Video: Putting CS metrics to work in the age of AI | Duration: 3632s | Summary: Putting CS metrics to work in the age of AI

Transcript for "Putting CS metrics to work in the age of AI": Hello. Hello. Love it. We are about to get started. We are right at the hour. Excited to have everybody joining us today. So before we dive in and while everybody's joining in, first, thank you for joining us today. My name is Mark Iafreight, and I'm gonna be joined by Ben Newell, VP of product at Pecoboard and Intercom's own Declan Ivy, Ivory. Excuse me. Declan, VP of customer support here at Intercom. We're gonna be talking all about metrics and how to put them to work in this new age of AI First customer service. And before we dive in to all the content, I just have some quick housekeeping items for everybody. First, we are recording the webinar. We're gonna share it out tomorrow via email along with a nice kind of summary recap, all the resources we talk about, some video clips, all that good stuff. You'll notice on the right side of your screen, there's a chat area. Feel free to introduce yourself, let us know maybe where you're joining from, and if there's anything you're particularly excited to talk about today. You'll also notice there's a little q and a tab there as well. Feel free to submit any questions you have throughout the entire webinar. We're gonna have some time at the end to answer those live. And there's also a little docs tab as well, that includes some useful links and resources, especially our new, guide, the new economics of customer service. And I'm also gonna be kicking off a few polls as we go through this to kind of, get some spot checks on how everybody's feeling about things. So now that all that stuff is out of the way, would love to do just some intros to our speakers today. Ben, you're our guest. Would you like to go first? Alright. Thanks, Mark. Appreciate it. So hi, everyone. My name is Ben. I am the VP of product at a company called Geckoboard. And if you've not heard of Geckoboard, then, I mean, 1, I'm very sad. But 2, if you've not heard of Gekabord, then what we do is we are the easiest way for, support leaders to build completely custom real time dashboards that they can share with their teams to drive performance, throughout the day. Yeah. And one of our one of our great partners, awesome. Declan, onto you. Hi, Mark. Hi, everyone. Delighted to be here. So my name is Declan Ivory. I'm VP of customer support at Intercom. I've been involved in the support industry for close on 40 years now. And I think I am more excited now than any other time in my career just because of the opportunity that AI represents to really transform how support is delivered. And I'm really excited to be talking about the, topic of metrics today, Mark. So looking forward to this discussion we're gonna have. Oh, likewise. Absolutely. Such an exciting time. Just before we get into the first kind of, like, set of discussion topics, I'm gonna open up our our first poll. You'll notice you can go to the polls tab. Go ahead and cast your vote. I'll put it up on the screen right now. This is just to help us maybe frame some of the discussion we're gonna be doing. The question is how have your CS metrics evolved with the introduction of AI? So go ahead, cast your vote, let us know what you think, and as we go through, I'll come back and we'll kind of review, some of the results as they come in. So, AI metrics, operationalizing storytelling, just to kind of set the stage. So as most people know, we have metrics so we can understand and clearly explain what's happening in our different businesses. And, obviously, we have to have accurate metrics, that can help us paint that accurate picture of reality. But just as important, which we'll we'll get into a lot today, is we need to actually put those metrics to work in order to make, business decisions. And so we we know we call that phrase like operationalizing, more jargon. But with the introduction of AI, especially into the CS world, especially with AI support agents like Fin automating inbound automating inbound queries, customer service teams are evolving at a pretty insane speed and that has a lot of kind of downstream effects ranging from what metrics people track to what reps are doing day to day, and even how the CS team works with different cross functional partners. And what that means is, 1, we have to reevaluate how we measure certain metrics and, in some cases, start tracking new metrics so we can paint that accurate picture of reality. And the second thing is we have to evolve how we operationalize those metrics so we can make better decisions and tackle all these new challenges that we're that we're facing. So after kind of, like, setting the stage there, Ben, I wanna just kind of start the first question off with you. As metrics have evolved with AI, how can leaders ensure that they're accurately telling the story of their customer service performance to all the different stakeholders? Cool. Thanks, Mark. Good question. So I think the the the thing that I always come back to with it is, like, there's, like, an internal and external perspective on this. Right? So, like, externally, you gotta remember that your customers see your business, and your customers don't necessarily care. Like, even if they're talking to an AI, they're talking to a human being. So it's always important, I think, to to never lose sight of the fact that, like, the fundamentals, like, am I getting, you know, good, accurate, friendly, helpful answers in a timely manner? Like, that's always gonna kinda be there. It's always gonna be there as, like, a an an invariant, like, whatever happens with the technology. But that said, when you're thinking about what's going on internally, like, it's very, very different. And then the way that you talk internally, whether that's, you know, upwards to to other members of leadership team, whether that's outward to the rest of your employees, you've really got to think about these things almost as it's very separate. It's almost kind of slightly misleading to think of it all as just, you know, under the banner of, like, your CS team or or something like that. Because actually, you know, AI and then, you know, these bots like Fin, they're much more like machines that you kind of come back to and and optimize on a on a kind of cadence. You know, a bit like, you know, if you if you're working at a product company and you have a product, you know, you you set things live and then you evaluate them and and you come back and then you iterate and make tweaks. I think AI and bots are much more like that. Like, whereas when it comes to, you know, the the human side of things, it's not. It it it is. Yes. There there are performance metrics that perform the standards, but actually, like, your ability to impact that in real time as things happen is is much greater because of the kind of human element to it. So I think when you're trying to tell that story, like, you wanna kind of pull them apart and say, well, look, what what's our AI for? Like, I think largely today, who knows where it'll go in the future, but largely today, we're trying to say, like, this is more about efficiency than it is about depth. And that's that's changing over time. I know like that that's increasingly not being the case. But I think for a lot of people certainly early in their journey, they're gonna start much more at the the efficiency end of the curve and going like, how can we take the simplest queries and just, you know, get them, you know, good enough answers or even better answers faster than we could by putting them through people, and that's gonna save us time and energy internally. And, you know, at the at the same time, then you say, well, what does that mean for for the people remain who who are left picking up the tickets that are left? Well, actually, if anything, they're gonna now get harder stuff. So especially if you're telling a story that, you know, let's say you've been reporting your CS metrics for years in, you know, all hands or kickoffs or whatever. All of a sudden, some of them are gonna might actually look worse as you put AI in because things like first reply time, things like even like CSAT, you know, like, if the tickets that get through to the human beings are actually the hard stuff that maybe you can't do anything about, like, people aren't that happy when you tell them, like, I can't fix your problem. Right? So potentially, actually, some things will get worse. And I think there's a story to tell there that's like, actually, the baseline is what's moving. Right? Like, it's no. No. Things haven't obviously got worse. And I think that that zoomed out perspective of, like, what's the whole thing doing? Looking at those things around, like, well, how many people are we responding to, and are