Beyond the AI Pilot: How Silverfort Built AI Into Everyday Customer Success

[Speaker 1]
Hi, everyone. Shani Brenner, VP CS at CIFA 4. In the next 40, 30 minutes or so, I would like to speak about a topic that I feel every CS organization struggles with.

And obviously, it's not whether to use AI. I'm sure all of us build great skills and beautiful dashboards that no one uses, right? The real question is, how do we turn these AI experiments into something the team uses on a daily basis?

So on the agenda today, I will introduce myself and where I'm coming from, why we even started this initiative, how we found where to start. I'll share some numbers and KPIs. Then I would like to walk you through a few examples.

And at the end, Kara will join me to deep dive into a specific use case. We're going to have some time for questions at the end, but if you need urgency, ask a question, feel free to do so. I will be more than happy to answer.

Good. So a bit about myself. Shani Brenner, VP CS, Ops at CIFA 4, cybersecurity company.

As an Ops, we are in charge of all the processes, cross-departments initiatives, tool stack, and also we lead the AI strategy for the entire CIFA organization. Outside of work, I play the guitar. I love photography, move to the other side so you can, love photography, good food, but I also work out at least four times a week, so I'm trying to balance it as much as I can.

I love traveling with my family. I love parties and raves. My husband keeps saying it's my midlife crisis and it's going to be over soon.

I'm not sure about that, but please don't tell him. But most importantly, I'm a wife to Doron, a mom to Lenny, Adam, and Nati. As you can see, Lenny and Adam are twins.

And yes, I had three kids under three years old for over six months. And if you were wondering how my afternoons look like, this is it. On the professional side, I'm an accountant in my profession.

I started my career at KPMG as an accountant, then I moved to Deloitte as a financial advisor. That's the point when I decided that I need to shift from the financial industry. I started at Fiverr as a CSM, moved to Checkpoint as a business operation, Google as a strategic partner manager, and today for four years at Silverfort leading the CS Ops.

Silverfort, in short, we're a cybersecurity company leading the identity security, whether it's human identity or non-human identity across every environment, on-prem or in the cloud. We also recently developed a new solution to help with AI-based attacks, so we're constantly improving and adjusting ourselves to the industry changes. All right, that was the backdrop.

Now, before I start, I wanna ask you a question. By show of hands, we're not going to use any tools, poll tools, by show of hands, where do you think CSMs spend most of their time? And I'll give you a few options and then we can vote.

First, talking or meeting customers, preparing for those meetings, or maybe firefighting after something happens. So who feels it's talking to customers, preparing, talking and meeting customers, okay. Preparing for those meetings, and fire drills, firefighting, supporting, oh wow, wow, I was not expecting that.

All right, so I asked the exact same question. Maybe our situation's really good, but I asked the exact same question, specifically at Cableforth. You probably won't be surprised to hear that it's not spending time with the customers.

Most of our CSMs spend most of their time by preparing for those meetings, surprisingly. That basically means they spend a lot of time cross-referencing between data. They spend time manual usage and health check before any renewal to understand if the renewal might be at risk or not.

And what surprised me the most was the amount of systems a single CSM can check before a single QBR. So let me show you how it looks like. At Cableforth, we love tools.

So we have Salesforce as our main CRM. We have Planet as our main CS platform. We use Staircase for sentiment and health.

We have EverAfter, Clary, Slack, Zendesk, Jira, Tableau. You got it. All of them are very, very important, but none of them is customer-facing tool.

Does anyone else feel the same pain? Good. I mean, not good, but it makes me feel a bit better.

All right, in addition to that, something even worse happened. Judgment was paying the price. CSMs spent less and less time with the customer.

They didn't provide any long-term plan. And most importantly, they were not able to spot risks in advance. And we started to see it in the numbers.

And that's the point when we understood we need a change. So we decided that whatever is related to analysis preparation planning will be owned by AI. CSM will spend most of the time with their customers, meeting customers, strategy, being the trusted advisor, everything that humans good at.

