Leading Post-Sales Teams Through the AI Shift

[Speaker 1]
I work at Relax as a director of CX and CS operations and I'm here to tell you about our journey so far in becoming a bit more AI-native company. But first things first, some things about myself. I've been working with Relax over 10 years now.

Started as a project manager, did that for four years. Then I went to the dark side, sales teams. Worked a few years as a field pre-sales consultant, helped our sales teams to sell our solution.

Then came to senses, moved back to customer success and now I'm responsible for all of our post-implementation processes, tools, customer insights and even the AI. I'm also the game side business owner. We have the Relax community that has gone four years, five years now.

We just ended the CS platform rollout early this year and just last week finished the rollout of Staircase AI on top of the CS platform. And as you can know already from my last name and why Georgia didn't pronounce it, because it's completely unpronounceable for most of the foreigners, I am from Finland. The land of 160,000 lakes, three million saunas and the one and only Santa Claus.

And I am soon to be a dad of three boys, so I have four-year-old and six-year-old at home and soon a third one, so this will be my last business trip for a while. Now, have you ever wondered how your local grocery store gets the milk that you want when you need it? Maybe.

So yeah, that's what Relax comes in and we work with one of the biggest retailers and manufacturers around the globe, helping them to make these decisions. What to have in their stores, what products, when to order more and where to order those products. And here are some familiar logos that we're working with.

And I'm not going to bore you too much about the Relax stuff, this is just to give you a bit more context that where we come from. When I joined Relax 10 years ago, 11 years ago, we were 160 employees. Today we're 2,400 plus working around the globe, 24 different offices globally.

And our subscription revenue has grown throughout that 11 years steadily to 30% plus every year, give or take, on average. And that's pretty incredible. So there's a quite a big change happening within the company over the course of my tenure also.

One thing that we are really proud of is our NPS score. And yeah, there's the industry average also to give you a bit more context. And that is one of the key testaments for our value of customer is a friend at Relax.

The rest you can read later. This is just to give you a bit more context of the change that we have gone through. And what I'm going to tell you next might be the biggest change that we are facing next.

Now, Relax was founded about 20 years ago, back in the time when Salesforce was pioneering the SaaS motion. Most of our competitors did still on-prem model, you know, big deals, multi-year licenses, you name it. But we took a bet to go SaaS native at the very beginning, and it did pay off.

And since then, the message has been very simple, renting feeds, owning a software. And over the last 20 years, the on-prem has completely vanished, as you know. But today, as I said, I think we're facing even bigger change than going from on-prem to SaaS.

And this is something that Chuck also told in his keynote. So we could have aligned our presentations a bit better if I would have known. But if you look at the market, you see a lot of these acts-as-a-service, platforms-as-a-service.

Chuck told hire-as-a-service. I think Gainside was talking about retention-as-a-service also. These are popping up everywhere.

And why is that? Because the traditional SaaS is cracking. It's crumbling beneath our feet like the bridge that there is in this picture.

The software itself, it's not anymore the selling point, because these days you can build on top or even with only having the data, anything that you want with cloud code or whatever you are using. So the actual software, it's not the selling point anymore. And at the same time, we need to also think again, the service models that we are offering to our customers.

If we start just bolting on AI on top of our current processes, the AI-native companies that have been formed within the last three or four years, they will beat us. So we need to also become SaaS-native ourselves and think what to actually deliver. And throughout this journey, it wasn't during the first steps of our journey, it hasn't been the technology itself that has been a hurdle for us.

What really sort of struck us at the beginning or during our journey is that we really need to start reimagining the whole work, how we are doing the work, not just bolting on on top of our current processes, AI stuff. I started my journey with AI about a year, year and a half ago. I used cloud for the first time about a year ago, September 25, only a year ago.

And what we've seen is that first people used AI as an extension of Google, and we're trying to get rid of that, asking the next question, the next question, going deeper within the AI. It's very easy to use AI today to make, let's say, a meeting summary out of that. Everyone is probably using that today.

But it's really hard to start thinking or reimagining the actual work that we are supposed to be doing and what hand over to AI to handle. Now, let me show you a few steps from our journey that what we have gone through. First, it was the access, getting our hands to the LLMs, cloud, whatever you were using.

Luckily at Relax, our leadership was actually encouraging us to get access, go nuts with it. And this has actually now starting to pay off, that we are, everyone is able to do something with AI and build their own. Then something that we landed very fast after that, and this is something that you all know, that access isn't the adoption.

