Name: How end users feel about AI Agents: The 2026 AI sentiment report
Uploaded: 2026-09-23T17:06:12.975Z
Duration: 43 min 56 s
Description: How end users feel about AI Agents: The 2026 AI sentiment report

Transcript for "How end users feel about AI Agents: The 2026 AI sentiment report": Hi, everybody. We're just gonna get started in a few minutes. I'd love if folks, could say in the chat where they're joining us from at the moment just to get the vibes going before we officially kick off. I'm gonna add my own one in here. I'm from Dublin, Ireland. We have people from, joining from all over. Okay. I am going to officially kick us off and say welcome, and do a little bit of housekeeping. So, good morning, good evening, good afternoon, good middle of the night if you're watching this back potentially, and welcome to our webinar today. I am Ruth O'Brien, senior director of AI support at Fin. So I lead the team who implement Fin the product for our own customer support team here at Fin, which is probably a little bit meta. And I'm gonna be your host today for the next forty, forty five minutes or so. So in terms of housekeeping, the session is being recorded, and we'll send it around afterwards along with a copy of the report that we're gonna speak about today. And the q and a tab is open now. So if you just look at where the chat is, over to the right of that, you can drop in questions for a q and a that we will hopefully have time to get to towards the end of our session. We'll try and reserve a few minutes for that, unless our chat gets, you know, super super deep and we end up going a little bit over. Yeah. I want to just introduce our speakers next. We have wonderful esteemed guests. So I've introduced myself a little bit already, but I'm gonna hand over to our guests in order of appearance here. Thanks, Ruth. Lovely introduction of esteemed gifts. I'm Caitlin. I'm director of research at Fin. So we look at, customer and our market perspectives across all kinds of topics Yeah. Finn, my researcher on my team, Maya, did an excellent job putting this research together, so I'm gonna be doing my best to represent that for you. Hello? Son? Yeah. I'm Simon. I lead customer and risk operations at ZayZoon. So we are a financial wellness platform at the intersection of fintech and HR tech. So we work with 15,000 plus employers across The US, which gives access to millions of employees, our core and wage access products as well as a number of other financial wellness related benefits. Hi, everyone. Thanks for having me. My name is Sneha. I lead some of our data products at Pendo. Pendo is a product experience platform, so we allow you to collect all kinds of interesting data on how your users and agents are using your product and and do some cool things with it. I'm excited to be here. Nice. Thanks, everybody. Really looking forward to the chat we're about to have now. So in terms of the next half hour or so, Caitlin is gonna take us through five key findings from the report. And after each one, we'll stop and talk about it a little bit. Then we'll close, with some practical ways support teams can actually build trust with people, and like I said, take on your questions. So, Caitlin, I am gonna hand over to you to talk about our first insight. Lovely. Thanks, Ruth. So just to start us off grounding us in the survey and the details of what we actually did. So we surveyed over 1,000 US based people who are currently employed across a range of different kinds of industries with the goal of understanding how they feel about interacting with AI agents, how much they trust them, and what would it take to make them feel more comfortable with this technology. So, researcher Maya, she supplemented the survey insights with six in-depth qualitative interviews, one to one, and each interview covered recent experiences with interacting with AI agents for online businesses, specifically, of course, along with their sentiment and their trust towards these agents. We made sure that we captured a diverse range of ages and genders and AI usage so that we could explore the data across all these dimensions and make sure that it's also representative of people. Halfway through the survey, we actually showed people a short video of an AI agent resolving a real customer query, and then we asked people the same set of questions again that we had showed them at the start. Now the goal of this was to measure how people felt about AI both in the abstract, drawing from their past experiences, as well as how they feel about it when they see how AI actually works and the full range of the technological improvements that we see at Fin as well. So respondents were split evenly between two video versions, one showing a chat version of that query I mentioned as well as one showing a voice agent too to test whether we saw any differences in how trust changes with different kinds of communication channels. So let's jump into the five key insights starting with how people feel about AI agents. So good news is that sentiment is warming. End users are feeling more positive than they did last year about interacting with an AI agent for customer service. 49% described interacting with AI for customer support as a positive experience, up nine percentage points from when we ran the study in 2025. And when they saw someone resolve a real customer support query, that number actually jumped to 74%. So interesting how that changed. 