And we started with the most painful point. So how did we define what are the most painful pain point? I'm going to walk you through five steps process, but honestly, it's much more than that.

Implementing and adopting AI, it's not a one-time project. It's a change of a mindset. So you guys need to remember it.

It's more than that. What we did specifically in Silverforge, so first of all, we shadowed CSM. I sat with a few CSM to understand how their day-to-day looks like.

I wanted to understand which reports they review, what usage metrics matter the most. Then we prioritized together the most painful tasks and also the ones that provide less value. We didn't chase after the most impressive ones.

I know a lot of people are doing it. We wanted to provide value. Remember the tool slide?

We needed to map those systems. We wanted to understand which system held which data source. And the last two points are the most important one.

We created small groups of two, three CSMs and one Ops, and together we build a solution. We didn't use any top-down approach. We build it together with them and not for them.

And whatever was a success, we used it as a reusable skill, reusable project, so we don't need to start from scratch. The last part, that was let it scale. And I do want to show you how it actually scaled because I'm pretty sure this is not what you would expect.

We saw that something is working. We literally just gave it to a single CSM to try. No training, no enablement, just try it.

The first reaction, I have to say, was definitely not excitement. People started to ask questions. They wanted to understand where the data is coming from, which is fine.

I would probably act the same way, although I'm a bit of a control freak, so maybe that's not the best example. But they checked it, and they realized the data makes sense. And suddenly they trusted it.

And then the magic happened. A different team member asked for specific skills. Or a specific link to something he saw that he's not familiar with.

The managers started to present a great dashboard we built together with action items. No enablement, no PR, no Slack announcement, just CSM to CSM. Numbers.

So as of today, we have over 10 systems integrated into Cloud only for the CS. We use Cloud as our main UI, and we're very happy with that. I'm perfectly fine when CSM don't need to jump between systems.

They are centralized in one single place, UI, and everything is behind the scenes. We created over 15 skills and artifacts that are being used on a weekly basis. I'll tell you a secret, we created 100, no one uses.

Out of those 15 skills, six of them are being used on a daily basis by most of the CSMs. And most importantly, we save them time. We save them, I would say, 20 to 25% of their week. Hours per week.

Examples. All right, whatever you're going to see here, all the screenshots are real screenshots. I blurred some sensitive information and changed some numbers, but other than that, everything is real.

That's the first quote I received from one of the CSM, came to me right after our latest release, and it was like, look, Shani, I never know if my customer request was actually delivered. Whether it's a bug fix or a feature request, they just didn't know. So we created our feature release notifier.

Very simple. For every version, all the relevant customers are being grouped by the CSM. It is modular, you can change the version and it will adjust automatically.

Clicking on a specific CSM will expand, and you can see which customer asked and what was delivered versus their ask. Very easy. No digging into Jira tickets.

Yes, Amit.

[Speaker 4]
Which feature request applies to which customer? Is that part of your release process?

[Speaker 1]
Yes, and it is part of the Jira integration and Zendesk and so on. We integrate everything together. Good question, thank you.

No digging into Jira tickets anymore. And also, that can generate a suggestion email communication to the customer, and the CSM just needs to review it. So this is also maybe very important to mention.

Specifically at Silverfort, we do not fully automate customer communications. That's a strategy, like our strategy for whatever is communicated outside Silverfort. So we keep the human in the loop to review whatever is going to be sent after.

All right, the next quote, and I feel most of you are going to relate to it. I find out what we promised the customer the same moment they bring it up. Every CSM I met had this moment.

They're in the middle of a kickoff session, right? Suddenly the customer says something he never heard of. And he needs to smile, to nod, pretending like he knows what he's talking about, but he doesn't really understand where it's coming from.

So we created the sales to CS handover. This agentic workflow reviews all the closed one deals. Based on that, it generates a one pager that includes all the important information that might have missed during the handover process.

Whether it's the reason that the customer bought Silverfort. What are the most risks or growth opportunities? Are there any commitment that we should know about?