they generally, on average, still pretty happy with how they're doing it? As long as those things kinda don't change or get better, you can explain why. You can say, well, look, the reason then that, you know, full resolution time is going up for our human, support tickets is because I do it I'm doing all the hard stuff. Like, you know, if you're doing a hard thing, it takes a long time to reply to it. It takes a long time to more back and forth with the customer. You're gonna have you know, you're not gonna solve as many on on the, first reply as you were before if you've gotta do a lot of, you know, complex back and forth diagnosing. So I think that there's a really important story to tell about, like, the roles that these two aspects of your team play, so that everyone, like, you know, has a has the right understanding. Because I think I think otherwise, it's very easy to be dismissive and sort of say, oh, well, you know, now people are slacking off because the AI is doing everything. It's like, no. No. They're not. People are just doing the hardest stuff. And then when it comes to AI, it's like, well, you know, you know, this is what they are doing for us. Well, you know, we need to show, especially internally, like, you know, to to to to some people internally, like, this is worth the investment. Like, you know, these things have a cost associated with them, both in terms of, like, you know, the product cost, but also time to set these things up. Like, you know, you don't just turn it on and go live. Like, there is an investment in the quality of your, content, you know, putting to your public health centers, like, content you've specifically written for these things. And then that ongoing piece around, like, you know, I've got this AI now, and I'm looking at it, say, weekly or monthly. I'm going, well, you know, how many conversations has it got involved in, and how many of them did actually do a good job of, and and so on and so forth. Then what am I gonna do? Well, potentially, I'll go write new articles. Potentially, I've got to go and, like, tweak the content, like, all that kind of stuff. So splitting is is key when it comes to the internal story, but combining is something we must never lose sight of because customers, they they just see your business. It's probably a lot roundabout way of saying it. No. That's that's great. No. I completely agree. And it's, we've seen this. And, Declan, I have a follow-up on that for you, but it kind of comes down to having to almost reframe the the the the kind of, like, background context to to internal teams, but not losing sight of what actually matters, which is a a stellar customer experience for your customers. Declan, yeah, let's do a quick follow-up on that. So how has Intercom specifically adapted our approach to measuring CS metrics to give that holistic view? And especially through that that lens Ben was talking about kind of like the the human and AI, elements. Yeah. I mean, fully agree, Ben, you gotta really think about the, you know, the AI agent, experience look at them from different lenses to some extent, but not losing sight of the fact of a more customer perspective. You're thinking about the experience and you're gonna be able to to measure that customer experience as well. So when it comes to an AI agent, like there's a whole set of new metrics that you need to think about and have as part of your kind of metrics portfolio. I kinda cover them at a high level. So so for us, the key things that we focus on when it comes to the AI agent performance or the VIN performance for us, It's really, first of all, around involvement rate. So what percentage of your conversations or customer interactions is the AI agent actually involved in? You know, because again, in in terms of looking at the the business impact of your investment in AI, this is a a key parameter, like, what what percentage of your workload are you able to apply the the agent to? And again, people will make different decisions around how and when they want to deploy an AI agent. They may choose to, deploy for a subset of their their customer base or for a subset of their kind of support transactions, And that obviously influences then the, you know, the involvement rate. But obviously, the aspiration is you're trying to drive that involvement rate as high as possible. You know? And particularly the the the Ben's point as this technology kinda mature, the new capability capability comes to the 4 that opens up more opportunities for for an increasing involvement rate. And just as kind of a, you know, an example, we were pretty lucky. We were able to take advantage of Fin earlier than a lot of our own customers because we're internal. So from that point of view, we we've had kinda at this stage nearly 2 years of experience of of using Fin, and we've been able to drive the involvement rate quite high. So we're currently hitting about 85, 86% involvement rate, in terms of our customer interactions. So that's kind of been one of the key metrics that that we've tracked over time, and we don't wanna drive it higher. So for example, there's new capabilities being delivered in FIN, stuff like FIN Guidance, where you can, you know, apply company policy and guardrails. And that will open up more opportunities for us to to to apply Fin and and to drive the involvement rate. The next key metric then is, okay, Fin is involved in the conversation with the customer. What percentage of those conversations are being resolved by Fin, I e it's providing a high quality immediate answer to customers. So resolution rate is another thing that that we've had to track. You know, again, something we didn't track prior to, you know, AI agents technology. But we've kinda tracked resolution rate. We've been very clinical around understanding, you know, how we can drive it higher. We appreciate the importance of knowledge and developing knowledge and making sure that within the the content that Fin has access to, you're enabling Fin to provide us comprehensive, an answer as possible for your customers that drives the resolution rate up. And again, we've kind of gone from when we initially launched Fin, we had about 27% resolution rate to, you know, last month we hit over 60% resolution rate. So again, we've been able to to drive that quite high over time. So 2 of the the key metrics that we look at it in terms of the scope of impact of an AI agent and how much of your workload it's actually taking account of and and, resolving for your customers. But we do look critically at CSAT for, for the AI agent as well. Well, starting, we've always looked at CSAT from a human perspective. We've made a very conscious decision. We need to also understand Fin's performance. Right? You know, At one level, it's just another teammate. We should be concerned about the customer satisfaction of of interacting with Finn. And we measure, CSAT for for Finn just as we measure it for the human support interactions as well. And again, it's ensuring that your customers have the opportunity to provide feedback and actually rate their their experience. So tracking CSAT for the part of the, experience is really important. But it's also, and this is where AI is your friend as well. It's also important to look at the overall customer experience. So in our case, we have built what we call a new customer experience score, where we use AI to actually measure the customer experience for every single interaction that we have with a customer. So as an example, we can use AI to to generate a CSAT even if the customer hasn't given you a CSAT rating. So, yes, we still rely on on CSAT ratings that customers give us, but we'd like to be able to rate the conversations where they haven't given us, a survey response. And And in that case, we can actually generate an inferred CSAT using AI. And building up things like inferred CSAT, we also use AI to generate what we call an inferred quality of resolution. We look at things like first response. Like, we combine a whole lot of parameters then into what we call the customer experience score. That's that's been kinda, you know, where we've defined kinda new metrics for the business. But I wanna talk about some of the traditional metrics quickly as well. Like, you have to think differently and Ben has touched on this as well. A good example is average handle time. You know, a lot of support organizations have this kind of desire to always drive down our average handle time and improve efficiency, etcetera. Reality, as Ben said, if you're getting more complex work into human support team, reality is the average handle time is probably increasing in this new world of AI first or an AI agents, world. So you gotta take account to that as well. You get some of these metrics are gonna be counterintuitive to actually go going the wrong way in your mind. But