61 also think AI has become more capable in helping people with their interactions with businesses over the last year. So it seems like people's feelings towards the technology will warm the more that they see it performing successfully. It's so interesting. Right? Because it's like now seeing seeing us is believing us. I think so many people, they experienced so many bad chat bots or weird IVR things in the past. You know? There was a lot of, like, poor sentiment and lack of trust there. Simon, when you first launched your AI agent, how did your customers respond to it? And was there anything specific to help people feel more comfortable interacting with it, or do you think that they were already comfortable? I I think it's been a shift over time. I I think for us, the the same sort of pattern, holds true, that we see a uptick in sentiment when they've seen one positive or one successful resolution. For us in particular, it didn't necessarily move when the features and thing got smarter. It moved when we changed what it could say in the challenging moments. So, like, if we had to say no or we couldn't help someone, that's where I feel like it really started to shift and move the needle. Okay. Interesting. Is there some sort of example that you have where you have to say no to people? Yeah. So, again, in in particular in our world, obviously, with earned wage access, it's basically relying on individuals having the ability to draw from those wages. So in a situation where someone hasn't earned any wages or is in a difficult situation, we we might have to say no that there isn't any wages available. So that's when we get into those difficult situations because often when people rely on us, they are in a difficult situation. They've got rent to pay. They're stuck at a gas station. So they can be very challenging interactions. Okay. Yeah. And, like, they can definitely, I would imagine, get, like, a bit more emotional. Anything to do with, like, money. Yes. Yeah. Absolutely. Okay. Sneha, how about you? The research showed us, the more people use AI in their personal lives, the more positive they feel about it. Does that show up in product adoption data? Yeah. It's interesting because, you know, we have product adoption data about how they use it in their professional lives as a b to b software company. We don't have the other side of the coin on how people use it in their personal lives. But, just anecdotally, I do think the people that use it more, whether it's personal or professional, are more aware of what it's good at and what it's bad at. Right? And you we all know this intuitively. If you use a product, you know, not that regularly and especially with AI, if you're not using it that regularly, you're you're more likely to say, oh, it didn't really work for this and and sort of give up. But if you're constantly pushing the balance of it and my friends especially in in tech who are, you know, more on the frontier of testing what agents are capable of, they're more likely to use it professionally too and more likely to trust it professionally. But when I think about my friends who don't work in tech or my husband, for example, who works in construction, he's much less likely to to use it for professional use cases because he doesn't, you know, he's not testing it all the time. I also think another factor here is, in order to be able to test it all the time, both for personal and professional cases, you have to have unfettered access to it. Right? If you're constantly trying to think about how to optimize your tokens and how to do things in the most efficient way, maybe you're paying for it out of pocket. I think it's less likely for you to do things that may be perceived as wasteful. Right? Like, you kinda have to get the practice in in order to feel like you trust it, and then you're gonna ask questions that, that it's more likely to answer. So I think it has to do with, like, practice and expectations, really. That's such a good point about how freeing it is to not have to worry about things like cost or token usage. Like, I think about for us at Finn, we've been really empowered and enabled with cloud code, and given lots of, like, challenges with it or, you know, lots of onboarding with it without so many of those constraints around, like, don't even think about trying this thing. You know? And it's just been so helpful to allow people to use like, get used to it, in kind of their own semi private way, you know, and, like, try and figure it out. And then they're able to actually start building on that to do more impactful things in their professional lives. So, definitely a really, really good point. Okay, Caitlin. I'm gonna hand back over to you, to discuss insight number two. Thanks Ruth and I think it's interesting because that point also follows through with a couple of other points that we're going to talk about so we'll probably come back to some of this topic. So moving on to the next insight is around complexity. So complexity is where trust needs to be earned. But we say complexity, but the reality is is that that could mean many different things to many different kinds of companies as well as individuals. So 54% of end users say that they would trust AI to handle simple or routine issues, but not complex ones. But from our interviews, we learned that complex can mean many different things to different people. They mostly had