Customer conversations hold an enormous amount of signal—but most enterprises aren’t measuring that signal with the depth or nuance it deserves. In this episode of The Dialogue Architects, Lauren Goerz talks with Etie Hertz, GM of Conversation Intelligence at ContentSquare, about how conversation quality—sentiment, friction, and tone—directly impacts customer lifetime value. Etie breaks down why CSAT fails, how granular sentiment scoring works, and what the future of support looks like when behavioral analytics meets conversational intelligence.
Customer conversations hold an enormous amount of signal—but most enterprises aren’t measuring that signal with the depth or nuance it deserves.
In this episode of The Dialogue Architects, Lauren Goerz sits down with Etie Hertz, GM of Conversation Intelligence at ContentSquare, to explore how the acquisition of Loris is bringing together digital experience analytics and next‑generation conversational intelligence to transform customer support, retention, and lifetime value (LTV).
Etie shares how Loris evolved from a real‑time agent assist tool into a full conversational quality and insights platform used by both CX leaders and AI teams. He explains why traditional CSAT is insufficient—low-response, biased, and missing data—and how more sophisticated scoring systems can capture the full emotional arc of a conversation.
Lauren and Etie dive into Loris’s granular sentiment scale (-2 to +2) and their Conversational Quality (CQ) metric, which evaluates tone, empathy, friction, contact drivers, escalation patterns, and even churn prediction. They discuss what makes conversations work—mirroring tone, acknowledging emotion, reducing repetition—and where they fail.
The conversation also explores:
• Voice vs. digital trends in support
• Personalization through behavioral + conversation data
• The role of cultural nuance in AI and human responses
• Where friction hides in refunds and policy conversations
• How widespread AI agents may reshape human interaction
A practical, insightful episode for anyone building or governing conversational systems in CX, operations, or AI.
Etie Hertz is the GM of Conversation Intelligence at ContentSquare, where he leads the integration of Loris’s conversational insights platform with ContentSquare’s global digital experience analytics. With a background spanning product strategy, AI‑driven optimization, and enterprise CX, Etie focuses on extracting actionable signal from conversations—helping organizations understand sentiment, reduce friction, and directly impact customer lifetime value (LTV). His work sits at the intersection of behavioral analytics, conversational design, and responsible AI.
(00:00) Meet Etie Hertz
(01:05) What ContentSquare Does
(02:26) Why Loris Was Built
(03:23) From Agent Assist to Insights Platform
(03:56) Customer Lifetime Value (LTV) Explained
(04:54) Measuring Conversation Quality
(07:49) Sentiment Scoring and CQ Metric
(09:41) Beyond CSAT: Why It Falls Short
(10:50) Root Causes, Attribution & Privacy
(13:13) Human Empathy Lessons from Data
(15:14) Humans vs. Bots: Tone Matters
(20:22) Teaching Empathy Techniques
(21:33) Refund Friction & Sensitive Scenarios
(22:15) When Chats Go Wrong
(24:14) Fixing Repeat Requests
(25:16) Voice vs. Digital Channel Shifts
(27:15) AI Agents Calling Support Centers
(27:48) Personalization Through Support Data
(30:05) Behavioral Signals Explained
(32:05) Conversational Commerce & Search
(34:31) Optimizing Customer Journeys
(35:45) Cultural & Linguistic Nuance
(37:51) What Conversations Reveal About Brands
(39:51) Closing Thanks
The Dialogue Architects is a podcast from Rasa, hosted by Lauren Goerz, exploring the craft and strategy behind designing conversational AI in the enterprise. Each episode brings together product leaders, designers, and engineers to unpack how dialogue is built, scaled, and governed in real‑world systems.
[00:00:00] Lauren Goerz: Today we're joined by Etie Hertz. Today is absolutely an exciting moment for me because we're talking with someone who has loads of experience with real human conversations. Etie is currently the GM of Conversation Intelligence at Content Square, and we're gonna really be just diving into this experience and ability to track, understand.
[00:00:20] Lauren Goerz: What works well in real human conversations and see how we can transform the learnings into the digital world where we are here at Raza, trying to automate conversations between humans and machines. Thank you so much for joining us, Tai. Glad to have you on our podcast today.
[00:00:49] Lauren Goerz: I had really love to kick off with a little bit of maybe an elevator pitch for Content Square, especially since we're sitting here with the head of general intelligence of conversation. I'd love to learn more about Content [00:01:00] Square, what you do, and really why is your product Awesome.
[00:01:03] Etie Hertz: Awesome. Thanks for having me.
[00:01:05] Etie Hertz: So Content Square is a leader in product analytics and digital experience. Our mission is to help brands improve conversion, retention, and customer journeys by understanding how their users truly experience their digital environments. So, as technology and consumer behavior is evolving, uh, we've sought out additional ways to understand engagement and, uh, conversational channel specifically.
[00:01:24] Etie Hertz: And that was actually why Content Square acquired our company Lo about, uh, six months ago. Uh, conversations are kind of the, the next iteration of what's happening online. And, um, at Lores we spent about six years building AI powered customer intelligence products and like Content Square, um, we're very, very focused on customer interactions.
[00:01:43] Etie Hertz: So LO helps brands understand sentiment, emotion, and friction within support, tickets, chat, and voice, uh, conversations. So bringing together the digital experience and the conversational experience makes total sense considering where the world is going, where conversations are becoming the new UI that people are accustomed to.[00:02:00]
[00:02:00] Lauren Goerz: Definitely heard that, that that chat is kind of, uh, the new app or something like that. That's, that's something that's been going around at least, at least in my circles here, in the kind of broader agent AI world. I'm curious to know a little bit, especially as you talk about going deep into the sentiment, the understanding of conversations, um.
[00:02:16] Lauren Goerz: Tell me a little bit, what was, you know, what problem were you seeing in customer service and how did you start to come about experiencing these problems yourself before you decided, okay, I've gotta solve this problem.
