened in the middle. And that’s just usually the biggest pain, is we’re just actually… we… we are missing some very important details about what was the core reason for the churn.
And so you don’t have a full accurate picture of that, but that detail of progress to success of saving it, or progress to not saving it, is usually never there. And that’s a core piece of understanding and having a better term prediction.
Certainly product adoption, other things go into that, but that does help Drive more context to predictive models.

Julia Nimchinski:
Awesome.

Elliot Hullverson:
Any other questions?

Julia Nimchinski:
Nadia?

Elliot Hullverson:
Cool, awesome. So, when it comes down to how do you get started, or where do you go from here, and again. Part one is… you know, depending on the size of the team, like, I know AI tooling is not cheap, and everyone’s in a different space with how much they can allow for, but as much as possible, enable your team to play, practice, develop.
Gainsight’s certainly done that. you know, before we were doing full enterprise licensing, you know, everyone was doing their own thing, and learnings from it, etc. But when it comes to going forward, if I have some recommendations. One is, Don’t blow the ocean. Pick your highest friction workflow first.
So, whether that’s for the CS team, account managers, AEs, product managers, support individuals, everyone, pick your highest friction point, start there. Get initial feedback first, to find what the standard approach is, and then build off that. And then iterate over time. I think that’s a really important thing.
And again, kind of the nice thing is everyone’s going to have a little bit different take on what they want it to be, and it, you know, they can take it and tweak it if they want, but still have to have some lens of standardization, so it makes sense internally. The other part is, you have to know where the data lives.
That’s… like, a really core piece. And so, again, understanding where the data silos are, where the data lives, how you can pull together the MCP, build it from there, and then enhance the friction point. And then make sure it is reusable.
Again, it’s… you know, a lot of people will think you know, you just take this one thing I built over here, and I just plug and play it over here, and they sometimes forget, like, you have to actually instruct it to change things in.
So you have to maybe, like, focus on setting it for yourself, or different parameters, but we want it to be a way that everyone knows what to do so they can reuse it, without feeling like it’s breaking on them all, every single time.
So again, those are some just the key areas where I’d say, hey, if you’re wondering where to start next, I would go here. And then just, again, just build on it, build on it, build on it. And if you find someone who really has a natural talent for it. Just let them run with it. But go through some governance models.
And again, that will really help spur the buy-in, and people getting really excited about how this can help simplify different tasks for them overall. So, that’s… that’s the slide where, again, I’ll just open up any initial questions. From the audience on any key aspects?

Julia Nimchinski:
Yeah, folks are asking about your customer stories. Can you speak to maybe some of your favorite ones?

Elliot Hullverson:
Define, like, from the VOC piece, or, like, how are we defining customer stories? Oops, sorry, I was gonna get out of here real quickly. I… I don’t know if I have a favorite thus far, but it’s helped me… give more… and again, we’re talking about this right here, guys.
I just want to make sure we’re talking about the voice of the customer piece overall, sorry. Is this what we’re referencing?

Julia Nimchinski:
I guess. Okay.

Elliot Hullverson:
So… For me, it’s… this has helped accelerate The lens of impact. Right?
Again, before… So again, I can’t speak to a specific story, but what I can say is me just doing this alone has helped me actually be a lot quicker about if I’m helping to own the product investment, product risk component of Gainsight, I need to be able to show where the next move’s gonna have the biggest GRR impact.
And if I don’t have clear line of sight of that. I’m throwing darts at a dartboard. And so this has absolutely helped me accelerate that component of it, but more importantly, it gives a lot more clarity to the product managers and engineers about how they prioritize the roadmap more cleanly, overall. So I think that’s the kind of high takeaway.
Now, as I get more into the role, I’m sure I’ll have a better particular use case, where we’re very clear about the closed-loop feedback with, hey, Mr. and Mrs. Client, here’s what’s coming up on the roadmap that you submitted, and then coming back and getting their input about, like, hey, how did that really help you guys the way we think it did?
So I think that is something to come in the future, though.

Julia Nimchinski:
Another question here, are these workflows just for sales, or can we use Gateside to automate other things? For example, consolidating all existing licenses and purchase history into one report?

Elliot Hullverson:
Yeah, I mean, you can certainly do all that. tons of capabilities through automation with Gainsight. So there’s a lot you can do. You can actually, you know, I would say I want to piggyback off of automations that go back into the system and drive other actions that are truly digital or automated.
And yeah, you can certainly build out an entire report on subscription history and things of that nature. So those are definitely in the wheelhouse capabilities.

Julia Nimchinski:
Awesome.

Elliot Hullverson:
Cool. Anything else?

Julia Nimchinski:
Not yet.

Elliot Hullverson:
Perfect. Well, happy to give it a little time back to anyone.