concerns about how the comp how the query would be handled, not whether AI was capable of handling it, which is really interesting. These concerns about complexity, they break into about three dimensions, Whether the agent can exercise judgment rather than just simply executing on something, whether the query touches on sensitive information, to go back to Simon's point earlier, I think that's really relevant, and whether it needs continuity across a multi part conversation. Yeah. So, like, the the sensitive actually, the sensitive and complex side of things is so interesting because I was part of a workshop a few months ago where we asked a bunch of customer experience, you know, customer facing people what makes something complex. And it doesn't only mean really difficult technical work. It also means emotional sensitive, Okay. like, all that stuff too. So, Simon, to go back to your point about, many sensitive queries in your world, is there something that you're actually concerned about in this space, like, or something that you really wanna make sure you get right outside what you've already told us about? Yeah. Ab absolutely. So, like, sensitivity is really the the the default state for us, not really an edge case. So as you were kind of just alluding to the, complex for us doesn't necessarily mean a a difficult, complex type of question. It's a complex situation, which is, again, someone is stuck. Someone is in desperate need of, assistance. So the way we've kind of really tried to approach it with with Finn is, like, we kinda split what Finn can decide from what Finn can explain, to really separate the two. Interesting. Have you been using human in the loop, actually? That's something that we're using ourselves now to try and, use AI to empower teammates, with some extra kind of guidance, and then they can decide if they're actually going to use that answer or use that decision that Finn would have made but, like, wants to, like, loop in a human. Is that functionality that you've explored yet? We've started to explore. Yes. Yeah. It sounds like for your use case, it could be super helpful. So, glad. to hear you. Okay. Let's move on to our next insight. I'll hand it back over to you, Caitlin. Thanks, Ruth. So people worry AI won't own its mistakes or use judgment. I think this is probably something that a lot of people on the call could probably feel familiar with in your own personal lives. When people explained why they wouldn't trust an agent, Their reasons pointed to the behavior rather than its ability. So two concerns that came up consistently, whether the agent is accountable for its actions and whether it can exercise the same judgment and flexibility that a human would. So people imagine AI operating in a silo with no human watching over it or to Ruth's point, not being able to loop a human in or get, additional input from a human. As one respondent framed it, a mistake by AI carries no accountability the way it would for a named human being. 65% said knowing that they could easily reach a human would make them feel much more comfortable, and 57% want to be told upfront that they're speaking with AI. These are both design decisions that you could control and implement. Back to you, Ruth. Okay. I'm just thinking about that accountability piece from our side, within our own support organization at Fin. We have a robust Fin QA process, like quality assurance process, across lower customer experience interactions or negative, CSATs that we get. And that's our way of ensuring that there is accountability. There's folks, who report to me on the AI support team. This is a huge part of their job, making sure that if Fin does make mistakes, we try and rectify those, and that there is, like, a level of accountability behind that. But Sneha, on the, people want to wanting to be told upfront they're interacting with AI, Curious to hear, your experiences in that space. Like, that's something I actually, like, I really think we should tell people if it's AI and not fake it being a person. Looking to hear, your side here. Yeah. It's it's, for all of these situations, my mind just goes to all the different, like, scenarios that you could be using AI for. Right? If it's like a, you know, chatbot or support agent, of course, you wanna know that it's AI because you want to know whether you can trust what it's telling you. But ultimately, it depends on like like in the study, when people are showed that it can be successful, they're more likely to trust it. So I think it's the right thing to do to say that it's AI because it comes down to the brand of your your company. In the long term, I think it's probably gonna have the effect of increasing trust if you disclose when you're using AI and when you're not versus decreasing it. And, it's about, like, trust in the company is so much of what makes an agent successful. I mean, if everyone is really able to make the same product, then what's the differentiator? It has to do with brand and how much people trust the company. So disclosing AI is a big part of that. So are you gonna use Zuck's personal intelligence agent or Siri's personal intelligence agent? It I think a lot of that comes down to brand. Now for us at Pendo, we've been working on building what's called a product adoption agent, which basically creates these personalized emails, and and in product messaging that nudge people to adopt the parts of the product