[00:02:26] Etie Hertz: So the, the moment in time when we started at Lores, um, there was a huge shift in the market from people speaking to customers via.
[00:02:32] Etie Hertz: Voice channels and increasingly through digital channels like chat, SMS, email and messaging. So we analyzed hundreds of millions of conversations and we worked with some of the best, uh, universities in the world to develop a system of models that would allow us to understand how conversations were going initially in, in, in the digital form.
[00:02:49] Etie Hertz: So our initial use case was, uh, real time agent assist. It would sit on an agent's desktop and it would guide them in real time throughout a conversation. So you have to imagine these agents, typically they went from, uh, a [00:03:00] voice. They would, most of them were not in the us they're all over, they're Asia South, south America, central America, and they worked a lot on improving their accents.
[00:03:06] Etie Hertz: Suddenly they were tasked with chatting with three conversations at the same time in English. Very, very difficult. So we really improved their lives. We made it easier and more seamless for them to interact. We, um, we boosted basically everything about their job, and we delivered very high impact and, and positive conversations.
[00:03:23] Etie Hertz: Um, our models became so powerful. That our clients started asking us to actually QA the conversations. QA is a huge business in customer support. How are agents doing? How are conversations going? And our models were able to do that really effectively. We initially did that for humans. We then evolved to also do that for AI agents and chat bots.
[00:03:40] Etie Hertz: Um, and then finally we have an insights platform, essentially, which is available for the head of CX or other C-Suite. Folks wanna understand what's happening in the customer conversations and what useful information they can derive from that to improve LTV essentially.
[00:03:56] Lauren Goerz: Okay, now let's break that down.
[00:03:57] Lauren Goerz: LTV. 'cause we've, we've got a lot of acronyms that come through, [00:04:00] probably your space and mine as well. What's LTV? Why does it matter?
[00:04:04] Etie Hertz: So LT v's. Lifetime value of a customer. Uh, so there's every ingredient goes into that metric. Essentially. What are people spending? What are your customers spending with you?
[00:04:11] Etie Hertz: How often are they spending it? Are they increasing their spend? Are you upselling them? And on the flip side, are you churning them? Um, support historically has been kind of seen as a cost center and a way to mitigate churn. Uh, but the more forward thinking brands actually use it to boost LTV when someone has a great interaction with a support, um, agent.
[00:04:31] Etie Hertz: They're more likely to spend more money and be a, a lifelong or longer tenured customer. So that's a really important investment, but un understanding the impact of those interactions with your customers is very, very difficult, but incredibly valuable. And so you'll see that, um, I would say five years ago, only like the best companies were doing it.
[00:04:49] Etie Hertz: And today, I think it's table stakes. Everyone's gonna have to do it to survive.
[00:04:54] Lauren Goerz: Okay, so when we think about tracking LTV, I guess one of the things that is challenging, at [00:05:00] least as I know in the kind of agent ai conversational AI space, is actually measuring the quality of a conversation. And I know, um, especially as we were talking earlier about kind of the, the conversation be an under leveraged growth, um, lever ultimately for, for these businesses.
[00:05:19] Lauren Goerz: Um, ultimately. When you, when you look at the quality of a conversation, how does that directly impact LTV? Is it, is it just, you know, can we just point to broad customer experience? Mm-hmm. While that was better, is there actually like key components that you can break down from your analysis that that shows, you know, this component absolutely affects LTV over time.
[00:05:40] Etie Hertz: So first I guess we should talk about quality. Um, historically the way people measured quality is usually through something like CSAT surveys. Which, um, have a 10% response rate. You get 15% if you're lucky. It's super biased. And so the first thing, actually the first model that we created was a very robust sentiment model that tracks not just every [00:06:00] conversation, but every back and forth within a conversation between an agent and a customer.
[00:06:03] Etie Hertz: So it's very, very granular. Um, tried and tested throughout the world and it's, it's, it's very, very useful. We've, we found it, um, be so powerful that some enormous brands are able to use our sentiment model to predict churn, likelihood of churn. So there's a lot of things within our sentiment model. Um, not only how did the conversation go throughout, did someone come in upset?
[00:06:24] Etie Hertz: Did they leave happy? But, um, how many back and forths were there? Did the sentiment change? Was the sentiment targeted towards the support agent or was it about a broader policy or product issue that the company has? Oftentimes CS will say, we, we've done our job. We had a great conversation. They're, they're not happy with the 14 day return.
[00:06:40] Etie Hertz: I can't change that. So those nuances are really impactful and you're able to parse them out and do very, very meaningful things. So, as I mentioned, one of the, you know, world leading brands was able to use this to predict whether a customer would come back and, and buy more from them just based on our sentiment model alone.
[00:06:57] Etie Hertz: We have other markers within the conversation that we're tracking [00:07:00] quality. Um, we group conversations in terms of like what questions are being asked, what are the root causes? We have conversation markers throughout. There are a lot of different things that, that we enable. For folks to be better understand the quality.
[00:07:11] Etie Hertz: Now, once you have that, if you wanna use that to derive LTV, you can do, you can run regressions and you can see if, if some of these things are correlated to repurchase orders or, or cancellations. Um, the big unlock here is combining Loris with Content Square, you're gonna have the front end analytics and you're gonna understand what people are doing on the websites and in the app, coupled with the subjective, very qualitative, um, data about how people are feeling as they go through these journeys.
[00:07:38] Etie Hertz: And once you start to try to correlate those activities, it's gonna be a much, it's gonna be much highly correlated, much more highly correlated to LTV than anyone's been able to do historically.
[00:07:49] Lauren Goerz: Now I'd love to double click on sentiment as a metric. 'cause I think that's something with within Agen ai we're kind of in the baby steps of working well with contact centers.
[00:07:57] Lauren Goerz: I would say we've got often a loaded [00:08:00] developers or you know, let's say business folks that are suddenly tasks with, you know, building this. Conversational interface and then suddenly find themselves working with a load of contact center people. It's not always the case. Sometimes these teams evolve from the contact center, but oftentimes it's the case where, you know, you're learning CSAT basics.