I've spent countless of hours working with my teams and other teams also sharing my, how I use AI and kind of creating that spark of ideas that, okay, you could use it like this and that. And it has been really, really fun to do that, even though it takes a lot of time. But access ain't adoption.

Next hurdle was obviously connecting the data. And today that's not an issue anymore. I think still beginning of this year, it was a bit of an issue, but these days everyone has an MCP.

Gainsight has an MCP in community, CS platform, staircase, all works. That's solved. But what we're still learning is that people shouldn't use LLMs as an extension of Google and really go deeper, asking the next question, understanding that why and learning at the same time.

We're also building something that SAP mentioned, or at least I think that they are referring to this one, a customer memory, a database where all the interactions with our customers live and they survive during handovers also. And without the customer memory, the AI is purely guessing. It doesn't have the context, it doesn't know the customer.

So we're building this one. And the sixth one, it's on purpose, unnamed. And this is something that I argue that this is the re-imagining of the work.

We're still taking the very first steps in this journey. So I'm not telling you that we are, we're done and we're there yet. We are really trying to figure out that ourselves too.

And if you think of all of these different steps, technology has been quite often the answer. But in the background, we need to also remember the people. The transformation starts from the people.

And we have actually seen this happening already. Most of us work with software developers. They have gone a major change within the last six to nine months.

The whole role of software developer has completely changed. They've gone through, or they have to move, level up in the abstraction. While it has been a very good, we have seen a very good productivity gains out of that move, it has been also a human loss.

The whole role has lost their identity. And my argument is that we're following, but the pace will be slower. Mainly because something that also Chuck told in his keynote, again, would have been nice to know a bit earlier that what he speaks.

It's about the customer relationship. That we are still responsible of the customer relationships. We can do that better, but we still need to do some changes.

I also argue that the implementation teams will actually go first and then us. But that's a whole nother story. The hardest part here, again, re-imagining the work.

And our products will also change. Again, what was said in the keynotes, platform as a service is, our customers, they expect to be able to build on top of the platforms. It's not anymore the software that we're delivering, but we're delivering a platform that the customers can customize themselves if they wish to do that.

And this just, again, shows about the pace, how fast these things are going today, how fast we are moving in this world of AI, and how fast this target will be reached. And to give you a bit more context about the speed and capabilities of AI, this is our CEO Mikko Kärkkäinen earlier or a couple of months ago in our R&D kickoff. And let me be clear, this isn't a roadmap announcement from Relics.

We're not going to anyone's business with this one. This is more like a statement of conviction and showcase that how much the economics of building has changed. 100% of the code these days or almost 100% is created by AI.

You can build very fast. And I will show you a few examples of that. But it basically changes the whole build versus buy or build and buy approach completely.

Also, one of our leaders. Now think what this means from our CSM work and post-implementation teams. If you can't keep up with everything that R&D is pushing, R&D has failed.

And this is in no means any criticism towards CSMs. No, no. It just shows that the manual approach, let's say with release notes or attending to every single enablement session, that cannot be anymore the plan. It's more about, again, what Chuck said, absorbing the information with AI.

And this brings us to the teams that we have at Relics working on the different or during the post-implementation motion. We have customer success, service delivery, and support. Each changing a bit differently, but the direction is the same.

I'm not going to go through what is there in the boxes because even though there are some changes within those boxes, that's not the key. The key is that all of these three teams work as one team towards the customer, sharing and gathering the one customer memory and utilizing it to make their life easier. Having the same context that the AI and the agents can leverage.

And with one focus, the customer value, showcasing that in every single step that there is. And this is an example of a bigger scale transformation that we are actually undergoing just right now. Now, a few examples about what we've actually built already.

Starting with this one, support, the obvious one, the most repetitive one, and usually having a lot of historical information data about previous tickets. Something that should be fairly easy to automate. And we have built a lot.

We have built with N8n, we're using Atlassian robo agents to do ticket triaging and categorization. We have Claude to do logging and handovers. And we're also using something called Quarks.

It's an internally built platform that we are really excited about. If you want to know more about that, happy to tell more about that. And one key here is that none of these solutions have been built from top up or they've been given by someone above, but they've been built in good coordination with our AI enablement team and Dylan Bolton here.

Could you show your hand? This guy. Good.

Thank you. So if you want to know more about these things, talk to that guy. But anyway, everything is built with the team and for the team.

And you can see some estimates of saved hours. Dylan can tell more about that too. But the key thing is that the usage, adoption, and ROI, all is tracked centrally.