that we think would be a good fit for them. Right? And in cases like that, we've been experimenting. I don't have crazy results yet to share, but we've been experimenting with if the email comes from Pendo product adoption agent versus, you know, just Pendo or just or even from a human, which one are they more likely to open. And I think those types of people are gonna do more and more of those tests and and and see what the results are. So I think it has a lot to do with your brand and authenticity as a company, and the companies that disclose it are gonna use it for the right reasons and find the people who are gonna be, happy with the outcomes. Right? Yeah. I'm even thinking about, you know, if you, like, discover a company's lying to you, you know, by, like, saying an agent is an AI agent is a human, like, you're immediately like, your trust in that brand and your relationship towards that brand is gonna be quite different. But I'm thinking about, I came back from a vacation earlier this week, and I was going through my email inbox and just the amount of stuff I get, you know, like reach outs from different companies and vendors. And some of it's from, like, honestly being some sort of, like, automated message, and then other is, like, you know, human name from whatever company. But I know that they use, like, automation and AI for that. So, like, in that sense, I don't really I don't think I trust either of them more. You know? So it's kinda interesting to see the way all of this is developing and how like, I'm even thinking about my own trust in I work in an AI company, and I still have these these feelings as, like, a a customer or a lead myself. Simon, how about you? How have you addressed the, like, being open about the use of AI? And then is there anything else that you do to try and alleviate any of these sort of trust concerns? Yeah. I think the the piece that we did was, the accountability was built before the scope. So we made sure we had fully documented processes, hard, like, do not approve rules, structured fallback processes, and then a a full audit trail to be able to review. So we kind of we we earn the trust through constraining Fin, not by deploying all its capabilities, which is a unique way to to deploy an agent, especially in today's world. But given that we're kind of dealing with, again, high emotional states and a heavily regulated world, it was the best approach that we decided to take. Yeah. So what I'm hearing is, you know, like, build on like, build some trust, then build on that, then use more functionality, and then, like, it can keep getting better and better rather than going, like like, wildly all in and potentially breaking trust with your. customers. Yes. I think we've I think we've all seen versions of that Yes. around. Sadly, yes. It comes back to that piece around, like, the responsibility that we have even with this technology being so good now and getting better and better. A poor deployment of it or poor decisions from the humans behind it actually can have such an effect on sentiment overall. So I'm gonna hand back to you, Caitlin, to take us through the next insight. This one is my favorite, Ruth, because I think that this is such an interesting topic at the moment, so I'm excited to dig into this one a little bit more. So we see that trust actually varies and people's perception varies by the communication channel that they're engaging with the AA agent in. So the same agent can be doing the same job but is actually trusted differently depending on whether it meets someone in a chat or in an email or over the phone. So chat is where people are most willing to fully trust AI. So we see 21% would let it fully resolve the query with no human involvement over chat compared to social media. So you know Instagrams even like WhatsApp where 12% would fully trust it, so it drops significantly down. Phone is a really interesting one, though. This is the most polarizing channel of them all with 20% saying that they would fully trust it and 12% saying that they would fully trust it. So you're either on or you're off with phone. In our interviews with respondents, we discovered that this is largely driven by how people use AI in their personal lives. So people who use AI less frequently they might turn to the phone as a channel of connecting to a human and kind of getting to the heart of the issue because the self-service and other channels has already failed them. So for these people, and we see this a little bit more in older individuals in particular, they might feel really disappointed when they are met with AI agent on a channel that they would expect to otherwise be human led. That can be very different for somebody who uses AI all the time in their personal life. Oh, phone. I have such a love hate. relationship with phone as a channel, first of all. Like, when I think about myself being a consumer, or, like, you know, customer of many businesses and services, phone used to be, like, my default historically because I had no faith in anyone ever replying to my email or my form or, like, if I did use a chat like, old school chatbots that it would be some sort of, like, loop of doom. And then when I eventually get to a person, if ever, it would be, like, I'd have to repeat myself. So if if you picked up the phone, at least you get straight to someone who was just gonna sort it, hopefully, ideally. But I've heard other folks