[00:08:16] Lauren Goerz: Um, I think CSAT is one of the most common that I see in agentic ai. Conversational interfaces. I'm curious though, when we talk about measuring sentiment, what does that measurement actually look like? Is it, is it, is it really just a benchmark against yourself or is there some kind of quantifiable, um, method of measuring sentiment?
[00:08:37] Etie Hertz: So. Sentiment itself. We, we, uh, uh, scrub five point scale to every interaction from a negative two to a positive two. And, um, what we're trying to do is mimic if, if two humans looked at the same conversation and had to score it, would they, would they come up with the same score? That's essentially what our models are doing.
[00:08:53] Etie Hertz: Then we feed that sentiment and that journey of a conversation into a metric we call conversational quality or cq. [00:09:00] And, and that should be a proxy for what you would get in a CSAT survey scenario. So if you go back to csat, if you tell me how you survey, I, I'll basically tell you what your CS SAT's gonna be if you wait 90 days to send it to somebody, versus if you send it five minutes after the interaction, if you send it with a picture of your dog next to your face, when you send the survey out, you're gonna get a higher score.
[00:09:18] Etie Hertz: You can manipulate CSAT in a lot of different ways. Um, and so being able to track, using our, our models to automatically track every single interaction live, that's gonna be probably a, a more accurate metric that you can do different things with. Csat, for better or worse is, is used by a lot of folks for promotions and things like that.
[00:09:36] Etie Hertz: It's not going away anytime soon, even though it's, as I mentioned, it's pretty biased.
[00:09:41] Lauren Goerz: Nope, that's, I, I see the same in, uh, conversationally, I teams broadly, because what'll happen is we'll often issue the CSAT at the end of the conversation. And so what happens is a lot of people abandon before that point, so all the bad experiences don't get captured.
[00:09:55] Lauren Goerz: Mm-hmm. Because these people just leave and we, they're either not logged in, we have no way to contact them. [00:10:00] So we're really only measuring people that get to a point where we might be have the chance to ask a csat. So I think that's something, you know, I think it'll change as, as we, as a kind of function space become more advanced.
[00:10:13] Lauren Goerz: In terms of like things we can do in the contact center, do we know who it is? Do we have a user id? Can we contact them later? But right now. It's a big black hole, I would say at least for, for conversational agents. How do you solve that or how have you seen teams solve that From the conversationally I perspective, because I think that that is a gap.
[00:10:29] Lauren Goerz: I, I guess sentiment maybe is, is the solution,
[00:10:33] Etie Hertz: um, yeah, sentiment and like you wanna get as close to the interaction as possible. And, uh, the beautiful thing about analyzing a hundred percent of your interactions, you're getting 10 x what you would typically get in a CS a, uh, scenario. So it's a much more robust platform and a much more robust way to understand how folks are feeling.
[00:10:50] Etie Hertz: Um, we haven't spoken about, but one, you know, one of the things we do incredibly well is we, we build, um, large clients and, uh, contact driver models. So we're essentially. I very [00:11:00] accurate at understanding why people are contacting you, what the root cause of those interactions are, and then when you layer on sentiment on top of that, you can start to find incredible, um, insights.
[00:11:09] Etie Hertz: So when people contact you about a certain product or about a competitor, or how are they feeling about those things versus when they're calling about your product or they're calling about your service, there's, there's a tremendous amount of insight you can derive once you start to layer these models together.
[00:11:23] Lauren Goerz: And is this then connected all up into kind of a user's unique user profile? Ultimately somewhere saved it, you know, where where does all this information live in reality?
[00:11:34] Etie Hertz: Um, so we're very strict with security and PII, so we, we redact everything. We have no, uh, personally identifiable information. Now, um, in the future, there is a way to connect this back to the conversion side of the house and, and use, you know, hashes or different proxies to figure out that this person that.
[00:11:51] Etie Hertz: Lauren or 1, 2, 3 came in and had this kind of experience with respect to this product. Um, that's kind of where this is going in terms of like personalization and being able to [00:12:00] deliver specific, um, services and solutions to specific users.
[00:12:06] Lauren Goerz: Okay. Okay. No, makes sense. Um, I think that's, that's also something we're early days I would say in terms of, especially when, when every single call for us runs through a language model and it's in real time, it often is getting sent to opening ai.
[00:12:20] Lauren Goerz: Different, different language model providers. So that's definitely, I wouldn't say like a fully solved issue in, in our space, but I guess you have the benefit is that you're doing this after the fact, um, in many cases in terms of this analysis piece. So that, right. The analysis is post
[00:12:32] Etie Hertz: interaction. We're using predominantly our own models.
[00:12:34] Etie Hertz: We do leverage LMS when, uh, when it makes sense, but yeah,
[00:12:38] Lauren Goerz: no, makes sense. Alright, well thank you so much. I think, I think that piece about sentiment is my biggest takeaway from kind of the section in terms of like how you could, what could we do on the measurement side in conversational AI to make ourselves better and really understand our users better, because I think we've really only scratched the surface.
[00:12:54] Lauren Goerz: So I'm, I'm really excited. It sounds like you've done this already with some conversational AI teams, so definitely a good, good reason to check out Content Square if you're, if [00:13:00] you're looking at really diving into that. Sentiment. 'cause I don't think there are many platforms out there that do a really, you know, let's say even using proprietary models to really hone in and figure out the, the, uh, sentiment behind our users today.
[00:13:11] Lauren Goerz: I would be really, really interested. This is the part I, I think I'm, I'm super interested to learn more about the actual things that you've learned from watching so many human conversations, from analyzing so many human conversations we're currently in an interesting space because.
[00:13:26] Etie Hertz: Um,
[00:13:26] Lauren Goerz: previously as conversation designers, as dialogue architects, we were very much blocked by our technology from delivering excellent experiences.
