Why most brands build chatbots backwards - Interview with Sophie Cheng of Sinch

0:00So welcome to the next edition of the Punk CX podcast. With me today I have Sophie Cheng who is the SVP of product marketing at Cinch. Hi Sophie, how are you doing? Hi Adrian, I'm great. Thanks so much for having me. You're very welcome. Now, for folks, obviously there's not probably that many, but for folks that don't know you or don't know Cinch, can you give me a bit of a thumbnail sketch or a bit of background on who you are and what Cinch is all about? Yeah, absolutely. So Cinch is a leader in cloud communications, which essentially means that we help businesses across the globe deliver business critical, or you could say mission critical communications through channels like messaging, email, and voice.

0:41And so ultimately, our mission really is to pioneer the way the world communicates, which means reimagining how companies connect with their customers to keep them engaged, informed, safe, and happy. And my role specifically at Cinch is to run the product marketing team, which means I need to deeply understand our customers, the markets that we play in, and make sure that we are building the right products and ultimately keeping them there as well in the market. Awesome. Now, the reason why I wanted to set this up and have a chat is because you, Cinch, actually, recently undertook some research that found that chatbots rank among the lowest in customer satisfaction yet remain key to digital customer experience.

1:29And so I wanted to ask you to tell me a little bit more about the research, kind of what the headline, main headlines are, and what's driving that finding. Yeah, absolutely. So that actually came from a study back in 2024, and we just did a new study for 2025. But honestly, the differences are nuanced. So I'll kind of give you the latest research, which is we surveyed consumers across 12 countries and got 2 ,800 responses, in addition to 1 ,600 leaders from various kind of verticals. So a really good base of responses, both from the business side and the consumer side. I think fundamentally, it's just important to think about chatbots within the entire communication ecosystem.

2:16So ultimately, customers want choice, and they have clear preferences. So which channels they want to interact with when. So that means businesses need to meet consumers on their own terms. And obviously, on the flip side, each channel does bring its own unique benefits when you think about your holistic communication strategy. So in regards to chatbots specifically, we have seen massive improvements year over year. If you think about what that experience looked like, you know, last year, capabilities were still pretty limited. I mean, just the growth in adoption and sophistication of LLMs is mind-blowing.

2:55We're seeing much smarter, more contextually aware and autonomous agents on the rise. And so with that, we also expect satisfaction to increase because obviously it's delivering a better experience. So that skepticism that we're talking about will definitely decline over time, but it has not disappeared by any means. Yeah, because I think the thing that they also, I get that, I get that sense of it's still, I say in air quotes on an audio recording, it's still relatively new. But the research also found, and this may be updated in the newer piece is that he talked about preferences is that 20 % of consumers prefer never to interact with a chat bot.

3:39And yet also only 29 % will only do so if I guess no other option is available to them. And so I wanted to sort of get your perspective on what's going on here, because that feels like it complicates the picture as well. It does. And when you say those stats, it kind of looks pretty bleak, doesn't it? But I'd say ultimately people want support, not friction. And so until agents and bots really deliver upon that promise and the experience that customers demand, they will remain skeptical. On top of that, I think what we're also seeing more broadly is consumers are becoming much more demanding. So they just demand timely, personalized and relevant communications and communications that ultimately understand their relationship history and context, right?

4:31So a conversation should carry over from one to the next, even if you're switching agents or channels. And so that really does paint a very complex picture and one within which bots haven't necessarily been able to keep up in the past. Now that being sent, we're definitely seeing some green shoots. Okay. So I would say overall, AI adoption is definitely accelerating. Adoption from a user standpoint. So what we actually found this year was that 42 % of consumers are willing to interact with checkbots. Okay. That's a big jump. It is. Yes. And it's kind of driven by a couple of different things. One, I would say adoption is generational.

5:12So probably not surprising, but millennials and Gen Z are embracing AI-powered interactions for speed and ease of use. Now, older generations, you know, tend to be a bit more skeptical, and they are particularly worried about transparency and data usage. So the drop is actually drastic where we're seeing Gen Z respondents saying 71 % of Gen Z said they would happily work with a chatbot, whereas only 20 % of boomers said the same thing. So that obviously paints a drastically different picture. And then when you think about how the general workforce is changing, but also our makeup of the different generations, that kind of speaks towards that acceleration of adoption.