saying that they app like, they just never want to speak with anyone on the phone ever. So I see why it's like a a bit of a polarizing topic. But now that like, I go to phone less now actually used to before because more businesses have good AI agents or at least, like, you know, they're deploying good AI agents in an okay fashion at the very least, and some actually are really excellent, which is great. And, like, forms and things like that are starting to become a bit of a thing of the the past. But I actually had to ring my bank for something the other day and their AI agent and I was kinda surprised by this because it's like, I won't name names, but, like, a brick and mortar, very kind of old school bank. They're kinda bit slow to adopt new technology. But they had an AI agent on the phone, and it was brilliant, and it solved my issue really quickly. And I was surprised that their deployment of what I know can be good technology, actually was so strong, which was great. So maybe I need to have a little bit more faith overall. I think what's interesting is with phone as a channel as well, for us, and our support team, we actually don't have inbound phone as a channel, and we never have here at Finn, which is I think when I tell some other support leaders about this, they're like, oh, controversial. How have you not actually adopted that? Especially because, in previous jobs I had at other companies, they were more kinda call center based. So it was really different for me coming to Fin and it being mostly, like, mostly chat and email, like, async chat and email. And, yeah, like, if I like, I've heard a few people say that they would actually never if they could go back and start over, they actually wouldn't introduce inbound phone at all because it comes with all these other bodies of work you need to do around, like, staffing and scheduling and phone systems and all these different things. But the good news is for folks who have it and can't get away from it is that there are actually some really good, AI agents. Like, our our FinVoice is excellent. Even though we don't have inbound phone, we were actually trialing, FinVoice in our Messenger Messenger recently. So without having a phone system, customers could, try it, and we were looking at the quality of the conversations folks were having with it, and it was absolutely excellent. I don't know if we're actually going to ship that. Like, it was something we were trying and experimenting with, but it was really, really cool. So just like phone I'm going on about phone loads now, but, I because he's like you said yourself, Caitlin, it's the most kind of controversial one. Simon, any thoughts on the channel side of things? If you wanna vent about phone, you can unless you love phone, but anything else from this part of the report? Yeah. I think, similar to you, make mixed feelings on, phone. There's definitely one side which is, I would rather not have to pick up the phone and call if I don't have to. And then there's obviously the side where you see some of those great experiences where, AI voice can resolve queries. For us in particular, we we haven't deployed voice. We chat and email, and that's purely down to the the type of work that we do. So more often than not, we will require some kind of documentation whether that is, like, a payslip or ID or some or something kind of that nature. So we'd actually have to force people to channel switch. So instead of having to do that, we've just decided not to open phone or voice currently and just default it to chat and email. Yeah. We actually do out outbound phone if we need to. You know, if we find that it will be better to hop on a call with a customer. But same as yourselves, like, a lot of the time when people are contacting us inbound, they need to share logs or screenshots or whatever that you actually just can't do over the phone anyway. So we found that we've actually just, like, never needed the inbound side of things, as a channel. Sneha, how about you? Like, what are your feelings towards channels, be it from, like, a business perspective or a consumer perspective? Fellow phone hater for personal use. Hate begin with the picking up the phone to call people. I hate doing support over the phone, mostly because it feels really like you you have to be there and give complete attention and you can't, like, really multitask. And as a mom, that's like that just doesn't work for me. But at Pendo, we don't we also don't use the phone for for the same reasons you all have described. I think a lot of the, trust, with AI comes from being able to go back and forth, which is why people prefer chat because if something doesn't work, I can just go back with you immediately as opposed to email or whatever where I may have to wait. So maybe maybe that's a factor, but I don't have any, anything groundbreaking to say on this front except that I hate talking on the phone. There's, a really interesting little nugget in there though and that we have we're focused in on our polarizing views of phone, but we also see very different expectations that come with email versus chat as well. So it's worth thinking about, like, what are your end users expecting when they come with an email based support request versus a chat support request. Because email, you probably find this in your personal life, is a lot more formal. And so we bring that expectation to our support experiences