Everything is being generated into one, yeah, kind of a one pager. Approximately 48 hours after the deal closes. And we add it automatically into the internal customer slack channel as a canvas so it's visible to everyone.

Also our CEO is part of the channel so it's visible also to him. I think that's the one I like the most. I spend more time building the account plan than planning the account.

And that brings us back to the beginning of the session. CSM spend way too much time on planning and preparing for the meeting. Here you can see a CSM account planning that can serve both managers and ICs.

The idea, again, very simple. Agent review all the managers upcoming meetings from their calendar. Then before the meeting, it will generate this summary before the one-on-one with these CSMs. They can review together the risky customers. When was the last QBR? Are there any flags we should be aware of? What are the open opportunities that we should follow up on?

That's the CSM view. So drill down into specific reasons. Why is the customer at risk?

What is the status? What are the opportunities, expansion opportunities, last QBR, and so on. And most importantly, they can take action.

We recommend the next step so the CSM knows exactly what it needs to do and the manager can review it. Now I want to introduce Kara. Kara, she's in my team.

She's been at Civafort for over three years now. She leads all our tool stack. She's staircase expert and planet expert, and she's great, and she will be able to answer any question you might have, whether it's a strategy or a technical question.

So, Kara. Thank you.

[Speaker 2]
Let's see, can you hear me okay? Great. So, like Shani said, I'm Kara.

I want to lead you through a thought process and a deep dive into what we actually built, the process that went into it, and where we took it. So, this is the main pain point that I was trying to solve. We have no visibility into how many hours are actually spent with a customer.

Because we're using different tools, different teams are using different tools, when you want to look at impact and time with a customer, you don't really know how much time a customer is getting from you. You don't know how much that customer costs to serve. So, I tried to look, where does the time go, really?

Because we have this customer, right? And then we have all the teams that touch the customer. We have a sales team that's part of maybe upsells.

We have a separate renewals team, it's not led by CS. We have customer solutions, they're doing the onboarding, they're also doing professional services. They're also doing maybe escalations.

We have customer support, that's the Zendesk, and we have customer success. So, all these teams are touching our customer, and we don't know how much time is spent with them. These are all the tools that they're using, like we said before.

So, we know which teams are using which tools. Renewals work separate from CS. Where is the time going?

These are the processes involved. So, this is how we look at a customer and go, hmm, how much time are we spending with them? What's happening there?

So, lots of teams, lots of tech, how do we solve it? So, I said, great, we need visibility. Let me build a dashboard.

Let's see what our time looks like. Let's see where time is going. Good enough, right?

I'll have the ability to search by region, by CSM, by segment size, and we'll see, hey, these are where our hours are spending. Do you think that's helpful? If you got this dashboard, is that helpful to you?

Not helpful to them. Interesting, but not actionable, right? Because it's not enough to compile the data together.

What you really need is to understand what to do with it. So, after I made that dashboard from all of our data sources, the feedback I got was, great, what am I supposed to do with this? And I said, solid feedback, let's go back to the drawing board.

So, after interesting, but not actionable, we focused on moving from raw data to insights, because that's what's really important, right? So, we start with identifying, and that's what I did solve. We had all of our touch points.

We could see who was talking to our customer, and we were also making assumptions about how their time was spent, right? Because if you're taking in a Gong call that says an hour of time, it's not really just an hour of time, it's also prep time. So, we identified where the time was going, then we focused on signaling.

So, flagging accounts that were over ROI or impact. So, these are accounts that you're spending a lot of time with, but also maybe accounts who bought certain packages of time, and they're over them. So, this is signaling to you what's happening.

And the last step of this, of course, is action. So, what are you supposed to do with it? We have identified it, you've signaled that something's out, but now what are you supposed to do?

And that's, a lot of times, the missing part of the AI process, is moving from identification into action. So, this is what I did. I went back to the drawing board and I did a few things.