it's actually a good thing. You know, your team are are getting to spend more time with your customers, working on more complex, more nuanced issues, and you actually want them to spend that time with your customers and add more value in terms of that interaction. So the fact that average handle time is going up is actually a positive in terms of the customer experience in those counterintuitive to the traditional way that we've looked at average handle time. Another interesting one which a customer actually raised with me a couple of weeks ago, first contact resolution. Again, there's kind of idea, well, you wanna drive 1st contact resolution as high as possible. Right? This customer is actually you'll said to me, no. Actually, I want I want to see 1st contact resolution get lower for human support. Because if a human is able to provide a first contact resolution, then the question should be, why wasn't the AI agent able to answer that question if it was kinda that simple answer? Right? And that that's a good kind of benchmark. Like, if someone can answer the question and go, is that something that, you know, the agent should have been able to handle? So is there a content gap? Is there something missing that's not allowing Fin or a any agent to answer that question for for your customer? So this customer of ours was actually saying, I wanna drive, first contact right down. I wanna see it actually decrease because then it means I am getting the complex work into the human support team. They are getting the work that's adding most value from the customer point of view. So that's just a snapshot. We talk for the full hour mark on on on kinda these metrics, but that's a snapshot of new metrics that we've kinda focused on when, you think about using AI, and also how we had to think differently about some of the traditional metrics, that support teams would normally look at. Yeah. No. That's that's fan fantastic. It's all great, and, I'm gonna put the the poll results up. And I just realized I I put 2 of the same answers on there, so that that's on me. But, a lot of folks were were kind of saying that they're still figuring out how AI is impacting those metrics. And we just hit on, I think, a lot of really good insights there. So part 1, what Ben was talking about, kind of thinking about it through this lens of, like, the the the AI and human, Declan, you touched on some of the the the new metrics you talked about, you know, FIN or AI involvement, AI agent involvement rates, the resolution rates, the CSAT, talking about some new metrics of kind of combining metrics to figure out the quality of those responses. But then we we again the I love hearing about how, some metrics are getting turned on their heads. Right? So, like, you know, the first contact resolution rate, response times, like, those going up aren't always a bad thing. So I think that paints a really good picture of not only the new metrics that we should be thinking about, but also how it's impacting kind of some of these legacy metrics that we've that we've had there. I'm about to move into kind of some of the discussion on operationalizing, so I'm gonna close that first poll. I'm gonna open up the second one. This one will, you can go ahead and start to vote on it now. This one is getting a bit more into kind of like that operationalizing side. So what is the biggest challenge in kind of putting those metrics to work across all those different teams? So feel free to respond to that one in the poll tab, and, I'm gonna kind of shift gears a little bit. So, Ben, I know this is kind of like core to a lot of what GECO board does. It's kind of what a lot of you live and breathe. When we talk about operationalizing metrics through the lens of both AI and human, you know, putting those to work to solve business challenges, maybe what are some of the most common mistakes you've seen companies make or maybe some of the best practices when actually operationalizing metrics and putting them to work for the business? Yeah. Absolutely. I mean, I think a lot of it comes down to, like, the cadence with which you're, you know, checking the metrics, the audience, and then the curation. So I think, you know, the the common mistake, if we take human support to start with, is that the cadence is too slow. So metrics are, you know, they're shared weekly or they're shared monthly, and they're looking back at what happened. The audience is too limited. It's often, like, just the manager or the managers or the leaders. It's not the the actual people doing the work. And the curation is is often quite poor because, you know, often it's kind of a kitchen sink of here's everything. Here's here's an absolute ton of metrics. And so, you know, what we see, work really well, and and I guess, you know, what we've largely built a lot of our ideas around is kinda going the other way. It's it's saying, well, actually, for for support, and and some of the domains that have, like, a a similar real time, you know, you can take action throughout the day kind of nature to them. Where is the team involved? And where actually you can think quite hard about, like, you know, what do I care more about at this point in time? Do I care more about, you know, fast replies or more care more about actually depth of quality? I'm willing to trade that off a little bit. Do I care a little bit more about, like, yeah, driving up first first contact resolution or tracking down personal care resolution depending on what you're trying to do. But but try you know, force people to to think about maybe a more limited set of metrics. Because the fewer metrics you have, the more agency you have to, to change them. And the easier it is as well for a team to understand what actually matters. If you have 32 different metrics, then how does someone looking at that go like, oh, yeah. Well, we care more about this than that. And so curating and then sort of saying, yes. Of course, as a leader and a manager, it's on you to always have a bit of an eye on everything going on. But really, when it comes to a team, can you pluck out and make clear to the team, like, these are all the metrics we care about the most? And then to actually make that work, to actually put that into practice and to drive the results, you've gotta get that stuff in front of that team in real time. Because the magic, I think, of this human side of things is that people can take ownership of that. Like, making something a metric and a goal and a shared team, you know, aspiration lets the team self organize around that. So instead of coming to the end of the month and looking back and going, oh, wait. No. Things built up really bad here. And, like, you know, we were we were, dropping dropping things left, right, and center back in, you know, 2 weeks ago or whatever. If the team have all this stuff right in front of them, like, if they can see the, you know, maybe doing, you know, multichannel support and the phone queue is backing up and up and up and up. If a team can see that in real time, then they can themselves decide, like, oh, maybe I should, like, stop answering these emails. They can wait a little bit, and I should jump on the phones as well and do that. Of course, like, you've gotta put the structures in place and, like, you know, you gotta give people the autonomy and the freedom so they know they can do that. But if you can shift to that way of working, like, I mean, as a leader and a manager, it takes a lot of your play in terms of, like, going around and micromanaging people and saying like, oh, you do that, you do that, you do that. Because, actually, the team can just look at this stuff and go, oh, this is what's important. This is what I need to do, and this is how we are gonna arrive our collective goal by doing that. And that's where, you know, tools like Geckoboard, plug number 1, come in because it's really easy to just make that curated view of exactly what it is that you care about, updates in real time, easy to share with the team, you know, whether you're remote, whether they're in an office, and you're broadcasting stuff on TV. So get cadence, audience, and curation. So real time, the whole team, and a fixed set of metrics. Like, that's super, super key when it comes to human support. And then when it comes to, like, you know, operationalizing your AI metrics, it is actually completely different. I think I think it probably is very much the other way because, you know, I mean, not yet. I mean, maybe maybe next week or a year from now, it would be different. And actually, if we, you know, give the AI its metrics in real time, it will, you know, respond accordingly. But but at least for now, it's more the case that you're gonna say, well, look, we've we've we've done some setup or the AI is in its current state with its, you know, the the training we've