too. And one little nugget that I wanna point out here, Ruth, is, like, you are proving in action that seeing is believing. Because when you picked up that phone and had that unexpectedly good experience, it started to change your view of, like, who can do excellent AI support, who can do an excellent AI support in a vertical that you maybe didn't expect. So even people that work at tech companies can still be delighted by customer support experiences. Yeah. Totally. And, like, I just think about, my past lives in, in call centers. And if we had thin voice as it sounds today and how good it actually is today, it would have just saved so much. You know, the people waiting on hold for, like, two hours, Yeah. some in some places I used to work, to just to get an answer to an FAQ rather than, like, all those kind of FAQs and probably semi complex conversations as well actually being resolved by an AI agent. And then the humans not having to sit in queues to talk to each other, and then when you do get through to somebody it's because you really need either relationship building or deep troubleshooting or whatever it would be. I only wish that we had what we have today back then. Customers would have been a lot happier, and my life certainly would have been a lot easier. But you're right about email. It's like even, like, the expectation for it to be more async than a messenger type chat. Like, when I send an email, I don't actually expect to get a reply from a business for a day, business day, like, TBD, Yeah. depending on the business, I suppose. But we. use Figgy over email, and it's giving those instant responses. And it's just interesting seeing how folks, react to that. Like, it actually works really well over email, but it's like I think sometimes people are kinda surprised to be getting the instant back and forth, and it probably just feels like a bit of a different experience. So, again, just like how all these things are changing and people's, sentiment towards it over time, it's actually really fascinating thing to to watch. There's also. there's a really interesting comment in the chat around, like, I I think there's an accessibility factor here as well with all of these different channels. Yeah. When you're building AI, you can't just build for, like, you know, usually, like, forward thinking tech people or young people or whatever. You have to build for the you have to meet your consumers where they are. So if it's an older population or if they're hearing impaired, like, you have to build accessibility into these experiences. So, sometimes, you know yeah. Anyway, that's I think that's a really important factor when you're building AI AI products. Yeah. Yeah. Totally. I mean, I think about, you know, when so much of the initial rollout of AI agents was sort of chat based, there's a whole cohort of people who that just wasn't gonna work for for various reasons. You know? Now having AI agents over phone, it can be, like, a complete game changer. So it's it's really cool to see all the developments here. Okay. Well, we move on to our final insight. I'll hand back to you, Caitlin. I told you that one was juicy, Ruth. Yeah. There's a lot there's more interesting stuff in there. Alright. So to to close us out with the last insight, beyond support, people ask whose side is AI on? So there's a little bit of a juicy one as well here. So extending AI beyond support introduces some different kinds of questions about trust. So whose side is the agent on? Our interviews with end users shed some light on why this new question of allegiance is arising because people worry about AI shopping assistance that will prioritize the business's financial goals over what's best for them as the consumer, assuming it's programmed to protect the company's bottom line rather than their own interests. This kind of calls back to some of the other points as well around the AI being able to use its own judgment too. At the root of it, trust is about stakes. It's just like in customer service, trust in AI in shopping is influenced by how high the stakes are. So if a purchasing decision that deals with personal information like health care, insurance policies, financial information, Okay. or something that has a very high cost associated with it, like a big ticket item, these are gonna automatically fall into a low trust tier for the consumer. People trust AI to help them to browse, but the moment that real commitment enters the picture, the trust starts to drop away. When we compare purchasing decisions and customer service use cases, we see a similar pattern, which means that many of the same levers can be used to grow trust in this new use case as well. Yeah. This one is an interesting one, isn't it? Because I think so much of it needs to come back to trust in the businesses themselves and the humans making the decisions about what AI is going to do and suggest. seconds. You know. It's not necessarily the AI itself. It's, like, what we choose to do with it, which is a very philosophical point, I think, across the across the world. Simon, what do you think about this? Why do you think people have this concern that AI agents are not, like, on their side? Is it something you've encountered with your customers, or do they feel the same way about this, across the the board, like, with humans as well? Yeah. I think they do, and I think it probably shows up sharper than it does in the kind