One, I focused on signaling specifically. So, this was for a solutions team, maybe to see who was above their identified usage amount, seeing which customers are costing us the most. But also, it gave us a high-level look.

So, you could search, maybe if you're a CSS and you just wanna see your time, you could search by this. You could search by region, by segment size. You can only look at those who are important.

So, flagged only. This allowed us really clear view into what's happening. But, more importantly, we wanted to focus on next steps.

So, this is a personalized view the CSM might see. They log in, they filter for themselves, and they say, hey, this is what I'm supposed to be doing. Review your touchpoints cadence.

Your hours are well above your peer average. The benefit of this is we also get benchmarking. We can see, hey, I spend more time than my peer, why?

And it's not a punishment, it's trying to help. So, this was really interesting for our team to be able to look at their segments and also by segment size. It makes sense that a strategic CSM would spend a ton of time with a customer, but a commercial CSM shouldn't be spending the same amount with a customer.

So, it's taking into account that as well. We also are looking at trends over time. So, this is a really big benefit of this is we can see where the time goes, signal what's abnormal, make changes, but then see how it changes over time.

Did you succeed in spending less time with a customer? And, by the way, it's not less value for a customer. It's not about spending less time because then you can take on more customers.

It's about having valuable time with your customer. And we also take into account health scoring. So, we wouldn't suggest stop spending time with a customer if they have bad health score.

So, that's just FYI. This also is a great visualization in which customers are taking the most amount of our time. And those tend to be the problematic ones as well.

So, that also is proof for us. Okay. So, all of this is great, but you can't make anything unless you measure if it's impactful or not.

So, how we measure success for this is we wanna see a reduction of flagged accounts by 25% within two quarters of rollout. So, this is a new process. So, this is how we've defined how we're gonna track success over time.

This is really good because we have visibility into reduced time. And it would be nice also to see within two quarters, within four quarters, what's changed for us. The next part is workload balance.

This is really important for managers. It is not, again, about saying, hey, CSM, you're not working well. It's about seeing which CSMs have equal work.

So, if I have five customers that are super difficult and taking a lot of my time, it might look like I don't have the same workload as my peer who has the same amount of customers, but is spending less time. So, it's a focus on workload balance in terms of difficulty as well, not just amount of customers. 30% of priority actions were completed within a month of flagging.

That's really important. Again, your signals and your actions are only helpful if you're doing them, right? So, we're trying to make sure that they're actually done.

And 60% adoption within six months. You saw from Shani's earlier slides that we had a lot of skills that were being used by 65% of our CSMs. So, 60% is a really good benchmark for us. Not every CSM likes this, and that's okay.

Okay, any questions before I continue?

[Speaker 4]
Okay. Tell me again, how did you model hours of usage per task?

[Speaker 2]
Great question. We'll save that for the end. Okay.

Impact and key takeaways. So, like I said before, visibility isn't a win, action is. So, it's not enough to add visibility.

We saw they did not like that. It wasn't enough for them. You have to provide signals and action in order for it to be useful.

And that is the number one thing, and take it one step further, not just action. You could also build the action in it itself. Trusted and unified data drivers.

This is really big. You saw our tech stack is huge. And so, when you get a signal from one system, it's not a holistic view.

The CSM doesn't know what to trust. But by combining them all together, you get very strong data. You get clear signals that you can believe in, and trust is the basis of everything.

If they don't trust the data, they're not gonna act on the data. So, providing a holistic approach is huge. Build it with them and not for them.

This is the most important thing. No one will adopt something that they didn't want, that is not helpful to them. And you saw, I built what I thought the solution was, and they said, nope, don't like that.

And by engaging them in the building process, by getting them involved, it's a much stickier thing. The other part of this is that I built this all myself. There's no R&D team here.

This was all with Claude. That's huge. So, if you wanna build a solution, you can build it alone.

You don't need other factors. You can take a CSM and say, let's sit down for an hour. Let's review your pain point.

Let's build something. You no longer need another team to help you with that. So, that's huge.

And now, we will take questions.