[00:13:33] Lauren Goerz: So a lot of my job used to be working around the technology, and now it's flipping in the sense, you know, when I, when I go and build a conversational interface, I'm thinking. Wow. You know, I'm, I'm more thinking how can I block it? You know, how, how can I limit it, which is a total 180 from what I, what I used to be doing.
[00:13:48] Lauren Goerz: I'd love to learn from your experiences now that we can have really natural conversations with machines. What have you learned from natural conversations with humans? What makes conversations better?
[00:13:58] Etie Hertz: Um, I guess initially when we [00:14:00] rolled out our product. The, my most surprising takeaway at the beginning was just the impact you could get from a human meeting, the other human on an emotional level.
[00:14:11] Etie Hertz: So people come in typically and support, they're upset. How do you match that tone? How do you make the, uh, the customer understand that you actually care about them? So like. People used to over apologize and be too empathetic in a sense, and that's, that felt really inauthentic and fake, and people hated that.
[00:14:27] Etie Hertz: Uh, on the flip side, how do you make sure that you, you do meet them where they are so that you can start before you even start to tackle the issue and what they're, what they're here to solve. They have to know that you're here trying, like with their best interest at heart. So that comes naturally to some, some agents, and it does not come as naturally to others.
[00:14:42] Etie Hertz: Plus you have this constraint where, um, oftentimes the agents have to handle several conversations at a time. So it's pretty hard to do that. Um. In, especially in like regulatory environments like FinTech, there's so many things that they have to say or not say. They oftentimes forget the, the emotional side of [00:15:00] it.
[00:15:00] Etie Hertz: And so even prompting them with certain things would, uh, would help them remember and would, would, um, would ultimately take them to a better place with the, with the end user. So there's just a lot that you can gain from matching tone between humans. If you wanna talk about with humans and bots, it's a whole, whole different conversation and it's much, much more complicated.
[00:15:19] Etie Hertz: Um. I'd say with respect to like humans talking to bots until a year ago, the experiences were pretty terrible. Most bots were rules-based. They couldn't do much, and they oftentimes tried to do more than they could do, and that would frustrate the human, so the human would find a way to, to match with a customer service agent.
[00:15:37] Etie Hertz: Then generative bots, you know, showed up about a year ago, and they're more interesting. They're better, they're higher potential, but oftentimes also brands are using them. In use cases where the humans don't wanna interact with that bot yet, so you can't really match tone with bot, uh, uh, humans don't typically want to have a machine be that empathetic with them.
[00:15:57] Etie Hertz: Like sometimes it'll make sense, and maybe over [00:16:00] time people will become more accustomed to it as they use Claude and Chachi PT at home as, as a consumer. Like maybe that will become more the norm. But I think that there's a big distinction between how, um, a human agent interfaces with a, with a consumer and an AI agent, interfaces with a consumer.
[00:16:17] Lauren Goerz: Hmm.
[00:16:17] Etie Hertz: I can get like way deep into this. Like you don't want, right?
[00:16:19] Lauren Goerz: No, I would, I would, I would, I would love to. You don't wanna, if you're up for, because I think I have so many, sym is something we've been trying to build in for a long time. I would agree that, um, there's, there's like the, but really the bare basics, right?
[00:16:32] Lauren Goerz: If someone comes in and says, you know. I, I've just an example, I had once I've, I've had a death in the family, I can't pay my mortgage. I mean, pretty brutal, right? And, and that happens. These people come in and they, they start chatting to you about it. And really, it almost, I think, I think in the, the reflex reaction in the past has been kick it to a human.
[00:16:52] Lauren Goerz: We don't wanna touch that. We're not gonna be able to deliver a good experience there. And that may well be the case for a while. I, I don't know the answer there, but I'd be curious, um, you know. Uh, [00:17:00] you talked about maybe apologizing too fat too much, and that made it sound inauthentic or not empathetic.
[00:17:05] Lauren Goerz: What does make an authentic, empathetic experience? Is it just something that is unique to the human delivering it and it's something that we just either manifest or have or don't have? Or can you teach it? Can you learn it
[00:17:19] Etie Hertz: for human to human interactions? You, there are just like a pretty tried and true set of techniques you can use.
[00:17:26] Etie Hertz: To kind of mirror the, um, emotional state of the end user. There are a lot of different ways to do that, and that's been studied and it's generally pretty reliable with respect to how humans and AI agents are interacting. Obviously this is brand new. It feels like we've been doing it forever, but it's only been, you
[00:17:42] Lauren Goerz: know,
[00:17:42] Etie Hertz: a short amount of time.
[00:17:44] Etie Hertz: A lot of things are gonna happen. I think as you, as a human just increasingly talk to AI agents, you're going to become less formal. Less emotional. You, you say thank you. When you speak to a human, the more you interact with the Claude Chet, you're not [00:18:00] gonna say thank you. You're just gonna ask them for the thing.
[00:18:02] Etie Hertz: And that's gonna probably permeate across society. Or stop. You're gonna not say thank you to other humans as much because most of your interactions with AI agents, like we're gonna train ourselves almost to communicate more often with agents, AI agents than humans. Um, so it'll be kind of weird if you're talking really.
[00:18:18] Etie Hertz: In, in a cold fashion to an AI agent, but that agent is, is empathetic and apologizing and say, thank you. It's gonna seem pretty strange. Right. And you also have the uncanny Valley issue. So I, there's a lot to learn about how, how people want to interact with AI agents. Some of them will wanna speak to them like they speak to humans, but I suspect a lot of them are just gonna want to get in and get out.
[00:18:39] Etie Hertz: Yeah. Um, it's, you know, we had, we had a, one of our clients was, I don't wanna give away the clients, so, but they, they would sell treats to animals, basically. And, uh, one of our data scientists one day was like analyzing a bunch of conversations, like moon's conversations. And he said, this is like a therapy session that sometimes sells treats, [00:19:00] right?