And if someone is not using, or if nobody's using a tool, it will be deleted or revamped and reviewed periodically. The next one is something that our CSMs are doing quite a lot, verified outcomes. And at Relics, we have a lot of data from our customers, retailers, and manufacturers, and we have the luxury of actually calculating and easily monetizing the ROI that the customer is getting from our solution.

But it's really data analysis heavy. There is a lot of data. Previously, it might have taken 40 to 80 hours even work to create a one verified outcome out of our customers.

What we did was we created a cloud skill or cloud plugin for our CSMs, something that starts with intaking the customer KPIs and context as an input. Then it does a bit of a validation that the data is good. We have a good period against which we are comparing baseline.

And no weak data gets through that. Then we are analyzing the KPI movement. Are they going up?

Are they going down? If they are going down, we are still surfacing those because CSMs need to be aware that the customer's KPIs might not be tracking positively. We need to know that because then we can act upon it.

Then the system goes and monetizes this, calculates the ROI, calculates the payback. And finally, what it delivers. It delivers customer-ready deck that you can go with and start the discussion with a customer that are you seeing the same numbers.

It creates a one slider for our sales motion to be used as a reference. It creates an Excel audit file that every single number is tracked and we can check those. And the latest one that we, I don't know if we've added it already, but we are also pushing that verified outcome into Gainsight CS timeline because we want to find those verified outcomes.

We don't want to keep them in our SharePoint or wherever. Days or weeks have become hours with this one, a huge saving. But something that one of our CSM pointed out and they were right.

If we, with this solution, we might end up showcasing our customers the numbers that we don't necessarily understand. That is correct. But we have a strict rule that while AI or Python in this everything, it does the math, but CSM or that human needs to own the claim.

They still need to be responsible of those numbers when they are bringing that to their customers. With R&D, as I told you, they've gone through a major change over the course of a year or so. What we came up with and how do we bring them closer to the customer?

We're using our beloved community. With the community MCP, we can create trend analysis. We can surface topics of the product manager's area, what the customers are talking, what do they have on top of their mind if they have issues.

The community provides the first contact point if the product manager wants to know more about it. They go to the community and start a conversation. Here's a good example of one of our R&D leads going into the community.

There's been some discussion about whether to do this and that. He posts two options or actually three options. Then a customer says that I would lean towards option two.

Then Tero here says that okay, that's a good idea. Let's do that. Obviously, they deliver through AI a lot faster than before.

This is again a good example of how R&D can be closer to our customer. Customer is a friend that you should be building with the customer, not for them. Again, I think this was mentioned at the keynote once again.

Really, this is the key mantra for our R&D today. Now, the age-old question build versus buy. To showcase a good example of this, build and buy.

I wanted to take this example to showcase what happens if the platforms do not open up. Community platform opening up. Our PSA system, it's not been, let's say, the best one that there is.

PSA, by the way, professional services automation. What our professional services teams used. We have created around it a workload estimate tool for our projects that connects our CRM and PSA tool.

The team actually built a mini game inside of the tool just for fun of it because it's easy and to drive adoption. I've played already like 100 times that little flappy game that there is. It's really fun.

Now, StarMap is our resourcing tool to do the resourcing before the project starts. Our PSA system never supported this one. We built it.

It took a couple of months to build. Compass did about two months too. The final one is time tracking.

This is something that we delivered around three or four weeks with our team. Our people, they have hated the current PSA system and the time tracking. Nobody loves that.

We created our own that connects your outlook and pulls some information that what have you been working on and makes it easier. The fourth one is a bit of a personal one. Over the last four or five years that I've had this current position, I've been always really sort of frustrated that I don't have a good visibility into the projects that we have internally happening at Trellix.

I've tried Jira. I've tried Asana. I've tried Notion and nothing worked.

A couple of weeks ago, one of our AI PMO team member posted something that was close to what I've been dreaming of. I contacted him, gave the specs and just last week he delivered the first prototype and working version of it. We are actually starting to test it hopefully this week after we're out of here.

Again, four years of frustration delivered in a couple of weeks and less than 20 hours of work. This is the name of the game. What happens if the platforms do not open up?

We're going to build around it. This is also my invitation to GameSight that opening up that platform even more so that us customers can build on top of it. We really want to do that ourselves and we're sort of itching to get hands off with the latest and greatest things.

The final one, something that I've already referred a few times, is the customer memory. I've explained that we need to have all the context in order for the AI to work accurately. Everything depends on that the teams have a common knowledge of what is happening within the customer.