6:02We're only going to see more and more of that. So that's number one. Okay. I'd say number two is situational, meaning AI adoption depends on what use case we're talking about. Right. Okay. Right. So, and you probably know this from your own experience, Adrian, but people are much happier, for example, to reschedule an appointment with a chatbot than divulge, you know, really personal health-related details, for example. So also not necessarily shocking, but it does speak to the fact that people are still very skeptical and don't necessarily trust this AI-driven experience from a data security and privacy standpoint.

6:43Right. Okay. And then the third one, this is one that we'll just keep repeating because it's so important, is consumers expect channel choice and relevance. So they just expect to be able to pick different channels and have that option at the ready, which both means having options to reach out to a business too, in terms of I want to pick if I'm going to send you an email, call you, interact with your chatbot, or also choice when I receive messages from a business. And so that just means it puts a lot more pressure on companies to really match those very high and increasing expectations. yeah no it's it's so it's a complicated picture that people's preferences are changing rapidly the people are coming more afay with technology but they still have these hardwired sort of preferences around kind of choice and it's as you say it's very situational and very contextually driven but i think there's also with the the whole technology adoption and implementation sort of side of things there i wanted to explore like perspective that you shared where you said And obviously, we want to make these things better.

7:55It doesn't want to just be one and done, as it were. But there's still a big, as you say, a big bit of skepticism around the use of chatbots amongst many, various different demographics. But you also shared with me a perspective where you said that you believe that most brands build chatbots backwards. And I was like, ooh, tell me more about that. What do you mean by that? Yeah, absolutely. So what we mean by that is essentially that a lot of companies approach building a chatbot from a tech first mindset. So they really focus on building a technologically advanced tool, let's say, rather than focusing on the customer journey.

8:39And so this ultimately just leads to bad experiences because those chatbots don't reflect the real conversational needs or the context. So what we really need to start doing is balance kind of that automation with empathy, let's say. So that means understanding the full conversation and relationship history. It means being able to pick up on nuances if somebody is getting angry or upset and knowing when to then actually escalate from an AI agent to a human agent, for example. And then ultimately, of course, also designing experiences that are very intuitive. You know, I would say that the good news overall, though, is it's clear that customers crave these conversational experiences.

9:27And I don't think that only means craving connections with other humans. It means they crave, you know, really good, timely, relevant, personalized responses in whichever manner they can get them. which means we are seeing that rise of chatbots or RCS, for example, which are giving you those more immersive experiences. Yeah. It makes me think, because I wrote something last year about this idea, and it's to do with chatbots. It was sort of inspired by, I was having a conversation with a bunch of people, and I said, I was quoting some research, and it was a Gartner research, I think, from last year, that said something like 70 % of all customers say that they would be happy to use chatbots to self-serve.

10:16But yet only 17 % of them, I think that's about right, were able to achieve their objectives when using some of these kind of tools. And when you then dig into it and you look at how successful they would have been for very, very simple tasks, that number only rose to about 30%, 35%, even for very simple tasks. And my point was, rather controversially at the time, was to say, hmm, your chatbots are rubbish. And I said, but here's why I think they're rubbish. And I said, but I approached that by asking them a question. It's like, how many, I said to them, or two questions I asked, I said, let me ask you, who are the conversation experts in your organization like who are the experts in having conversations with customers on like a daily weekly hourly kind of like basis and they're like well probably are people in service and support the things and i said okay how many of those people have been uh involved in the design and implementation of your conversations within chatbots and then it was like tumbleweed no and i said my and the thing i wrote on the back of that conversation i said to be successful in within there's two types of ai i said there's artificial intelligence and then there's agent intelligence and i think you sometimes have to combine those things to be successful to sort of leverage that that collective super intelligence that you have in your installed base so people that are really experienced in dealing with customers on a daily basis and also the power of the new technology that's emerging.

12:08And I think that speaks to your point around building chatbots backwards. If you don't get the right people, it's not just about the data and the technology and the journeys and all these different things, but if you don't actually have the right people from your organization involved in delivering these things, then you're probably not going to achieve the outcomes that you want. I couldn't agree more, Adrian. It definitely speaks to that. You have to build it from a customer first mindset and not from that tech first standpoint. And that's a great point. You definitely should be bringing in your frontline teams that interact with customers every single day, actually know the nuanced questions that they ask or the areas where they get hung up.