[Speaker 4]
It stands.

[Speaker 3]
Start reading out the questions so that we get through as many as possible. So, let's start with the first one. Did Silverford CSMs transition from firefighting to planning?

And if so, how did that journey look? What's that you mean?

[Speaker 4]
Yeah.

[Speaker 2]
Okay. I don't know if anyone can ever fully transition from firefighting to planning because firefighting is part of being a CSM sometimes. I think what we focused on is one, much better churn signalings or risk signaling for the beginning.

So, a lot of the firefighting was happening because you didn't know something was going on with your account. That happens a lot with maybe lack of visibility to tickets or escalations. So, I think that the AI building risk indicators and surfacing issues really helped with the firefighting earlier, especially when it came to churn risk.

Planning. Yeah, it's an ongoing journey, but we try to focus more on strategic planning and planning in the beginning. So, that's how we did it.

Yeah.

[Speaker 3]
Would you like to add anything on or? Perfect. Cara did that justice.

So, where did the team initially expect AI to help, but it turned out not to be the right fit?

[Speaker 1]
I can take it. But we improved it. All right.

At the beginning, I have to say we went, we decided to go with the most sophisticated thing. We didn't start easily. And we said, okay, what if we're able to spot those risks based on keywords and maybe usage that is something that sometimes we're not 100% sure if it's right or not.

And at the beginning, we created this tool skill that basically shows which upcoming renewals might be at risk. And then it turns out that it was not 100% accurate or some of them were already churned. So, this is where we actually shifted and we adjusted the prompt to basically show is exactly based on usage and on adoption of our tool.

So, we can flag only the ones that are not fully adopted that might show upcoming risk or upcoming churn.

[Speaker 3]
That makes sense. And actually, the next one is one of my favorite questions and one that I had in my head as well. How long did it take to build and deploy the examples?

[Speaker 2]
Wow, wow. It really depends on what. So, like I said, for things that you don't need an outside source to build, you can build them in a day.

And that's really the huge impact of AI. For things that are cross-functional or that require different systems, like building a Slack canvas. And that, by the way, that Slack canvas also gets sent to other systems and automatically gets logged.

Those things require other teams, so they take a little longer. So, it really depends on what you're building.

[Speaker 1]
But in addition, I would add that for every example that we showed, we did a few iterations. So, it was not just one thing. We launched the first version, we received feedback, we relaunched it, we received again feedback.

So, it was a few iterations. I can specify a specific amount of weeks and days, if you like, but it's between months to a quarter per project.

[Speaker 3]
Yeah. Perfect. Just trying to get to the next question.

So, what challenges did you face trying to get all of the systems integrated into Cloud? And there were many. So, what dependencies on other teams were blockers?

[Speaker 2]
I can take it if you want. We have a, we at Silverford are really big AI adopters. So, our company strategy and our company ethos, like really it's pro-AI, that makes a huge difference.

The blockers really are security. We're cyber security, right? So, we have to go through a security process before we're allowed to connect any MCP.

And that was the biggest blocker. Once they were available, it was just go ahead, have fun.

[Speaker 1]
Unless it keeps disconnected.

[Speaker 2]
Oh, yeah, that's another problem.

[Speaker 1]
And then you open a ticket to staircase to increase the amount of staff. Yeah. I asked them specifically to increase the amount of funding.

[Speaker 2]
Yeah. And that is the problem, it's like all these MCPs that live in Cloud, they disconnect. And if your CSM didn't check that, the data might be like, hey, something's wrong here.

So, that's the other technical part is just reconnecting.

[Speaker 3]
The next question everybody has is, for your holistic approach, how did you reconcile conflicting data from different sources?

[Speaker 2]
I can take it. So, remember I said we had all those systems and all of those fields and all of those things from different places? I actually defined to Cloud in a master document what is important, what takes precedent over something else.

So, all of those things were mapped as part of this process. So, we understand if the field from this system says this and a field from this system says that, this one has more weight.