given it, the guidelines we've put in place, the content we provided. We're gonna set that live over a certain period, and that's gonna be based on probably, like, the volume of conversations that you're getting as to how long that period is. You know, if you're super high volume, you can probably do this every week, every day if you're really high volume, or maybe it's gonna be monthly or quarterly depending on that. And you're gonna set it live, and then you're gonna kinda leave it, I would have said, and and let it do its thing because any 1 or 2 conversations aren't really gonna sway your opinion of of how it's doing. You need to see a bigger trend. You need to this it's a statistical problem. Right? It's not a sort of, you know, one conversation at a time because your ability to steer it one conversation at a time is a lot lower. Like, you're trying to steer it statistically. You're trying to steer it at a much higher level. So you gotta let that come through and then look at that in-depth. And then, you know, I think Declan was raising these points earlier, but, like, do we have a content gap? Like, do we need to add some additional cart rails? Do we need to, tailor where it is and isn't getting involved? Like, all those kind of things. You've gotta get pretty deep at you at that point, and that probably is gonna be just the manager or or the person you've delegated responsible for that because it's going deep. Right? It's not a surface glance at a few metrics, look at it, and know what to do, and move on. You've got it. You've got a yeah. The metrics are gonna guide you, but you're gonna know whether the metrics are trending in the right direction, whether they're meeting the internal benchmarks that you have. But then you've gotta go and do the work to figure out what it is you've gotta do about that. And if they're not meeting there or if you're not trending in the right direction, why is that? What actions are you gonna take? And so you actually need to, yeah, almost back off in a way. You know, I think so sometimes we see this. People want to put the AI metrics on their dashboards in real time and say, okay. Fine. We can do that. But what are you actually doing with that? It's probably not gonna be super, you know, operational for you throughout the day other than maybe interest. But if you're really trying to make it better, yeah, wait, evaluate, go deep, make your next hypothesis, figure out what it is you think needs to change, make those changes, and then repeat again. And really the limiting factor there is just know, the amount of conversations that you're getting through, in that time in order to, you know, update your assessment of how things are going. Yeah. That that's, like, fantastic. We've we've heard a lot of that from customers when they when they've been talking about the the kind of bifurcation of of human human, kind of, like, authorization and then AI as well. And that that, like, iterative, motion of of, like, working on AI in these kind of, like, set periods of time to, like, set plans to how we're gonna change things. It's very aligned very well to, like, what we've heard from people, like, in in the field doing this, and, I wanna come back to something around that. I wanna talk a little bit more about feedback loops in a second. But before we go into that, Declan, to you, would love to hear how kind of the integration of AI into our support strategy has really influenced the collaboration with with other teams, teams like customer success, sales, and even r and d, because it has a lot of implications. And so have the the metrics or the meeting cadences kind of evolved to reflect this new reality? Oh, I think you're on you're on mute, Declan. A rookie mistake. Sorry about that, Mark. No worries. No. Just a re really good question. I think when it comes to, you know, what's the impact of, an AI strategy and how does it influence collaboration? I think it's broken down a lot of silos across teams. And I wanna talk about a couple of aspects of this here at Intercom. You know, so typically as part of a customer journey, there are multiple touch points with any organization. And from a customer perspective, they tend to think that most of these touch points as some form of customer service or customer support, even those different teams delivering it. So here they come, we we're very much committed to providing what we call an AI first experience, where it's practical across the entire customer journey. You know, therefore, we are collaborating more actively with other teams such as sales, success, and billing. And it's really to ensure that we're leveraging AI to provide immediate and high quality responses to customers where it makes sense. And we're also working to ensure that if AI can't answer a question or an issue for a customer, that the customer is immediately routed to the correct team to handle the question. I the team with the right skills and the right background to handle that question. So we put a lot of thought into how we design that customer journey. So for example, we've established what we call an automated and practice support team within the customer support organization, but it's this team that is regularly collaborating with other parts of the business that to ensure that we do think about the end to end customer experience when an AI agent is involved. For example, we have a conversation designer as part of that team, who drives much of this collaboration and is really thinking about things from a customer experience lens. Now it comes to r and d functions or our product, development functions. There are also ways that we can collaborate differently in kind of an AI world. First is around what we call new product introduction. So we have enhanced our new product introduction process so that we're prioritizing the development of content for Fin. And this ensures that for every launch that we have, we have sufficiently high quality content that we drive a minimum resolution rate of 50% right out of the box when a new product is launched. And if you're thinking about that's the time when you're gonna receive the most questions from customers, it's hard to anticipate what those questions might be ahead of time. So it's quite a aggressive target for us to hit. Like you'll I developed the content to a level that we can resolve 50% of queries for new product and capability. And that's where, required a lot of collaboration with r and d teams. We've had to get them to think slightly differently about the type of content they provide to us because we gotta think about how we'll thin consume and how will an AI engine actually make use of that content. That's been an interesting aspect of kinda, deploying AI across the the business. Another really interesting example is our feature requests. So for example, we're now making Fin aware of existing top feature requests for each of our kind of product areas. And that means that Fin can quickly identify the feature request for customer already exists. And if it does, then they can actually prompt the customer to go to our community platform and and upvote us, the feature request. And again, that's giving more tangible information to our r and d team around what's important from a customer perspective when it comes to new features, helps influence the the roadmap. And we're trying to automate some of that going forward. So apart from just, you know, asking the customer to go to community platform, we're gonna look to have the, AI agent actually make that upvote on behalf of the customer if that's what the customer wants. And then AI insights, like, we recognize that support needs to be a viable voice of the customer, particularly back to the r and d teams. I mentioned already that we developed a new customer experience score that's using AI to generate a CSAT score and a quality of resolution score for every customer direction. And I say we combine these other attributes for an overall customer experience. But we look at that customer experience through, you know, customer account level, but we also look at it at a topic level or product level. And we can feed that information back to the r and d teams, and it gives them far more actionable insights into, you know, what what is really influencing the customer experience, in terms of of how they interact with their part of the product or or their topic area. So there's just some areas of where, you know, going down this AI first strategy has forced us to work differently with the rest of the organization. And it's been really interesting to see that collaboration evolve over the last 2 years. And I'd say many silos being broken down. We're all aligned and focused on what is the customer experience that we're driving and how are we using some of the insights that we're gaining through using AI, to to really