of purchase purchasing shopping scenario. Like, for us, the trust kinda p shows up when when when we can't help someone or we have to say no. So the the kinda read on AI at that point is, like, AI or the bot is protecting the organization and not helping me, which is something that we we run into, fairly often. So it's something that we've had to kind of work through, and the the fix hasn't necessarily been the language. It's approaching what Finn can respond with and lead what the customer can do and what the path to a resolution is versus what they can't do or what the decision is. Yeah. It's like that piece about, like, you think, if you go and speak with somebody yeah. I even think about, like, say, face to face in real life. If you go and speak to somebody there, they might make an exception for you. But the idea that, like, if the computer says no, you know, and try to break down that that barrier. So, yeah, how about you? What have you encountered, with this insight? Yeah. I think we're all wary of being sold to, and every company is innately trying to sell something, right, which is where this sort of concern and fear comes from of whose side is on, especially in a b to c world where it's shopping or transactional. But even in b to b, the ideal scenario is that what the creator company is is is, offering via the AI is equally valued by the user or consumer. Right? And good companies with good brands and good AI strategy are not optimizing for making, you know, max dollars off of this one AI transaction. They're optimizing for customer lifetime value or, you know, retention or making someone a happy consumer. Those are really the metrics that you wanna be, you wanna be optimizing your AI for. So so whose side is it on? If it's on the side of the consumer, then it's gonna be on the side of the creator because that's that's the ideal world. But, you know, we've we've all been trained with thirty years of ads on the Internet, and we're all a little skeptical. It's the same skepticism skepticism I reserved for sales outreach. If I get sales outreach on LinkedIn, I assume, like, yeah, this product, they probably don't know enough about me or my specific use cases, so they're probably gonna try to sell something. And and we're all innately skeptical, and I think that's good. That creates stronger products because you have to get past that hurdle of initial skepticism in order to get there. Right? I'll tell you one, thing that AI is actually really interestingly good at. One specific use case is helping me to cancel services. I feel like in the traditional world, it's always buried somewhere in the nav. But if I can go to an AI agent and just say, hey. Help me cancel, it does it so much faster. And companies that do that, it makes me trust them. Even though I'm trying to cancel my service, I'm more likely to come back because you didn't make it so hard to cancel. You know? So that's one specific example of, like, I know it's on my side. Yeah. The shopping side of things is interesting, for me because, like, if if I'm on a shopping type app and it actually does a good job of being like, here's some other stuff like this, and sometimes those things are potentially cheaper or have a better review. I'm like, oh, actually, that was better. Even if it's. trying to sell me something, I was like, this is actually a better experience, overall. But then I think about some of the, streaming services that I use, and I think or, like, social media, and I think it's, like, targeting me potentially, like, on age, demographic, like, all that sort of stuff. And I'm like, oh, I don't like that at all. You know? This isn't like, it's it's different depending on the service or, whatever decisions are being put into what sort of suggestions are happening there. You know? Mhmm. Okay. So we've covered our insights. Are we happy to move over to q and a folks, or is there anything else in this final insight that you wanna cover? We have a couple of questions in the q and a tab. If it's alright, I am going to read one of them here. So has anyone created or tried guidance to tell their AI to acknowledge mistakes and tell the customer that it will send the interaction to an audit team for accuracy? I'm curious to try that, but also don't want something like that firing off too often, for invalid reasons, essentially. So, Arthur, thank you for that question. I'll hand over to, our guests here. Like, any anything you've tried or seen in this space around AI actually acknowledging that it's made a mistake and then saying it's going to do something about it? We've not tried that yet, but definitely curious to explore that a little bit. We kind of rely on, the kind of either CS score quality sentiment or, again, a manual human flag on the accuracy of the response, and then it goes through for review. We've not tried it that way. Yeah. I imagine, similarly, when I've worked on support related, AI, I think that's the same thing. If it's unsure, it'll flag it for a human first and just let the custom rather than make a mistake and then fix it. But kinda depends on whether you care more about, like, speed or accuracy. If you care more about speed and you think that's the more important lever to pull, then maybe that's worth the trade off. Right? That's how I would think about it. Yeah. It's an interesting day to get this question because I was speaking with, someone on my direct team about this