[Speaker 1]
And also we uploaded your product that we actually didn't show, it's a good one also. We uploaded a lot of material that new CSM when he's onboarding uses and then it can map properly the data.

[Speaker 2]
So, it's also something we use already. But that is another point is, if you are looking at something like licensing and your pricing or your licensing changes, then you need to make sure that your skills are also updated. So, the best way is to make sure you're building something from a live document so that when you are updating it, it knows it and it doesn't require manual updating every time.

[Speaker 3]
Yeah, that makes sense and sounds like a long process too to do it. Where does the dashboard live and how do you make it easily accessible to the team? Do people see it as reducing tool overload or does it sometimes feel like just another tool?

[Speaker 1]
So, that's a new thing that I'm going to initiate pretty soon. We're going to have a CS hub where all the dashboards are going to live there. Currently, they are spread across different managers, team members or also in box.

But we do want to have an internal hub where everything will be centralized and mapped based on the relevant topic. And what was the rest of the question? If they feel...

The second part of the question, do you... It doesn't feel like another tool, definitely not because we have so many tools and now they just live in Cloud and maybe Slack. So, definitely not another tool.

[Speaker 3]
That makes sense. And how do you host and deploy the apps that you build?

[Speaker 2]
Well, they're not really apps. They're more like... I think it's the same question.

[Speaker 1]
Yeah.

[Speaker 2]
They live in Cloud for the most part. We can host them externally. We don't have a problem.

Like we said, we're creating a hub for that. But if we continue to grow a lot of processes there, then we might need to think of other solutions. Yeah.

Yeah.

[Speaker 3]
The next question is, what surprised you the most once these AI workflows moved from pilot into everyday use?

[Speaker 2]
I have one and I'm sure Shunyi has many. I think that the thing that surprised me the most is they love a small solution. I thought we would be building these huge AI processes because in our mind, and I think in ops, we use AI all the time and we're really engaged in it.

So when you do something really small, I had someone who was manually updating emails in Excel spreadsheet and I built them a skill to do that for them and that was wow for them. So I think that, I won't say think smaller, but impact can be smaller for people too for their day-to-day lives.

[Speaker 1]
And I was also very surprised with the engagement. But again, it's not like one thing that suddenly everyone uses. They really need to hear from their teammate that something is working.

Otherwise, they just don't trust it, which is ridiculous, but it is what it is.

[Speaker 3]
Yeah, it's sometimes the simplest thing and the old school method of, well, it travels faster than anything else. So sometimes you just have to hear it from your piece. Okay, next question that we have here on the list is, given that you already have PlanHat, which could combine all of your data sources and integrate with your Cloud instance, why did you choose to create skills in Cloud rather than use PlanHat as a single view of the truth?

[Speaker 1]
The main reason is, yes, we do have PlanHat, but we have other data sources or data indicators that literally don't live in PlanHat. So we needed a centralized place that will collect everything, and we can see in one place all the relevant information from different data sources. So it's true we have PlanHat, but it's still lacking a few things that we sometimes do not send directly to PlanHat.

[Speaker 2]
And I can add to this that AI gives us more analysis of what we're seeing. So it takes in all of the data sources. Analysis also reads the news.

So it tells you if anything's happening about the customer that's not documented somewhere, and internal Slack channels where discussions are happening about the customer. That is the number one thing that gets missed, and this takes that into account.

[Speaker 3]
Yeah, that makes sense. Okay, so we've got a couple of more minutes, so I'll try and squeeze in as many questions as possible. So what did you learn about the quality of your underlying customer data once AI started depending on it?

[Speaker 1]
Is it being recorded?

[Speaker 3]
It is being recorded.

[Speaker 1]
No, just kidding. We have- It's amazing. Perfect data.

Some of the data is great, but we have a few that we need to improve. That's, I think, relevant for every company.

[Speaker 3]
If you want the details, please feel free to see her after this session. She will give you all the details when it's not recorded. So moving swiftly on, what were those 15 plus skills widely used?