[00:19:00] Etie Hertz: So people just wanted to con contact and they wanna have a conversation. And maybe that's gonna be true with AI agents. Uh, they're gonna wanna just. Communicate with AI agents, even though it's not a human. But I think for the most part that the whole, the whole way that people communicate with AI agents can be very different than how they communicate with humans.
[00:19:18] Lauren Goerz: I think from our perspective, maybe like insights from what we're seeing is we're, we're definitely seeing that switch in terms of how people talk to us. Um, there, there's an expectation of ability. Which sometimes is not met by, by a lot of these experiences that are kind of still in development. But it is interesting 'cause we're seeing loads of people talk to us way differently than they did previously in the sense of, you know, like long paragraphs saying thank you, things we didn't have previously nearly as much.
[00:19:44] Lauren Goerz: So I think we're starting at least on our side to already see, you know, as I go through logs, this, this change a little bit and, and I think a data point too, I remember, I can't remember if it was someone from OpenAI or Anthropic were saying, you know, they were wasting billions of tokens. You know, per day on people saying [00:20:00] thank you to these agents because they were scared of, I don't know, agent AI apocalypse or something.
[00:20:04] Lauren Goerz: But nonetheless, I think, I think that that is changing, and I think one of the reasons why we were never speaking like a human to these agents is 'cause they couldn't speak like a human back to us. So I think there's an interesting case that, that it might change in the future. But, but that's my, my perspective for now.
[00:20:17] Lauren Goerz: Let's see how long, long that holds on, I guess on the, the empathy perspective. Mirroring is actually from the conversational design perspective, something that is part of our design procedures. So I think in many ways a lot of the things that we learn and do, follow what contact center agents learn. So I'd be, I'd be curious to learn if there are any other kind of tips, thoughts, tricks in terms of how you teach contact center agents to be better agents.
[00:20:42] Lauren Goerz: I'm curious if we can maybe find some parallels in our space.
[00:20:45] Etie Hertz: Um, mirroring obviously is, is definitely something There's, uh, oftentimes when people are upset and they explain why they're upset, you need to kind of say it back to them and exacerbate the issue so that they, they really understand that you get it.
[00:20:56] Etie Hertz: There's basically just a bunch of different techniques to, to signal to the other [00:21:00] side. Like, I'm on your team, we have an issue. If I can solve it, I'm gonna solve it for you. We're not enemies. 'cause typically people come in like, it's a zero sum game. You're trying to get something for me. I'm the brand. I don't wanna give it to you.
[00:21:10] Etie Hertz: And it becomes this, you know, you don't want it to be this negotiation. You want it to be this problem solving thing that we're both doing together. Um, now when, when you have start to have agents on AI agents on both sides, that's gonna be tricky because that probably ends up becoming a negotiation. If you have an AI agent and I'm a, an AI agent, responding like empathy is irrelevant.
[00:21:33] Etie Hertz: You're gonna want a refund or so, you're gonna want something and I'm gonna wanna probably not give it to you. Or I'm gonna have an issue giving it to you, some friction, giving it to you. And so what is that gonna look like? Um, super interesting. Again, maybe mi maybe mirroring as an overall umbrella is the right is the right move.
[00:21:48] Etie Hertz: If someone comes in cold, you respond cold. 'cause maybe it's a bot. Um, a lot. There's a lot to learn there. I also think, yeah, it's gonna get super tricky because, uh, volumes are also gonna explode. Everyone's gonna have [00:22:00] their own consumer AI agents running around doing all kinds of things, buying and returning things and shopping, et cetera.
[00:22:06] Etie Hertz: Um. Okay. This whole thing is gonna evolve in a very, very unique way over the next, I think, year or two.
[00:22:15] Lauren Goerz: I guess even then we've, we've talked a little bit about what can make conversations better. I, I'd be curious to understand also what you've learned about what makes conversations go really bad really quickly.
[00:22:25] Lauren Goerz: If there's any kind of inflection points that almost always results in something horrible, or maybe it's even a certain domain or a vertical that's just really hard to serve. I'd be curious to hear about your learnings there.
[00:22:37] Etie Hertz: I'd say the number one thing is, again, people wanna feel heard, so anything that you do that indicates you're not listening, um, like if you repeat yourself a lot, if you, if you force the, the user to explain their issue again, like you didn't get it the first time, that's probably true of a human agent or an AI agent.
[00:22:54] Etie Hertz: Um, you, you obviously, you're, you're familiar with folks talking to a chat bot and then getting upset. The chat bot doesn't get it. [00:23:00] Like the chat bot's not listening. It's not picking it up. Yep. Um. People in every domain, they wanna feel heard and they wanna feel like the brand is here to solve their problem.
[00:23:09] Etie Hertz: So anything that you do that doesn't necessarily do that is probably bad. Um, you wanna be problem solving, you wanna be thoughtful. Again, they, my analogy essentially is like the, the user needs to feel like they're sitting next to the agent, not across the table. Mm-hmm. Like, we're in this together, we're gonna solve it together.
[00:23:26] Etie Hertz: Anything that you do that doesn't make me feel like that as a user, is gonna upset me, whether that's a human agent or an AI agent. And so I think, you know, people were hesitant to use AI agents historically. 'cause again, they were rules based and the outcomes weren't great. Um, but if you can problem solve effectively with AI agents, people will get over it.
[00:23:45] Etie Hertz: They'll be willing to deal with a lot of the age and can solve their issues.
[00:23:49] Lauren Goerz: Super interesting, and I think we're even seeing, like I, I've seen in, in the past with the, with the A, within the AI agent space, it's been typically just classic customer success. I was recently talking to someone [00:24:00] at one of our events who had built a negotiation agent purely for negotiation.
[00:24:04] Lauren Goerz: I mean, that is extremely open-ended. There are any number of things that could happen in that type of conversation. So I'm really hopeful in terms of like. What we'll actually be able to do in a year from now and two years from now. Um, on the, on that front, in terms of things going bad, having to repeat yourself, that's something that we're still solving for.