In the center of everything, we are using GameSight CES and Staircase to consolidate all of the information into one place and then again using the MCPs and other means to surface that information. We're bringing in a lot of information and data from different sources, so we're not using multiple different MCPs but trying to consolidate everything into one, even the voice of the customer from the community. Staircase AI handles our emails and transcripts and some of the incident tickets, categorizing them and doing some sentiment analysis on top of it.

This is the sensitive part when we are tapping into people's emails that we need to be really careful with. The next hurdle will be to share this insight so that we don't reveal any of the sensitive or confidential data objects like opportunities or emails. We want to keep those within the account team, but we want to share the higher context, for instance with the support.

If they are working on trying to figure out an issue with some issue, they might leverage that information, let's say if you've been using cheat sheet, that level of information, higher level of abstraction information that there is. The goal is simple, one place for all of our customer insights and something that the whole team can leverage more. For all of this to work, the foundation is the new gold, how I put it there, the transcripts.

Who would have known by the way, a year ago we were still here scribbling down our notes or in our customer meetings, now we're fighting for getting our hands to the transcript for us to do the meeting summaries. This is really interesting development that we've gone through within last 12 months, but here is the more uncomfortable part, that how can we use these transcripts without making the customer feel uncomfortable of sharing, sharing their insights, sharing their worries, etc. Because if the customer feels that everything that they say is kept, shared, analyzed, you name it, categorized, they will start using any trusted advisor approach with our customers.

The most valuable thing in our line of work is that the customers can share freely and they are not afraid of being, or we bring those things back. And I want to close now this presentation with a discussion that I had with a colleague of mine the other day. We were having lunch and we were talking AI, as we all do.

We were talking about asking the next question and we were talking about transcripts. And then he said something that really, really, really struck me. A good relationship is as much about forgetting as remembering.

And that really hit me, because not only I've spent a better part of my session here telling you about AI that never forgets customer memory, transcripts are the new gold, but he was right. Between us people, forgetting is actually a feature. That we let go, we know when to let some things go, when it's said during a bad day.

Forgetting is how we forgive. And that's my answer to the question that was in the previous slide. Customers can share freely with people that they can trust.

And overall, this transformation that we are doing, it's people first. We need to also think customers as people. And I think for both Gainsight and Relics, we have kind of a common idea there in the background, purpose and mission and values.

I think Gainsight is something that being the living proof that you can win in business while being a human first. Did I get it right? Yes.

Whereas at Relics, it is customer is a friend. And none of these things, neither one of these slogans, you can hand over to a model. So the key is to understand what to hand over to AI and what to keep ourselves.

And my argument, let the AI take the admin and we take the friendship and relationship. That's it.

[Speaker 2]
And in the spirit of forgetting, maybe you forget that I can't pronounce your last names.

[Speaker 1]
It's very easy. Idan Heimo. Idan Heimo.

[Speaker 2]
So I really, really, of the many things you said, I think what struck me is like, yes, building all that context, building all that mindset and way of building a habit in the organization will still be that human like argumenting and challenging what actually are we doing that information and how are we enabling each team to build a piece of that puzzle to help with the overall focus of like customer value. All right.

We're going to jump into question. We're going to have probably a bit less than 10 minutes. Thank you so much, team.

Right. We have the most voted one. Can you share a verified CSM example of an AI driving measurable revenue impact?

For example, revenue retained, expansion generated through AI enabled QBRs. How do you attribute the impact to AI? We have loads of questions.

[Speaker 1]
Oh, yeah.

[Speaker 2]
I'll let you tackle it.

[Speaker 1]
Oh, thanks. Gee, thanks. To be honest, this is like having a really a KPI that we have seen moving.

I think it's still about the efficiency. Are we being faster? Yes.

Do we produce better quality, but does it show in retention, for instance? Too early to tell, to be honest. I honestly believe that for only the last six months, we've had some tools to tackle this, even though we are talking about starting to use AI 12 months ago, it was still early player playing around with the tools and not really a progress.

I think for support, we can already see that trajectory, hopefully soon also with CSMs. As said in my presentation, it is a slower progress for our post-implementation teams. It's not going to happen just like that. There are bigger topics there.

[Speaker 2]
There's a lot of productivity and anecdotal good feedback that I think a lot of the teams, especially that I'm also working with, are gathering, but the revenue or business impact will come. It's where in the market, we probably need to challenge ourselves and keep discussing what's the way that you found helpful and what kind of metrics can we move and how? Yeah.