focus on the right things to improve. Yeah. No. That that is all fantastic. I love even being at Intercom. It's great hearing, like, kind of how all those different teams, like, interoperate kind of, getting into some of those feedback loops, and that's kind of a perfect segue. Ben, kinda going back back to you. I mean, I know Intercom, we do it our way. We have every kind of company is a little bit different, but there's always kind of cross functional teams that are that are involved with this. And so question to you around maybe part 1, like, how do you make sure that different teams have access to those right metrics? And then the the second part maybe being, what are some of the other ways or best practices you've seen companies set up, quality feedback loops? Not not I shouldn't say not about CSAT quality, but just generally good feedback loops to make sure that people can iteratively improve on both AI performance and human outcomes. Yeah. Good good one. I mean, this is a good chance for plug number 2, for the wonderful tool that is Gekko. But, yeah, I think you need often one of 2 ways. Right? The metrics are in your source tool. So, you know, hopefully, for most people watching this, that's Intercom as your support tool of choice. And either you need your cross functional partners to be, you know, coming into that tool, to have logins to that tool, which which can work. The other one is that you can use Twerk Echo board to, you know, surface the key ones that they need to see and maybe even have multiple dashboards for different teams that you wanna share that stuff with. But there's always actually, like, you know, the the the classic one, which is, you know, how often do these people need to see this thing? Is this something they need all the time? Do they need ad hoc access to this? Or is this actually something that as a manager, you can just, like, pull out and share with them on a fixed kind of cadence and say, like, you know, hey. We're getting together to talk about, you know, x, so let me bring Symmetrix to that meeting. I think, you know, depending depending on what you're doing, how frequently you're doing stuff, like, you can kinda choose your own adventure, from from within those paths. But, you know, anything that kinda pushes towards, yeah, more data transparency, you know, paired with understanding. Like, there's there's always this, like, tricky balance between, you know, you sort of want to share everything because, like, transparency is good and sharing is good, but then also, like, like, nuance is important. And, and sometimes you share everything and then people, like, draw their own conclusions and all that kind of stuff. So I think, you know, you you do wanna strike that right balance between letting people have access to the metrics, being kind of open, but also making sure that people aren't gonna, you know, go off on the wrong, you know, drawing the wrong kind of conclusions. And I think, you know, often that's just when people come in, like, really well just to to sit down and go through stuff together or or, you know, walk through things kind of, like, in a meeting. And it's kinda come to the, you know, the second part, like, around, like, effective feedback loops. So I think I think you're gonna have many on you. So, I guess I mentioned earlier, like, you know, we think at Geckapod, like, we think that that a really good feedback loop when it comes to metrics and and people is live. So, you know, especially for support teams, you can change what's happening in the day. Like, if a problem is is arising in that day and you can see that problem arising, you can do something about it if you have the, you know, the autonomy to do that. And so I think that feedback loop is is invaluable. It doesn't replace the more traditional, like, you know, weekly reviews or monthly reviews. Like, you will or, you know, you need to at some point look at the broader trend and make bigger decisions. That's where you're gonna make bigger decisions perhaps as a leader around, like, you know, how you allocate people's time or how you're allocating things between channels or the relative priority in various metrics or, you know, even how it's interplaying with your company strategy. Like, maybe you've been very focused on efficiency for a while, and actually now you're gonna go focusing on quality, which, you know, I think we touched on earlier. But, you know, as as AI takes more and more of the easier tickets and it leaves the harder things, it leaves more opportunities to talk to customers. Like, can you seize those opportunities that you have when you're talking to your customers to not just, you know, give them what they need and leave them having a good support experience, but to also actually, like, go further and say, well, hey. We're talking to you right now. Like, did you know x, y, and z? Or, like, you know, in helping you do this, we noticed, you know, this thing in in your account, you could probably be doing this better, or you could be adopting this feature that would really help you, blah blah blah. So almost going on to kind of, like, I guess, onboarding or adoption or kind of success type things, off the back of that. So you have yeah. You have that that still have that bigger strategic cycle review, but that really should be for your for your big changes. And then your your sort of real time day to day one is is more around having good feedback loops to teams, in particular. And then for AI, it is who who who is on that that team? Declan touched on the the different, I guess, like, teams and and some that you've created at Intercom around, you know, having, you know, having Finn do a pretty sizable portion of, of of your amount stuff. So and and the skills there, some overlap with your traditional support roles or success roles, and some are a bit different. Like, I suspect that not many teams listening have a conversation designer, in in their team. Maybe they need to. So starting to think about, yeah, who's gonna be involved in this? And maybe, to start with, it is, yeah, like some folks in products, some folks in sales, some folks in support. Like, it can be even other folks in leadership. But who is involved in in, you know, driving up the metrics associated with AI? Because, yeah, potentially, the things you need to change in order to achieve your goals there are gonna affect, like, your sales process, or they are gonna affect, like, the product itself, or or things like that in your website, like, depending on how you're deploying this. So get the right people involved. Establish your cadence, which is, for me, I think is based on it's based on having enough volume to to say whether or not the things you did last time have actually changed or not. So fix your cadence based on your volume. Get your right people, bring them together, have a discussion, go deep, and then, come up with your plan and execute. Repeat. Yeah. No. I I love that. You you touched on something towards the end there too. Like, I think one thing people should be keeping in mind is the the the traditional feedback loops, like, you should be thinking about how to evolve those, but you mentioned, like, there's new roles. There there might need to be, like, a new team or a new person or a new role that kind of gets brought into that feedback loop depending on whether you're looking at, like optimizing your AI, involvement rate or your resolution rate or if you're working more on the the team side. I think it's another important thing for people to keep in mind is that, you wanna know what you don't know. So you might have to be introducing, like, a a new role or a new team or or new people into that into that feedback loop, to make sure you're actually able to continuously improve in the right ways depending on what you're actually going after. Great. Declan, like, back to you kind of piggybacking on that. I know, at Intercom, we have a pretty strong emphasis on actionable insights. And so how do you and the CS Oregon share metrics aren't just tracked, but, kind of, like, operationalized in that way that's driving AI performance along with the human team development? Yeah. A really interesting area. So so here at INCOM, we've established what we call an, a support optimization function or team. And this covers a number of areas that includes support enablement, quality assurance, and process management. But probably the biggest thing that this team is doing is not just tracking metrics, you know, that we've previously talked about, but also building what we call a continuous improvement flywheel mechanism. And that's effectively, you know, enabling us to build a list of prioritize improvements that we can drive to improve the overall customer experience. And again, we, you know, we are moving to use AI to drive this in that we are using