12:49And if you do it well, you really can deliver a much improved customer experience, but it can't just be bots that are these simple if then statements. Yes. And so, I mean, I guess underlying that statement around most brands should build chat bots backwards must be a belief or a conviction that we can like, here's actually what we should do instead. I mean, can you walk me through that sort of like design and implementation kind of journey about what brand and how the sort of journey that brands should be going on in order to deliver these more effective tools and better outcomes? Yeah, absolutely.

13:33So I'd say overall, obviously, you need to start with your customer and their problem. So you need to understand what are the main use cases that your customers come to you to solve, right? Is your influx in support requests typically, for example, around rescheduling appointments, or is it understanding what health benefits your insurance covers, things like that. That's data that companies typically have at their fingertips. So definitely start with your customer and their problems. And that means you will ultimately deeply understand their needs. But then you also kind of need to layer that in, in terms of their preferences.

14:11So it kind of goes back to, again, you do need to give them that choice of channel. You need to make sure you're bringing in relationship and conversation context, which carries over between those sessions and channels. And that is a very complex thing, which means you need to have data from in-person conversations that have happened that maybe you've added to your CRM or ERP. It means you need to understand what somebody's recent purchase history, or maybe even just browsing history was. You need to understand if their recent experience was positive or negative, for example, did they leave a review somewhere or fill out your NPS scorecard?

14:48So you need to kind of take that entire context for one single person into account, which is now when you think about how AI can help, tremendously useful, because that is just impossible for one person to kind of go through and do that in real time. For example, if you were a human answering to questions on a chatbot, or if you're saying, responding to a phone call, And then obviously, you kind of also need to have things like understanding the in-moment interactions. And what I mean by that is, what is that sentiment? Because if somebody is starting to get angry, you need to know when to triage or give them choices to kind of switch channels.

15:31So I'd say it's not about people being against bots per se, right? It's much more, people don't want bad experiences, which I think we can all appreciate. We have problems, we want them to be solved as fast and as seamlessly as possible. Now, one thing that I think it was fascinating that came out in our research, Adrian, was two-thirds of businesses seem to think their communications are integrated. So when I say seem to think, they said, yep, all of our communications are nicely integrated, which means between channels and across experiences. And yet the biggest frustration from consumers was that they constantly receive unwanted, outdated, irrelevant messages.

16:15So there is clearly this massive disconnect that we need to solve for. I think it's interesting that you talk about that and that's consistent and it's been consistent for years over different domains. I remember back in that, there was an old piece of research that came from Bain from probably more than a decade ago that said, oh, 80 % of brands think that the experience that they deliver is great, and yet only about 8 % of customers agreed. And it's this sort of thing is that they're stuck in this, I don't know, their own little bubble, as it were, about thinking about, I'm not sure there's a cognitive bias that's related, that overestimation bias or something that's related to that.

17:00I mean, so that's a problem. How do you crack that? Totally. And it's kind of wild that we're talking about this, you know, decade or decades later still, but I guess that's where it is nuanced. Things still have improved, I would say, over time. But you've also obviously got those rising customer expectations because people know what good can look like. Yeah. I think they also just start demanding and expecting that. And the flip side, I think, which is probably more daunting nowadays is there's a lot more choices for consumers. So one bad experience with a brand or a company is often enough for people to switch.

17:45So, you know, this is a very risky business if you don't get it right, let's say, which really means companies need to focus on building trust. And I think that's absolutely where it kind of comes back to, you need to put the customer front and center. And so if you're not in a position yet to be rolling out communication channels that will actually support your customers in the way they need to receive or expect to receive support, then maybe you should focus more on improving that experience first, I guess I would say. Don't build a chatbot for chatbot's sake. I know, indeed. I mean, for me, I think it feels like you've got to collapse the distance.

18:27There's almost that perspective distance. And I sometimes, I also going to think about the idea that you've got to almost like change your behavior to change your perspective. You almost have to go and seek the insight. So get out from behind your laptop or your desktop or your kind of away from your dashboards and stuff and see things how they actually are. I mean, I remember speaking to a, who was he? It was the CPO at Zopa years ago. And they're like an online peer-to-peer lending kind of platform. And they did this fascinating thing where, because they didn't really speak to the customers on the phone, they didn't have live interactions.