So, imagine you go back 12 months, what advice would you have given yourself when you were starting or bringing the teams into the AI journey?

[Speaker 1]
Access is not adoption. That's the first thing, and I wish that we would have done that a bit sooner. Then again, the MCPs came about six months ago, so that's not the kind of, or that was the breakthrough when we started to get access to the data, then it made a life much easier.

But to have those sessions that how do you really use AI, we would have done those earlier. Then again, you never know what Anthropic will release next month, so it is a bit of a black box. Throughout the spring, it was like, okay, that idea went out to the window and towards the next one when co-work was presented, for instance.

All of that, it's all about re-imagining the work again and again and again when something new pops up.

[Speaker 2]
Of everything you've measured, and maybe we can phrase it also you've implemented so far that you've just exposed, what was the hardest to prove that was actually working and how did you crack it? Since you did have to gather a lot of teams and probably a lot of resources and investments, so how did you manage to prove the value as you were going?

[Speaker 1]
Good question. I think Dylan might be actually a better guy to answer this from the support side, but I think for many, it was still the adoption and the resistance to start using the tools. There was always a reason that, no, it doesn't work with my customers or it doesn't work with my data.

There is always this basic human resistance with every single new tool. You can think of AI as a new tool and you need to do the transformation and all that change management. That has been one of the hardest parts and it's getting even harder when we are designing the processes once again.

I don't know if that answered the question, but still.

[Speaker 2]
Let me just come around, yes, a few more minutes. Yes. Right, the implementation of your new AI solutions, has that reduced your post-sales team headcount or has it enabled more customer interaction and value delivery or both?

[Speaker 1]
Obviously, the book of business will grow within our CSM, so we do not recruit as much, even though we still see that 30% growth in our overall ARR and subscription revenue, so that has balanced it quite a bit. The recruitment rate has stopped. As you can imagine, from being 160 10 years ago to 2,400 today-ish, it's been quite a lot of recruitment, but not anymore.

I think that's something that we are also seeing quite a lot in the news, that there are unfortunately quite a lot of layoffs.

[Speaker 3]
All right, we have two maybe last ones. How do you tackle AI pushback?

[Speaker 1]
With patience and showcasing how the system works. No, yeah. It's hand-helding, it's showcasing, it's really going almost team by team, showcasing what the system does.

I did last week six or seven enablement sessions for the Staircase AI with each country team. In our session, it was the first one and we will book the follow-ups later on. It's just being there, hands-on showcasing what is there, what is the value, why are we doing this for Staircase AI.

At the moment, it's still about the customer memory, that it survives if there are hand-offs and that we can start building the AI that really works and knows the context and design the work.

[Speaker 2]
The last one, I think it's very interesting because it's connected to this one. What about the learnings? You're doing these sessions, right?

But also, what's maybe a more programmatic or systematic way of not just uncracking how AI is working, but how do you enable it and talk the narrative on the learnings back to the organization?

[Speaker 1]
Obviously, we want to do or we have planned at least to do a lot of, again, enablement sessions, but it is also a bit of a wild, wild west at the moment. What I mentioned there, the one platform called Cortex, that is a platform where you can use all of the data that we have at Relics about ourselves, but in the future also about the customers. On top of that, you can build whatever tool you wish to build.

It's a white-coding app also that we built for internal purposes. It's been online for a couple of months and it's been still developed. There are almost 500 different apps there today that people have created.

For instance, we created an app for following up our NPS survey and the results and insights that there are. Someone even built a retailer tycoon game inside of that platform. You can build whatever you want.

You can imagine, out of sudden, we have 400 new apps and everyone's doing this and that, maybe some overlapping here and there. It's bloody difficult today. This is something that we've been discussing also with Dylan here, that we're in a place where we still want the people to explore, go nuts with the AI, but we need to start also consolidating and centrally push the actual apps that and the best practices and the workflows, reimagining the work that the people should be using.

For instance, for staircase, one of the CSM, one of Chris's CSM, sent me a message that could you share some best practices, how to utilize it in day-to-day operations and day in the life of CSM. That's the next thing that we need to start gathering and sharing those globally, those best practices. It's a bit of a balancing act between going nuts and here's your tools to use.

Proven tools, perhaps.

[Speaker 2]
I think we're out of time. I know we have more questions. I would invite, we have some brilliant minds from the Relics teams around, so I would encourage you to connect and maybe just get some answers directly from them.

Once again, thanks so much for coming and for telling us about your journey with AI. Thanks again. Let's give it up for Timo.