[Speaker 1]
What? Okay. You want me to list those?

Okay. First of all, we have the CS onboarding system. We have what we just showed.

We have the product indicator. This is something that is huge. Product teams came to us, and they started to ask to talk to customers to get some insights on a new feature we just released.

So we created a skill for them so they can just receive high level feedback on this specific new feature or product they just released. What else?

[Speaker 2]
The prep. We have QBR prep for the customer, but also we have one for managers or for products. So if you're joining a meeting and you have no idea what's going on with the customer, you get a much higher level view and prep for that meeting.

[Speaker 1]
We have expansion analysis and opportunities. We have, those are pretty straightforward.

[Speaker 2]
And adoption. So we look at adoption by product, but also by region, by growth, by everything. So a lot of what we're doing is also visibility and benchmarking among different segments.

So tracking how customers are adopting our tools and impact related to that.

[Speaker 3]
Perfect. A couple of last questions here. So how is the security concern regarded by your IT team when sharing sensitive data with Claude?

[Speaker 1]
Once it passed all the security requirements, it's fine. I mean, we use Claude via our SSO, and then it's under the security.

[Speaker 2]
Yeah, it's a closed system, so.

[Speaker 4]
Just one question.

[Speaker 1]
Of course.

[Speaker 4]
Anything you want to say?

[Speaker 1]
I don't think we use it, no. No, currently I'm in core.

[Speaker 3]
Okay. So that leads us to the last question, which is how did you manage to clean the data from different systems? That's a big one.

[Speaker 1]
We just did a cleanup. No, later, I mean, before. So it's maybe the two parts.

First of all, we tried to clean as much as we could. Again, there's a lot of ops involved there. So whatever we were able to clean, we did.

But other than that, it's just very easy to pull what you just need. That's the great thing on, you need to get everything. You define exactly what are the most important things you would like to pull in from which system, and those are the ones that are being pulled.

Don't pull all the data.

[Speaker 2]
Yeah, and I will go one step further and say, I built myself a personal skill. It is not for the CSMs, but it flags data issues between systems. So if I have two systems that feed each other and they're showing two different things, that means something in an integration is broken.

So I have a personal skill that also tells me when that happens. So that helps with data integrity. Perfect.

[Speaker 3]
We're gonna combine the last two questions that basically came in right now. So who owns the skills and is there a CSM owner too? And then the other one, the last one, are there any proactive alerts?

[Speaker 1]
Definitely. I mean, we have, and I encourage CSM to create their own skills also. I mean, I'm more than happy to do it together with them, but if someone wants to create their skill, they can do that and then they are the owner.

And proactive alerts, definitely. Like I think there's too many maybe proactive alerts and it becomes noisy sometimes, so we need to reduce them. But like for everything that you can imagine, alerts, risks, opportunities, growth.

[Speaker 2]
Yeah. For the CSM-owned skills, we try to make sure, like if it's really successful and they're sharing it with maybe their peers, then we try to make sure it's centralized and everyone has access to it. But we still let the CSM-owned ones go.

We have a dashboard that's named after a specific CSM. We call it her dashboard and we love it. And that's great.

It also gets them excited about AI, which is good. And one thing about alerts is, yeah, we get a lot of feedback that it's too much noise. So I would be wary about how many alerts you get and from what systems, that's another pain point.

Every system gives their own alerting. So if you're gonna use alerts, pick one.

[Speaker 3]
That's something that we face in Gainside as well. We've got fatigues with alerts. So maybe some of you are familiar with it, but that wraps us up.

Thank you so much. Sometimes it just blows my mind how much is out there and having heard it from someone who's actually built it, that I think that was very productive for us. But thank you again, Kara and Shani.

And if there are any questions after the session, please feel free to see them as we will have a half an hour break or so before we carry over the next session for AI That Works for You. Any feedback that you have for the session, please feel free to put that in the app. But again, if I can have a big round of applause for Kara and Shani.

Thank you. Thank you.