earlier, where we're trying to figure out a way of looking at, like, negative, FIN CSAT specifically. So where a customer has told us, in a survey that they didn't like whatever happened in their FIN interaction. And we're like, how can we follow-up on those at scale or learn from them at scale? Because, obviously, like, we have, like, like, big volume of conversations. We can't manually review everything. And the person on my team today was showing me how they were using our operator functionality to run the negative, conversations through that, and ask operator to analyze where Finn made a mistake and, like, what could have been done differently. And some of the flags came up that the knowledge in our help center was incorrect or unclear. So that person went and updated the knowledge and then, asked operator to suggest a response to send back to some customers, and then they were able to take some of those responses and say, hi. You know, I'm one of the people who's responsible for Fin. I see how it didn't do a good job here. Thank you for telling us. Like, you know, sorry about that. Here's what we don't actually rectify it. So next time a customer has this type of question, we're, like, gonna be able to answer it properly. It was really cool thing to see today. And we're trying to figure out how to process size that. And, obviously, it would have been better if Finn, could have flagged that itself, you know, rather than us going in and being like, okay. A human's gonna decide to take these negative surveys and run them through operator. But it was, like, kinda one of these, earlier steps that we've, just kinda, like, closed the loop on the negative sentiment from the customer and try and actually do something about it. So hopefully, next time we all speak, I'll have an actual, like, full beautiful process and results and everything that we can show on this. I'm like, we improved customer sentiment by this much, you know, but it's really, really early days at the moment. Ruth, just out of curiosity, did you see the response back from the customer when you went back and said, oh, actually, we've changed things as a result of that? Not yet. asked because yet. I've was sent this morning. Yeah. It was only sent this morning, so I'll just, come back. to you, Anish. Yeah. Glad to see. Yeah. Super. Because. I've I've definitely I've definitely seen instances where we have gone back in other forums and said, oh, we have changed something as a result, and it's actually led people to be above and beyond happier in this intimate that even if we just resolve the query straight up originally, Yeah. like, the close the loop effort that we've we've implemented there really seems to land very well. So I'm very curious to see how you go as well. Yeah. I mean and also that was one of those ones where, like, a human on our team reached out, like, via their like, they an AI helped them create the response and solve the problem, but they reached out to say I'm responsible for Finn. I see what happened here. You know? So keen to see, like, I'm hoping the sentiment will obviously be quite good off the back of it. Because it's something that we've been doing with our human teammate interactions for, a number of years where anytime a customer leaves a negative survey for us, we actually we reach out to them and be like, I am acknowledging what you said and, like, here's what we're trying to do to either improve it or here's why this happened, or here's the help you didn't you know, like, whatever the reason is anyway. It's just with our fin interactions, our automation rate is so high that to be able to do that manually on every single one is is really difficult. But I just think that, like like you said, the sent the hopefully, the sentiment that customers have towards our business and our brand that someone bothered to actually check this, you know, or or bothered to implement, ideally over time an AI system that does a lot more of this flagging itself. It just means a lot to be like, you're not just trying to deflect everything. You know? You're actually trying to, ensure that customers get actually what they need, and then if they don't, then you're trying to rectify that with them. So, hopefully, the sentiment report will reflect that loads of businesses do that the next time we run it. Okay. Folks, we are just at time. We've just gone over forty minutes, so, we will start to wrap up. I'm just gonna show you a quick summary of the key key findings from our report and our survey. These will all be shared with you afterwards, so everything is going to be sent out, the report and all far more details about everything that we discussed today along with the recording. So if you are on the current recording now, or on the live chat, here's you can download the report here by using this QR code, but like I said, William added around afterwards again. So I would love to thank our guests today for the great conversation. Folks, you were absolutely brilliant. Hopefully, our audience enjoyed it. Thanks for all the engagement in the chat and for the questions, folks. And hopefully, in another year or so, we'll be able to say that customer sentiment is even better towards AI, as long as we're all doing our jobs and implementing it very well. So see you later, everybody. I hope you all have, a good day, morning, night, evening, wherever you are. And, yeah, catch you soon. Bye. Thanks. Thanks.