insights from, for example, the new customer experience score that I that I mentioned. We're also building up insights from ongoing demand analysis, like what's the work that's coming into support team. And also we're beginning to improve our QA processes. You know, for so for example, we're now doing QA for FAME conversations as well as for human conversations. We're also trialing a function called auto QA that will enable us to QA up to a 100% of our our customer interactions. And these kind of 3 core sources, of kind of insights are being fed into kind of a single continuous improvement, prioritize list of activities so that we can really ensure that we're working on the right things on a continuous basis to drive improvements in customer experience. And though there is can include product improvement, they could be kind of skills or knowledge gaps across the the support team. It could be kind of process improvements that have been identified. You know, so it's really, really important that you know, you look at how you can scale that, you know, and and AI is your friend in this regard. And particularly when it comes to the the fin QA process, we really need to identify opportunities for improving fin. And and Ben touched on some of this, like, we wanna make sure that we are driving, you you know, the quality resolution, the involvement rate, and and sometimes you need insights to to drive that. So this has been a key part of our focus, Mark, in ensuring that we have this ability to really identify the right things for us to prioritize from a continuous improvement point of view. And again, as I say, AI is underpinning or or supporting what what we're doing in the space. But I didn't even touch on one thing. I really like Ben's point around trying to operationalize these metrics in a way that they're you're almost being able to react to them in real time as well as being able to react to them, retrospectively. And again, that's where we'd love to move to, particularly with things like customer experience score. Like, not only only are we generating customer experience score retrospectively after conversation ends, Can we actually generate it in the life of the conversation and use some of those metrics and insights, you know, to to drive action in in real time, to drive the customer experience? So, yeah, very aligned what Ben is saying around, you know, really think about real time metrics and how you can use those to to drive improvement all the time. Yeah. I I love that. I I kinda wish I'd asked almost you that question first because it it almost seems like what you just described is almost kind of like step 1 of developing a function to be able to understand where you wanna be developing the feedback loops and where what are the things you wanna be getting done and then kind of like moving into, like, how do we actually, tackle those challenges and problems. So it's almost like getting the clear picture and then dividing that up into the areas you're gonna attack and then developing your individual feedback loops based on what those individual challenges are. So it's, and then the the real time metrics is is great too, but, like, it's interesting seeing that kind of, that that kind of, like, workflow kind of, being being played out in in real time. I'm just being conscious of time. So what I'm gonna do now, we're gonna kind of move into, a bit of a discussion around kind of the evolution of the CS Org in the age of AI because one of the things that we've we've realized is that, AI has enabled CS Orgs to kind of meet rising demand without necessarily having to add additional cost in the form of significantly increased headcount. And that has huge implications for a business, including how you calculate ROI of the sport org and AI and how do you determine resource allocation, but it also kind of necessitates changes to both short and long term strategies for CS teams. And so, Ben, question for you around this. You know, beyond the pure efficiency gains, which we've talked about, how do you think companies should approach measuring kind of, like, the broader ROI of kind of, like, AI or these new strategies in customer service such as the, you know, impact on CSAT and long term loyalty? Yeah. So I think, I think there's a there's an interesting perspective on on it, which is the you know, instead of seeing it purely as like, oh, it it's the same stuff, but we're now doing a bunch of it more efficiently. And therefore, like, the the amount of stuff has kind of decreased or the amount of of time or effort here has has decreased, which is the kind of, I think, the, you know, economic argument, like, do more with less kind of thing. And, but I think there's there's this other perspective, which is that, you know, if you've got this team who you've had for however long and, you know, you're kinda happy with them working quite well, and and now you're adopting AI and it's taking away these, you know, x percentage of of that workload, then how do you take that team and their their knowledge and their skills and deploy that to actually grow the value of of what it is that you're able to do with those people. So, you know, not the same work, but now a bunch of it's being done faster. But say, well, actually, there's new opportunities that we have. And I think the the the things we I think but Declan and I have kinda touched earlier around, like, those conversations where, you know, you and your company have a chance to talk to one of your customers deeply 1 on 1 to leave them with a very positive, experience of interacting with your company and to potentially look at how they could be getting more from you. As I said, like, can they be adopting more features? Like, that has both, you know, good implications for them. They learn they continue stuff, has good implications for you. You said you're driving up the adoption of the things that you've spent time on. Can you give them some best practices, some guidance? Like, can you help them out with other things? Can you, yeah, just, like, nudge them for any fee any other feedback that they have. Like, here, let me request. Like, make them feel heard. Make them feel valued. Like, all that is the stuff that, yeah, it takes more time. And I think under the sort of old regime of just like, you know, just do everything faster. Like, there's there's more stuff coming in. Just do it fast. Get it, you know, get it done, that kind of thing. You've got fixed fixed resources and and this big, big warping to get through. It's really hard to justify doing that. But now that some of that pressure is being taken off by AI, you can say, well, yeah, actually, our goal isn't just to, you know, deflect and complete tickets. Like, our goal is to, you know, create happy customers. Like, the goal of any business, really, any business that intends to stay in it for the long haul and and sort of be successful and sustainable is to kinda create and maintain customers. And I think there's a really powerful opportunity to to do that here. And and so when it comes to what's the ROI, like, yes, there's a very clear efficiency ROI. Like, yeah, we now do 50% of our tickets are handled by this magic machine. Isn't that cool? And and it is. And then there's this other ROI, which is all the new stuff you can do. So you you can you can do what you had with less and you can do more. So you're kinda getting a a twofold, a twofold benefit, I would say, if I was marketing this. Yeah. No. That fantastic response. Completely agree. And you you touched on kind of this idea of roles evolving to be, like, more consultative, building that long term loyalty because they can now have the time and the energy and the headspace to actually do that because they're not answering the same 50 tickets every single day. Yeah. Declan, I I know we've talked a bit about this before in in in past webinars, but at Intercom, how have you started to think really about quantifying the ROI of AI and customer service, but beyond just the the the cost savings? Yeah. I mean, I think there's there's many dimensions to this, but just to kind of touch on a few of them. So for example, in our case, we've been able to radically increase the the service levels that we're willing to offer to our customers. Again, you know, be because we're now using combination of fin and human support and, you know, as a really strong combination, we can actually offer better service level objectives to our customers. We're offering an overall better customer experience in that. You know, we're we're now able to offer support 24 by 7. We weren't able to offer that before. We're able to offer, you know, multilingual support. We're able to offer, you know, to jump on a call with a customer if we need to do troubleshooting. We didn't have the time or bandwidth to do that in in the past. You know, so