19:12It's all sort of done as web-based. They brought a whole bunch of customers into their offices and had them actually do some live transactions in front of them. But then there was a whole group of people behind this, you know, one-way mirror sort of thing. They were watching it. It was a bit like a market research kind of like studio. And they were watching them doing these things, and it was designers and engineers and marketeers and communications experts and everything else. And thankfully, the room was soundproofed as well because people in the team got to see the customers of theirs actually interacting with what they designed in real time.

19:49And they were screaming at the window going, no, it's over there. Do this. Do that. It's simple. But there's a completely different reality when you see things in the wild. And I think that feels to me that that's the sort of kind of different types of effort you've got to take to grab that insight to be able to collapse that gap between what you think is good and how people actually experience it. Does that make sense? It definitely does. And I think it's one of those things that you just have to, I guess with most things in life, it's trial and error, right? You can put all the research into things, but at the end of the day, you ultimately have to put it out into the wild and have your real customers and prospects test whatever experience you're putting out there.

20:37So yeah, it totally makes sense to me. Yeah, sorry, I'm just one of those things I kind of bang on a little bit about. But like going, sometimes you have to get out from behind your desk and go and talk to people and see how things work in real life. Because that's where you'll get a real insight in terms of to see, to your point, around you get to see the real friction, right? A hundred percent. But I also think that, as you said earlier, that the technology is improving exponentially, and it's mind-blowing in terms of how fast it's improving. New models are coming out every day, different updates and all these things.

21:17But this technology is also not without risk. And I wonder if when we're talking about, particularly chatbots, but I'm sure it's applicable for other channels and other different functionality capabilities, I mean, what are some of the pitfalls that brands should be on the lookout when they're wanting to lean into the use of AI assistance or generative AI in general? Yeah, that's a great question. I would say, one, definitely be transparent that you're using AI or AI agents. If you're trying to couch your bot as an actual human, people see through that very quickly and they get really frustrated.

22:00and we know there's enough frustration around, you know, communications already if brands don't follow the experiences that customers expect. So make sure you're, you know, transparent from that standpoint, because trust, privacy, and kind of accuracy are definitely the top concerns that we found in our research, both from a consumer and business standpoint. So it's interesting, there's concerns on both sides, which is good. That means businesses are thinking about this in terms of we need to make sure we keep our customers' data safe. We need to make sure we're not building models that might hallucinate or make wrong recommendations, for example.

22:41And then from a consumer standpoint, it's obvious that they're concerned because they don't know if they can trust the technology yet. So they need to kind of have a couple of positive experiences with AI agents overall and then a particular brand kind of start building up that muscle, let's say, to leverage and trust those chatbots. So you definitely need to make sure you've got guardrails in place as a business while you're building these AI agents. What type of data can they pull from? Making sure you've tested all sorts of variations, training it on your own internal data, for example, to make sure that it really will only pull from relevant support documentation for example to make sure that you're providing those accurate responses to your customer base and then knowing when to when to triage i think that's definitely a really important thing as well perfect so all of this stuff is all very very kind of like um right it's great it's interesting it's useful but i always also like to rather than just talking about things in general terms i also like um to ask people if they've got stories that they can share with me about brands that they might have worked with to show kind of like how their approach has had an impact as it were because it makes it makes it real if you know what i mean so if you've got a couple of stories about brands or instances that you've worked with kind of brands that do the around the kind of chat more friendly you can tell us about and what they're doing and how they're benefiting what outcomes they're driving that'd be that'd be great yeah absolutely um i mean we have dozens and dozens of examples cinch has worked with customers on building these types of personalized experiences for decades, really.

24:32And so one example I could give you is Intermarche. So you might know them, they're a French retail giant, and they really wanted to increase customer engagement with their brand. And they were looking to create a more immersive or interactive brand experience, let's say. So a lot of people are now turning to online shopping, obviously. But then with that, you're like, how can we make this more engaging and ensure people, you know, keep coming back? So one of the things we worked with Intermarché on was developing a recipe generator bot. So this is like a more fun application of how you could use an AI agent where, you know, it knew what was in your shopping cart and could then suggest recipes that you might want to consider cooking, but would also then know if you, you know, need extra ingredients, for example, to be able to make whatever we were just suggesting in terms of that recipe.