[00:24:22] Lauren Goerz: Let's say you really do have to make a customer repeat themselves. What's the best way to handle it? We haven't cracked memory entirely. It's something we're working on really, really closely here at Raza. This kind of global contextual memory across all of these different systems. Sometimes you don't have it.
[00:24:35] Lauren Goerz: How do you fix that problem?
[00:24:37] Etie Hertz: So for humans, it's, it's much easier. You can basically, like, like Lauren, I'm gonna say back to you what I think your issue is, and correct me if I'm wrong. So obviously I'm being empathetic. I'm trying to, I'm trying to be on your side. Uh, let me know if I missed anything that, and then I can basically validate that I've understood the issue.
[00:24:51] Etie Hertz: So even though I might be, we're forcing you to repeat yourself, there's a good reason for it. If you're an AI agent, it's a little bit more difficult 'cause you [00:25:00] should have picked up what Lauren said in the first place. Um, I don't know how comfortable you'd be if the a, a AI agent asked you to repeat yourself again.
[00:25:09] Etie Hertz: Like probably you get away with it, but I don't know that you have as much leeway as you would with a human.
[00:25:14] Lauren Goerz: Yeah. Okay. Fair. Fair. And I'm curious also, when you look across modality, so we're talking voice or IVR and, and chat, digital web, WhatsApp, whatever, you know, the, the many different types of, uh, written channels that we have.
[00:25:29] Lauren Goerz: Are there any trends that you notice that are also important in terms of serving customers differently to people who message agents who message behave much differently than agents who talk on the phone?
[00:25:40] Etie Hertz: I think the users are different. Um
[00:25:42] Lauren Goerz: hmm. Interesting.
[00:25:43] Etie Hertz: Typically, younger folks don't want to call, they want digital service.
[00:25:49] Etie Hertz: Um, maybe an older consumer is more comfortable and accustomed to speaking on, on a voice call. Voice calls are much more expensive. They're much longer. You can only have them one-on-one. So you have to be a certain kind of, of brand to [00:26:00] essentially enable that kind of voice. Probably a high ticket. It's worth it for you to, to staff your, your, um.
[00:26:06] Etie Hertz: Your team like that. But increasingly, more and more folks, again, the younger folks want digital. And as people become more accustomed to talking to, like again the chat gpt and clouds of the world, they're gonna bring that experience to their shopping and to their, um, to their day-to-day lives as they interact with customer sport.
[00:26:23] Etie Hertz: They're gonna be used to chatting and used to chat responses. I, I would expect that is gonna grow dramatically. And obviously when you have AI agents, you can do AI agents in voice now too, but probably they're more digital. And so I guess in general, I would expect digital channels, the, the volume of digital channels to grow exponentially and voice will continue to shrink.
[00:26:45] Etie Hertz: I think as people have less and less time for that.
[00:26:49] Lauren Goerz: Hey, it's super interesting to hear this from an outside perspective because in right now, I would say voice exploding in, in the broader kind of, let's say, agent AI vendor space because probably because of [00:27:00] the cost issue it's so expensive to handle. So any opportunity to automate that is a win.
[00:27:05] Lauren Goerz: ROI immediately for, for a lot of organizations. But fair enough. It sounds like, you know, actually, like if you're really looking long game, it might not be the channel of choice for your
[00:27:14] Etie Hertz: category. What's gonna, what's gonna happen to today? Like 30% of people say they're willing to let an a I agent complete a purchase for them.
[00:27:20] Etie Hertz: So you're probably, your default would be like, let them run around and chat and try to buy something or sell something or whatever. But maybe that channel is, is um, is hard to get through. So let them, what do I care? I'm a consumer, I'll call you. And that's gonna cost you five times what a chat would cost.
[00:27:35] Etie Hertz: I don't care. But now you have to figure out how to respond to all these voice calls that are being run by AI agents as well. Um, and then it's gonna be a matter of like, what are tokens gonna cost you versus service servicing? All of these AI agents that are inbound, it's gonna get crazy.
[00:27:48] Lauren Goerz: I'd love to dive in kind of last segment for today really and look, look deeper at the future of personalization in conversation.
[00:27:55] Lauren Goerz: This is something that in chat, we already have loads of data. We're [00:28:00] getting loads of data. Our customers are telling us exactly what they want in their own words, which is fantastic, and we're really not leveraging it. And I think, um, over at Contents Square, as you mentioned earlier on in your intro, you're doing some pretty cool things about Consum, you know, learning your consumer be behavior and then implementing that back into the user experience.
[00:28:16] Lauren Goerz: I'd love to hear more.
[00:28:18] Etie Hertz: Sure. So historically what would happen is we would watch you, let's say we know that, that this person, that Lauren, you know, number 1, 2, 3, is on the site doing a certain amount of things, moving their mouse, looking at at certain products that are interesting or not interesting, potentially engaging within the app back and forth.
[00:28:34] Etie Hertz: But in between all of these interactions, there're oftentimes support conversations that were never being picked up before. So you would have all this objective data and you try to derive. Insights from that. But what if I can then tell you, Hey, by the way, in the last three weeks while Lauren was doing all those things, she contacted support three times, and here are all the things that she said.
[00:28:53] Etie Hertz: You not only understand so much about what you want, what you were looking for, why you got frustrated, why you succeeded in purchasing or [00:29:00] failed, but you might be a proxy for another thousand people that have the same journey on the website or in the app. You just happen to be the person that interacted with support.
[00:29:09] Etie Hertz: So you're the one who gave us all those golden nuggets that now inform how the journeys of all these other people went. So just imagine from a brand standpoint, having all of that data combined, it's never, it's never existed before. You, if you want to, as a brand, you'll be able to customize your entire experience for every single user.
[00:29:25] Etie Hertz: Um, it's, it's incredible.
[00:29:28] Lauren Goerz: And this is pretty cool for us as well, on on the agent side because when we go to automate a service, what will happen in the future, and this is really quite literally, we're like on the ground building this right now, is being able to pull in all of that data and then inject it at the right moment into the conversation and allow a language model to flexibly use it.