without adding any additional resource to our team, we we we can actually deliver better quality of service, you know, out of the box. That's definitely one thing. You know, and that ensures a better customer experience overall. But not only that, we're also driving, you know, what we call practice support and consultant support with our customers. So from practice support point of view, like, we're looking for triggers as to to where we can work with our customers that help them get more value out of the product. Like, we have this concept of a customer milestone framework where, we know how well a customer is using Intercom. We expose that information to our support team. They can engage with a customer in the life of a support conversation to actually say, hey, you know, you have you thought about enabling this feature or, you know, should making this configuration change because it'll help you in this way. So begin to provide that proactive support to customers is is a value. And then also consultative support, like, we're identifying maybe where customers aren't currently using Fin today, and is an opportunity for them to, you know, use Fin in a way that will actually help their business. So again, we're trying to make, our support team more consultative in that regard, and really trying to drive better value for the customer out of using, Intercom. That's something we will be able to track and measure over time or around the effectiveness of that. But we are also focused as, you know, Ben said, there is an efficiency, piece to this. So part of it around, in our case, it's more cost avoidance than cost reduction. Right? So the way that I kinda play it back to the business is, well, if we were trying to deliver this level of service, you know, to this, range of customers, because we also opened up the the the type of customer we support as well. So for example, we're providing better support for trial customers than we were before. There's some non customer support queries that come in that we actually answer as well today that we couldn't answer before. So we were trying to do all that work at the new service level objectives. We would have to add x number of heads to our team. In our case, like, by the end of, this year, that would probably be about 40 extra heads, which the business is a saving of somewhere between 3 or 4,000,000. So that's the tangible cost avoidance figure that we can generate for the business. So it's not cost reduction in that we wanna actually use the bandwidth that we have freed up on our team for more valuable work. But if we are trying to do all this, you know, just with humans and without AI, you know, that's the the the cost of avoiding this effectively in our case for the headcount. That's substantial. You know, and we will continue to track that. But what I'm really interested in seeing is how we can actually demonstrate that we're delivering more value for our customers over time and that leads to, you know, stronger loyalty, stronger attention, less churn, etcetera. And really what's happening, a lot of the lines between success and support are blurring. And that I see we're freeing up bandwidth for our team to actually provide more success type interactions with our customers, particularly our scale customers. If you think of self serve customers, they don't have a customer success manager. We can actually provide a lot of success type interactions for them through consultative support, through proactive support. So it's really changing the nature of the role. And certainly, you know, the team here have embraced that. They really love this opportunity to deal with customers in a different way. And I think we can play that into a very strong return on investment message back to the business. So so well put. Absolutely love it. So I am going to quickly just review the results of that second poll and also, just a reminder for folks, we're about to kinda go into the q and a part, so if you go to the q and a tab, you can ask a question if you haven't already. Also feel free to upvote questions that are in there. If you wanna see those answered, we'll get to those, next. So that second poll was about, the biggest challenge in operationalize operationalizing metrics across teams, and interestingly, it's it's kind of, similar there across the board, but, the one that's kind of standing out is combining those AI specific metrics with the traditional human performance metrics. And it's kind of good to see that because that's a lot of, like, what we talked about. That's a very common challenge that we've seen teams, teams facing. So before we do the q and I, I've got one, like, rapid fire question. So just, you know, quick 32nd response, but we'd just love to to talk to looking to the future. So, Ben, first to you and then Declan, you know, as AI continues to develop, it's it's it's impossible to fully predict the future, but what are some of the maybe, like, emerging challenges or things that maybe you're excited about, just throughout the realm of of metrics? It doesn't necessarily have to be through through the lens of of customer service, Ben, for you, but, what are you kind of looking forward to, with with the future? Yeah. I I think there was an interesting point around, like, I guess earlier, I kinda made the the the distinction. I said, well, you know, it doesn't really matter to feed these metrics back into the AI in real time. I mean, maybe it reached a point where it does. You know? I I may maybe that we reach a point where actually feeding more of your sort of operational support metrics back into the AI in real time can affect its performance. It's it's a little different because it's it's not, you know, it's obviously not quite so bandwidth limited. You know, you can answer a 100 questions in parallel at the same time. But there may be other interesting, other interesting aspects of that or, you know, to maybe to Declan's point, like, if it is able to assess in real time the sentiment of that conversation. Is it going well? Is it not going well? Like, it may be able to adjust in real time its tone to to rescue that interaction or to sense as a problem or even to, you know, to bail out if if, you know, you've got a really sort of hostile customer on the end who just wants to talk to a person that's getting really irate. So I think there are interesting feedback loops where, you know, it's not metrics in the sense of evaluating performance, but it is like they are operational metrics that, you know, you you can look at these, for a team or or for a given conversation, given interaction, and feed that back in and say, well, yeah, maybe that will end up being, an interesting vector to, to explore. Love that. And and Declan, how about you? What are you what are you excited about for the future? I think I'm really excited about how AI agent technology is evolving. So, like, things like Fin today are very good, what I call answer bots. Like, they will provide a really good answer based on the content that you've provided. And I think we're moving where these bots are becoming far more personalized and task oriented. You know, so you're able to, you know, interact with operational systems through what we call fin actions. You can pull in information that's really, you know, identifying that customer and personal that customer, and that can help the AI technology then develop a a more precise answer for that customer. And then potentially take action on behalf of that customer. So back and update an operation system. So you move into that realm, there's a whole new set of activities that can be automated across the business and not necessarily just support, activities. It's probably other activities across other parts of the customer journey Like, we can really look to automate when we move from this kind of answer bot to what I would call a personalized task bot type environment. And things, for example, like, what we call fin 2 and things like fin guidance, fin tasks, fin workflow, etcetera. They're all contributing to us addressing a whole other set of activities that can be automated using AI. That's what I'm excited about. It's it's really moving into the the next phase of AI technology and how it's, enhancing and improving customer experience and customer interactions. Yep. Absolutely love it. Great. So we're gonna hop into the live q and a, got some questions here. Again, feel free to go to the q and a tab upvote. I'm gonna do this first one here, this is from Dustin. Have you had experience with users making their experiences with support marking their experiences with support as negative due to, quote, talking to a bot? If so, what would you do to improve that experience? Declan, maybe that might be a better one for you? Yeah. I I mean, it it it's certainly a factor like there there are some people