25:29And so it was definitely one of those things where they just saw a massive spike, I think, in customer engagement. And, you know, still waiting to hear, I guess I would say on retention numbers but it was a campaign that was very well received and just kind of a more fun and lightweight way to get consumers used to interacting with a chatbot. Right and was that the sort of thing where then somebody was shopping online that it would that it would pop and say ah have you thought about all these things here's some recipe ideas would you be interested or how did that kind of engagement sort of thing kind of like kind of work?

26:08Was it that they had to go looking for it or did it just go, have you seen, do you know, that type of thing? Right. Well, it essentially started with their newsletter. So a lot of people kind of sign up to get their regular newsletter. And so they really knew these were fairly engaged folks. And that's how they initially rolled it out, where from that newsletter, you could kind of click through, you'd end up, you know, it'll give you the, when I say newsletter, we're talking like deals of the week, for example. You'd click through and then you'd start shopping online. And And then kind of as I want to say, it was once you kind of got close to the checkout experience where it all of a sudden like pop up and be like, hey, these are the things you could make.

26:48Or consider adding this so that you could end up making, you know, you're missing tomato paste to make spaghetti bolognese. Right. Okay. No, no, that makes sense. I mean, because again, that's kind of part of just enriching that whole kind of like that part of the journey for people that are there. they're open to that now that's that's cool because it it does show that you can be creative with some of these kind of like this technology and it's not just a purely functional thing where people are going oh my i want to do this and therefore you design it to achieve that but it actually can be there's something which can potentially enrich the you know the experience if they're going through like another kind of like process um i agree i mean any more though you want to share with me?

27:31Yeah, I mean, I'd say there's tons of very functional applications as well, right? So I'm not advocating for only start with lightweight kind of fun things. There's really tangible, helpful ways that you and your company can be leveraging chatbots today already. So another example is we work with a lot of banks or financial institutions. And obviously here, we're talking about sensitive data, right? So when you're talking about somebody's finances, you need to be pretty secure in that experience, let's say. But it is one of the most common use cases, I would say, where banks are looking to leverage chatbot experience to support their customers and get them faster responses.

28:18So a bank that we worked with in Belgium, their name is Belfius, we actually helped them simplify their insurance claim processing. So that was a process that used to take quite some time. And now customers were able to visit their website and log their insurance claim online, get support from that personalized chatbot that would obviously understand that individual's background or relevant details. and ultimately what we saw was an increase in completed insurance claims and a decrease in time it took for the customer care team to support customers so ultimately what that means is you know people are getting their claims reimbursed um the customer care team doesn't have to get involved as often but it also happens at a faster clip so ultimately you know you're definitely improving that customer experience and satisfaction overall.

29:16Yeah. No, thankfully, I haven't never had to submit very many insurance claims, but I do know from the experience of having to do it that it can feel quite onerous and also a little bit back and forward. And so having something that can guide you through the process, like, oh, we need to see this, I mean, you need to see that, and oh, this is your policy, but tell me about this, what happened, and send me photos of that to guide you through the process, to gather all that sort of right detail to accelerate the process, but to reduce that sort of back and forward interaction, which would normally happen.

29:52Agreed. Yes. It can make it a much more seamless process. And you can just ask a lot of basic questions, which sometimes maybe people don't even feel comfortable asking if it were a human agent on the other end. Insurance is always a very complex topic. And so these are excellent use cases. to make it easier for people to get the help they need. Perfect. So, Sophie, we've covered a bunch of different things around trying to make these experiences more customer-centric, and therefore more effective. And it's not just functional, but it can also be creative, and it could be also efficiency-driven, and so on and so forth.

30:32And I think there's some really good stuff in there. So is there anything else that we've missed that you'd like to add or highlight before I ask you some quick fire questions at the end? Yeah, you know, I would say one stat that came out of our recent research was 97 % of businesses are using or planning to use AI within their customer experience in 2025. That is up from 63 % last year. So that really is a seismic shift if you think about how businesses are thinking about leveraging AI. But I think the crucial point is actually not that everybody's thinking about it, but more that adoption isn't enough, right?

31:13That real transformation is only going to come when companies integrate AI, automation, orchestration, that kind of rich intelligence layer around an individual customer and putting that customer at the center. So to me as a marketer, but also working for Cinch, this is an incredibly exciting time because I actually think it's never been more important, but also never more achievable to kind of create these types of experiences that customers really will love. Excellent. Excellent. So some quick fire questions before we wrap up. So we talked about a bunch of different things, but I always like to ask people to almost give me, to boil it all down or give me their best advice.