[00:29:47] Lauren Goerz: In the past, we couldn't be flexible, right? We had to have that information in and then deterministically operate on behalf of it. But now. With memory, once we have this kind of global memory layer that we're able to activate and [00:30:00] leverage these, these, I, I think these systems are gonna be so much smarter than they were previously.
[00:30:05] Lauren Goerz: And that's pretty neat. So I'd be curious, you know, when you talk about behavioral signals that that happen, you know, and I think this is something we are all going to need to learn in the future as we go to serve, serve customers with conversational interfaces. What are some of the behavioral keys that, that you've seen in the past?
[00:30:21] Lauren Goerz: I know you've said mentioned, for example, rage clicking or cart confusion. Mm-hmm. These are all new words for anyone that's coming from my background.
[00:30:28] Etie Hertz: So when someone hits a site, there's certain things some people like, just imagine it. I guess the easiest way for me to explain it is imagine an in-store experience.
[00:30:35] Etie Hertz: Some people they, they walk in, they know what they want, they're buying it, they wanna get out as fast as possible. Some people wanna browse. They wanna hang out, they really, they have an affinity for the brand. They wanna stay in the store. They wanna see what's new, they wanna talk to, to someone to explain more about it.
[00:30:48] Etie Hertz: Um, so people have kind of a different journey depending on what they're looking for. And you can get some of those signals. You can get a lot of those signals online by watching what they're doing within the site. Are they actually trying [00:31:00] to get something but they're unable to and they're rage clicking 'cause they can't figure out how to do it.
[00:31:04] Etie Hertz: Are they navigating to different pages? Are they looking at certain. Products and then other products that are similar. Like there's a lot of things that you can pick up from the digital experience, but again, if you can marry that with the conversations on the, on the support side, that kind of lights up all those interactions to a different level.
[00:31:19] Etie Hertz: We really have a better understanding. We haven't even spoken about the fact that commerce now is gonna be become conversational. You're gonna stop using your finger and dragging it. You're gonna be typing, you're gonna be chatting into brands on their websites. You're gonna have like a chat GBT like experience on most brand sites.
[00:31:33] Etie Hertz: So that conversation is gonna be much richer than it's been historically, and you'll have a brand will have even more data to understand how people behave on the site, what they're looking to achieve. And not everyone wants the exact same thing. So do you wanna treat different customers differently? And then back to what we said earlier, if you, in, if you got to the point when you edit LTV, scribe to every single interaction, I know that Lauren is really valuable.
[00:31:54] Etie Hertz: I can spend an hour on the phone with her because she spends a lot of money with me. Someone else is never gonna buy. They come in [00:32:00] and out and I, I can't waste hours on the phone with this person. Or they're sending their AI agents at me. Right.
[00:32:05] Lauren Goerz: I'm curious what this means for brands, because you have a, you know, you have the broad over overarching perspective on this.
[00:32:10] Lauren Goerz: Um, so one of the things that we're seeing as kind of a net new use case that people are starting to wanna do is exactly what you described, this kind of product search use case. Something that we really couldn't do well was very deterministic. Kind of like retrieve rank, show the different products.
[00:32:24] Lauren Goerz: There's probably a lot of opportunity there from a business perspective to optimize that, to make that better, to get the most out of it. What does that, you know, what are, what are the opportunities there for businesses in this kind of product? Conversational product, search, product recommendation, experience.
[00:32:37] Etie Hertz: It's like a whole new science. I mean, if, if you expect, if you, if you assume someone's just coming in, they're looking for a widget and they want to get out, then okay, then you have to be really good at serving up the widget as fast and seamlessly as possible. But most folks are gonna say, well, if, if you're looking at this, maybe you wanna look at other things.
[00:32:53] Etie Hertz: Maybe you wanna add more things to your cart. Maybe I wanna do certain things to strengthen my relationship as with the brand and the user [00:33:00] within that conversation. There's so many different things. Again, analogize to an in-store experience. There's a lot of things that you could do. You could let someone run around by themselves and never talk to them.
[00:33:08] Etie Hertz: You can offer to help. You can ask them if they wanna look at similar products. Um, that's gonna be. Pretty analogous to what you're gonna find online soon. When folks have a better understanding of what people are looking for, how much patience do they have, how much are they willing to stay on a website or on an app to interact?
[00:33:27] Etie Hertz: Um, yeah, all of it's gonna change.
[00:33:30] Lauren Goerz: I guess are there, are there any other key behavioral signals? Um, are there any other key behavioral signals that you've noticed? So for example, like rage clicking, I'm assuming this is just someone that's clicked really fast. Maybe I've done it, I'm not even sure if I have, I've never heard of it before.
[00:33:43] Lauren Goerz: What are some of the other like clickable readable signals that you get from users? Is there moving about in a webpage,
[00:33:50] Etie Hertz: you have journey maps. So we we're able to basically track where someone comes in and how they leave and, um, if, again, if the end game is for you to buy this. Item, there's a [00:34:00] certain journey that that might take typically.
[00:34:02] Etie Hertz: And if you fall, maybe something's wrong in the site. And the app experience where it breaks. Yeah. Where you're looking for something but you can't find it. And then you're running around and you're getting diverted to other pages. So there's something that tells us about the experience that you're having from a digital side that we should improve to make it easier for you to find the thing you're looking for and get out.
[00:34:19] Etie Hertz: Hmm. Um, there are a variety of different things that we can observe online with all of these different data points. Um. To understand if the, if the outcome is, is what we were looking for.
[00:34:31] Lauren Goerz: I guess. So it's, it's, I think you made a good point there. We've talked a lot about personalization, but actually optimization is the other side of this.
[00:34:37] Lauren Goerz: How can you optimize your experience? What's some of the things that you've learned from conversations that have helped you optimize experiences elsewhere?