Julia Nimchinski:
If it’s all, and the demo is over, let’s just dive in into the roadmap, Elliot. What can you share? What are you excited about? What’s coming next for Gainsight?

Elliot Hullverson:
Well, we just had a huge release, last weekend, and so a couple of the big ones for that are… sounds silly, but, like, full automated email capture. So bringing in you know, all the context of email threads in the game site, and again, that lens of information helps really drive more accurate outputs, overall, so that was really exciting.
I think the other really big one, that’s top of mind for me is, again, the Gainsight MCP access is now officially live, so if you use any LLM, OpenAI, Copilot, Claude. Gemini is their capability, but it’s through their CLI, so it may not be available. Sorry. Sorry, dogs.
So those were the really big ones, because this really opens up a layer of creativity on how you use Gainsight data in conjunction with other data from your tech stack, and really allows people to really extend what they can do in a day, or… or, sorry, reduce the time it’s taken to do other activities that normally they’re responsible for doing.
So that, to me, is a huge lift, and huge, awesome opportunity. And we’ve also done a lot of really great things around our DCS program, product line. So. And really becoming more and more, integrated with all the systems.
So I think that’s the other big piece, is we really know that the future is about driving scale, and really leaning into how our entire product set, from Staircase to CS, PX, Community, Skilljar, NorthPass in between. all work together to drive the best bang for the buck possible, out there for those who are Gainsight clients.

Julia Nimchinski:
Awesome, and can you speak to the broader functionality of Gainsight? Our community is, just executives in go-to-market, so sales, marketing, CS, RevOps. And there are so many things you cover and do, so, perhaps some of the most innovative use cases that you’ve seen on the market as of now.

Elliot Hullverson:
you know, one of the big ones to me is about how you engage with your audience. And this is taken from different angles, like, from an LMS perspective, how, like, education, like, what information are we getting from there, or what information should we be developing? Community. What is our primary target audience?
Maybe… executives is, and you’re using your community the right way, maybe you want to expand outside of that.
And then, you know, everything from the Gainsight CS side, product adoption, I think one of the key things that this allows us to do is, one of my hope… the things I would really… I want us to build is a content hub, is pulling in all of the information from all of our products.
Into an area where it helps you guys understand what content to create for your clientele that’s met at a persona level, and you guys can then choose the medium it gets pushed back out at.
So you guys are driving the highest level of engagement, you are extending your engagement to different audiences, but you’re putting the right information that’s gonna actually bring them into the… to the environment.
So I think, to me, that’s one of the best, forward-looking things that I want us to develop and really embrace, with AI, because content is disparate everywhere, and everyone knows to do things at scale, you gotta have the right information going to the right people at the right time. It’s a really hard thing to do.

Julia Nimchinski:
Definitely, and Elliot, just last question here, we have one minute. I’m curious, how would you frame Gainsight’s uniqueness? We had some questions before, comparing Gainsight’s to other tech solutions in the CS space, retention. So, yeah, in your eyes. How are you different? Why you? Why not?

Elliot Hullverson:
For, you know, the… I’ve been with Gainsight to see every single acquisition come into flight. And I would say the best part about Gainsight is the vision of what we wanted to do has always been there, even actually for, like, a lot of people are talking about outcomes now.
We’ve been talking about that since 2018, but I think the biggest part to me is we have really understood the vision of picking… acquiring complementary products that enhance the broader vision, which is how do you drive CS, customer success, at scale, and do it really well?
And to me, that is one of the clear differentiators of Gainsight from a product standpoint, is we’re the only one that encompasses everything. But we also are the only ones that really bring… I wouldn’t say only ones, but we also are the market leaders and the thought leadership, right?
So, you know, we’ve been developing this and crafting this, and we’ve seen the maturity lens in all of our CS clients in the ecosystem. But again, I think the other part is we truly bring all the expertise to the table. to help drive the product component of it.
To me, that’s the biggest differentiator, especially when the world… how things are moving so fast. Thought leadership, being human, and the context and knowledge you bring to the table is really what’s going to separate feature function From Value. That’s the name of the game.

Julia Nimchinski:
100%. Huge Gainsight fans, and we’ll share the links and all the announcements for… if I’m not mistaken, you have a conference, right? Coming up.

Elliot Hullverson:
Yes, just a little one, just a little one. So, yeah, Pulse is coming up, end of May, we’ll be in Vegas. If anyone is interested, reach out, let us know. We’d love to have you join us. It’s gonna be packed, lots of great information, not just about Gainsight, but from a broader, agnostic picture about the world we’re living in today.
You can imagine a lot of AI is going to be thrown out there, so if you have a drinking game, go for it. All those fun things, but yeah, it’s gonna be a great power-packed session, and hopefully see you there.

Julia Nimchinski:
Amazing. Thank you so much again.

Elliot Hullverson:
Awesome. Alright, thanks guys.