who don't like talking to a bot. So I think there's an important concept here is that you've gotta give customers choice. You know, you can't force them to talk to a bot. You gotta make sure they have the option to talk to a human if they really want to talk to a human. That's one way of kinda guarding against this this negativity. The other thing is around also encouraging customers to engage with the bot in a conversational way. An example of this is people today still have this historical view that in a chatbot world of all of old, you have to be very deterministic around. I gotta really get this bot to go around the right branch or flow. Right? So I'll give it a keyword or I'll give it a very terse set of commands. And you don't optimize an AI agent in in in that world. You really have to have a conversation with it. And we found, for example, by encouraging customers actually talk to the bot as if it's a human, we actually increase CSAT for for Fin quite a bit. Like, we increase it by 4 or 5 percentage points just by encouraging customers to think differently about that interaction. So, yes, there are some customers who just don't wanna talk to a bot and you gotta give them choice. Other customers, you gotta get them to try and engage with the bot in a conversational way because they'll get more value out of it, and ultimately, they'll get a better experience. Yep. Absolutely. And I've also found that once people start to have one good experience with with a really good AI bot, they they're much more likely to have a good one with them. This is an interesting one from from Barry. At what stage of the AI involvement do you track the CSAT score? Is it after the customer stopped interacting with the chatbot? Declan, maybe that one also for you. Yeah. Again, yeah, typically at the end of the interaction of the chatbot, we will track CSAT. If it ultimately gets hand over to a human, we also track CSAT at the end of human interaction as well. You know, so it's important that that that you're measuring both parts of that journey. Yep. Love that. This next one here, definitely for for you, Ben. Somebody good good shout out here from from Tobias. So Intercom and GECCOBORT have a great integration. It's in the app store if you wanna if you're already using both products. If not, definitely check out GECCOBORT. But, Ben, do you have any road map for adding more metrics in GECCOBORT? I'm using both now, but missing basically what is being discussed today. Yeah. Absolutely. Good good feature request, in the q and a there. So, yeah, we absolutely, on on our radar, as I mentioned earlier, it's not something we prioritized thus far because GECKOARD's SLAANT is more real time immediate feedback to the human side. And I think that the, you know, tracking some of those AI metrics is broadly better suited to a slower cadence. That said, I know that many teams and maybe, maybe your team is always one of these, do also use Geckleboard for that start of reporting too. So, I'll follow-up with you. I'll I'll find you in intercom, in our intercom, and I'll follow-up with you afterwards to have a chat about what it is you're trying to achieve and, what it is you're looking for. But thanks for flagging it. Great. Interesting one here from from Jenna. Kind of in the weeds, but I love that. Declan, this one's for you. You were talking about, like, uploading product feedback collection process. They're asking you, is this happening directly from chats with Fin? Is it a product feedback capture natively of intercom? What is you I think you had mentioned the community. Can you give some more insight there? Yes. There's something we're we're working on. We haven't built it yet. So so basically at at the moment, what happens is Finn can identify if there's an existing, feature request. They can suggest to the customer, hey, you should go to community platform and upload it. What we're trying to work on is use Fin actions to actually interface directly with the, community platform. So Fin can actually say to the customer, hey, this is an existing customer request. I can upvote this for you on the community platform. And again, using fin actions, we should be able to talk to the community platform and upvote it. You know, being transparent. We haven't developed that yet. That's what we're working towards. That's the next phase of of and we feature with but that's the type of power of something like fin actions. Like, so not just telling the customer go do this action. You're actually saying to the customer, I can do this action on your behalf, and I can actually upvote, in this case, upvote on the community platform for you. So that's what we're trying to build. So hopefully, that that answers the question. It is using the FinActions capability. Let's say it's something we haven't implemented yet. It's the next phase of of how we wanna handle feature requests. Love that. This next one's from home. We actually I would love to hear this both from from you, Benny, and from Declan. The kind of your opinion, the single biggest area you believe AI will contribute most to support. And I know maybe maybe we can think about that, maybe looking a little bit ahead because we did talk about the, you know, resolving like, there's just, like, informational queries being able to resolve those, so maybe looking a little bit ahead. So, Ben, maybe just starting with you, single biggest area that you see, AI contributing to support, and then, Declan, you can follow on. Yeah. I mean, I it's interesting to to try and consider, like, where where things land. I think that there is, I think, a big spectrum of of possibilities from, you know, how how far the complexity curve can it go? Because I think the outcome, if it's like 90, 95, 98, 99%, can be done by, you know, a really good bot supported by really good context, like, that really is materially different to even the state where it's like 80 or 70%. Like, it is it's really, really different. So I think it's yet to be seen where that caps out. I think it's gonna be quite company dependent. I think, you know, obviously, the the more complex your your product or service is, the harder it's gonna be to kind of rise up that up that curve. But I think that's gonna be really fascinating. But I think that there's there's another side to it, which is that, like, I guess at the moment, we we we talk about we're having to feed these things with with context. We have to, you know, create good articles for them. We have to create good knowledge guides for them as well. But maybe in the future, these these, certainly more agentic systems may be able to actually use your product and service themselves and then generate the content for you that can feed the AI itself. I mean, that's probably looking many hops far out, but it it's kinda interesting to consider. Yeah. Can I think it's not unfeasible to imagine that it may be able to take on even some aspects of that purely from a kind of brute force? You know? It's it's clicking around a browser trying to figure out how does this thing work kind of thing or or drawing parallels with with, other products and tools that it's it's seen in the past. So that's a little bit more wild, but, you never know. That's certainly one thing we've really learned. You really do never know. Love it. Great. And, Declan, do you wanna do you wanna comment on that? Yeah. So and, unfortunately, there's 2 things. 1, the ability to provide a very personalized experience, particularly as things like Fin actions that I need to interface with operational systems. I think that's one of the biggest areas of impact. It's a much different experience for the customer when it's personalized. And secondly, I think using AI insights to really allow support to be a really strong viable voice of the customer back to the rest of the business. I think that's the other area where AI will have a huge impact on support and enable sport to really move from being, in many cases, being cost center to being really a real value driven center for the the business. Yeah. Yeah. Love that. We are about at time. So first, if you're already using Intercom and GECCO Board, definitely go check out the integration in the app store, intercom.com/integrations. Go check out GECCO Board. Fantastic product. Ben, thanks so much for for coming, spending time with us. If anybody wants to try out Fin, just click. There's a button at the top of the screen. Just click start a free trial Fin. We'll send you information on it if you wanted to do that. But thank you everybody for being here. Declan, thank you for your time. Ben, thank you so much for for joining us. Hopefully have you back on on soon. And, yeah, I hope everybody has a great day and thanks for coming on today. Thanks, Mark. Yeah. Thanks. Thanks for watching, Mark. Thanks for listening, everyone. Thank you. Bye.