[00:34:48] Etie Hertz: I mean, the easiest use case is people who are coming in, they wanna buy something. Let's make it easy for them to do that. Or they wanna return something, let's make it easy for them to do that. So how do we, how do we make that journey the shortest path between two points as opposed to them running around and [00:35:00] trying to figure it out?
[00:35:01] Etie Hertz: Um, but it's super nuanced. I mean, there's a lot of companies, for example, that have a support team that make it really hard for you to cancel your subscription or return an item that's part of their like. And, you know, uh, way of doing business. Other people want customer service to be really, really excellent and they wanna make it super easy for you to get the thing that you're, that you're looking for.
[00:35:20] Etie Hertz: Same is true on the conversion side. Some brands want you to hang out. I want you to spend more time, I want you to have an affinity for the brand. Um, maybe I don't want you to find the exact same, the exact thing that you're looking for right away and get out. I wanna show you other things along the way.
[00:35:33] Etie Hertz: Um, so it really depends on what the brand is solving for.
[00:35:36] Lauren Goerz: And
[00:35:36] Etie Hertz: there's a lot of things that go into that. Is your demographic young or old? Are they online or, or not? Um, everyone's got their own flavor of how they interact with their customers.
[00:35:45] Lauren Goerz: That actually gives me another question, which I think we haven't covered today, which is interesting because here at Raza we serve actually quite a wide variety of geographies.
[00:35:53] Lauren Goerz: Languages also. That's one of the things that Raza framework is really good at, is really serving different languages well. So I'd [00:36:00] be curious to learn from even like a cultural perspective. I live in Germany today, originally born in America, so I know there's, there's loads of different cultural perspectives and how I interact with people today here in Germany.
[00:36:09] Lauren Goerz: Then I would in California, nonetheless. I'd be curious, what have you learned from these conversations? What, what element does culture does, language does age, demographics have, and, and what are some of the key things, quick fixes you can make to make these experiences better if you do know a little bit about the person?
[00:36:27] Etie Hertz: So when we started, essentially our job was to help folks, agents around the world interact with the US clientele. Um, and they had studied, a lot of them have gotten training in, in terms of how do you interact with an American audience. So a lot of the suggestions and things that we would do to help them, they, they were all kind of in the same vein.
[00:36:46] Etie Hertz: Now we've added additional languages. Now Content Square is totally worldwide, and so that's, that's a little bit new for us. Now, we have to go expand in terms of, um, our understanding of different cultures, the language part's, not that it's basically a solved problem. The [00:37:00] cultural part online is not difficult, but in, in, in conversations it's more nuanced.
[00:37:05] Etie Hertz: So I, I would say that we're still, we're still learning that piece of it.
[00:37:10] Lauren Goerz: Okay. Super interesting. Yeah, would love to hear your learnings as as they come up in the next years for sure, because I think that's something we we're seeing also internally. We've got, sometimes what's happening now with language models is the center of excellence teams will pop up and as you said.
[00:37:24] Lauren Goerz: Translation is in many ways a bit of a solved problem here. So it makes it easy for, you know, a team in one country to serve 10 teams in other countries. So it's an interesting case where you, where you have one team that might not have the cultural awareness of all of these other countries, how do you actually design for that?
[00:37:39] Lauren Goerz: And I think that's an interesting case for all those conversation designers listening. For you to be involved, obviously, because having these local opinions, local identities, local understanding, I think is still massively important. But I think through data we're probably gonna learn a lot about ourselves as well.
[00:37:51] Lauren Goerz: And I guess that that'll be maybe one of my closing questions. I, I guess, what have you learned about humanity so far? A bit of a broader question. After watching all of these different [00:38:00] conversations back and forth, any any key learnings that you've had that actually surprised you over the years
[00:38:04] Etie Hertz: about humanity in general?
[00:38:06] Lauren Goerz: Sure. People talking.
[00:38:12] Etie Hertz: I think people that interact well, I think people are generally very nice and, and well intentioned, even if they're frustrated in the moment. So if they can see the other person on the other line as a human like themselves, then think the whole interaction is different. Um, it's much more respectful. You wouldn't, I, I, I don't think that I would've expected that at the beginning.
[00:38:34] Etie Hertz: I would think that people would just interact with a support agent and just, you know, like. I need you to fix this for me. Do it, do it quickly and, and whatever. And lemme get outta here. Um, people not only kind of appreciate the human aspect, a lot of people need that. We saw that a lot in COVID, for example.
[00:38:52] Etie Hertz: People were alone oftentimes not, not interacting with anybody else, they just wanted someone to talk to. They needed the human connection. And literally, like a lot of customer [00:39:00] support volume was just people trying to talk to other people. So that aspect of it was really, was really lovely. Um. And so I, I guess what I'm more interested and concerned about now with the proliferation of AI agents when they're all running around and you're not gonna even know if you're talking to an agent or a human, what is that gonna mean for humanity in society?
[00:39:19] Etie Hertz: Are we all gonna become colder and more empathetic? I hope not, but it's, it's definitely possible. It's, it's certainly a concern and you probably will kind of see the, um, the seeds of how humans. Engage with other humans changing in, in like in our world, because there's just so much volume and so many conversations and so many interactions tracked.
[00:39:43] Etie Hertz: We're gonna learn a lot about like the evolution of human interaction, like just from our day-to-day jobs, I think. So I'm hoping for the best, but I don't know what's gonna happen.
[00:39:51] Lauren Goerz: Amazing. Thank you so much for joining today. It's been an absolute pleasure to talk with you. I think there's probably, you know, a thousand different conversation designers that would love to [00:40:00] hear your insights about human real human conversations because ultimately, at the end of the day, that's exactly what we're trying to do here.
[00:40:05] Lauren Goerz: Have slightly more human conversations than we currently have with machines. So thank you so much. Appreciate your insights and wish you all the best and can't wait to hear about your learnings in the years to come.
[00:40:15] Etie Hertz: Thank you. Thanks so much for having me.