Latest / The Adults In The AI Room / Earley AI Podcast Ep 78: How AI Is Revolutionizing Customer Feedback and Engagement for Large Enterprises

Transcript

Earley AI Podcast Ep 78: How AI Is Revolutionizing Customer Feedback and Engagement for Large Enterprises · The Adults In The AI Room

• 0:01
  SPEAKER_00: This is the Early AI Podcast hosted by Seth Early.
• 0:04
  SPEAKER_00: Join us as we delve deep into the passions, expertise, and experiences of thought leaders and practitioners to talk about what's possible with artificial intelligence.
• 0:14
  SPEAKER_00: The Early AI Podcast is sponsored by Early Information Science for a digital transformation journey with design and deployment of innovative technology solutions, as well as vector.
• 0:26
  SPEAKER_00: Now enjoy the show.
• 0:28
  SPEAKER_01: Well, welcome to the Early AI Podcast.
• 0:31
  SPEAKER_01: My name is Seth Early.
• 0:32
  SPEAKER_01: I'm your host, and we are going to be exploring how artificial intelligence is reshaping the way organizations manage information, create value, deliver better customer experiences.
• 0:44
  SPEAKER_01: Today we're going to be talking about the intersection of customer engagement and customer feedback with AI.
• 0:51
  SPEAKER_01: And we're going to be talking about how the emerging technologies can be applied to delivering more meaningful interactions and being able to get better feedback from the marketplace about our customers, their experiences, and the competition.
• 1:08
  SPEAKER_01: What is the competition doing?
• 1:09
  SPEAKER_01: So joining me today is George Swetlitz.
• 1:12
  SPEAKER_01: He is CEO and co-founder of Write Response AI.
• 1:16
  SPEAKER_01: George has a lot of expertise in natural language technologies, enterprise AI adoption, and applying advanced models to solve real business problems.
• 1:24
  SPEAKER_01: His work is focuses on building systems that go beyond generating text to really understanding how customers are interacting, what kinds of feedback they're providing, and again, what that competitive landscape is looking like.
• 1:37
  SPEAKER_01: So, George, welcome to the show.
• 1:39
  SPEAKER_03: It's nice to be here, Set.
• 1:40
  SPEAKER_03: Thank you.
• 1:41
  SPEAKER_01: So let's talk a little bit about I'll want to get to exactly what you do and what your company does in a minute, but but from your uh your exposure in the marketplace and you're talking to organizations, what are the types of things that they're not quite understanding?
• 1:55
  SPEAKER_01: What are some of the big misconceptions that you're finding in the marketplace about AI and about how emerging technologies are impacting uh the customer experience and that whole feedback mechanism of uh getting that customer sentiment, understanding how they're interacting uh and what their experiences really are.
• 2:13
  SPEAKER_01: But what are people missing or what are they not getting?
• 2:16
  SPEAKER_03: Yeah, I think what people are, you know, people know what they know and they don't know what they don't know.
• 2:20
  SPEAKER_03: And so as they get exposed to things and as they get exposed to AI, they learn about what works and what doesn't work.
• 2:26
  SPEAKER_03: And what's happened in this space around customer experience is that people, leaders and organizations, don't really, they haven't really, they don't really understand what's the best way to leverage AI in the customer experience, customer sentiment space.
• 2:40
  SPEAKER_03: And they're focused on the kinds of things that they see AI do, which is sounding human, as opposed to what I counsel people, which is about being contextually relevant, right?
• 2:52
  SPEAKER_03: Using AI to be relevant to your customers as opposed to sounding human.
• 2:57
  SPEAKER_01: That's an interesting distinction.
• 2:58
  SPEAKER_01: And obviously sounding human is important, but it has to be relevant, right?
• 3:03
  SPEAKER_01: It doesn't matter if you're sounding human and uh and you and you're you're you're not addressing my problem or you're not uh giving me any insights, but sounding human is fine, but give me something that's relevant and meaningful.
• 3:14
  SPEAKER_01: Tell me about how you look at this marketplace and how you look at customer experience and give me a little bit of a sense of the kinds of things that you do uh in this environment, how you're leveraging AI and how you're impacting the customer experience.
• 3:27
  SPEAKER_01: It's a little bit of a uh company commercial, but not not too not too safe.
• 3:32
  SPEAKER_03: To build on what you just said though, you know, it's the flip side of sounding human is, you know, when you call a customer service agent and they read you back the script.
• 3:41
  SPEAKER_03: Well, they're human, but what you say is, well, they're robotic, they're robots, they can't help me.
• 3:46
  SPEAKER_03: And so what you're trying to do with AI is get the best of both worlds.
• 3:50
  SPEAKER_03: You're trying to be relevant to somebody in the space or in the place that they're in, right?
• 3:57
  SPEAKER_03: Kind of that's how I view the world.
• 3:59
  SPEAKER_03: And so, you know, what we do with AI is to do the things that large organizations need to be relevant to their customers.
• 4:08
  SPEAKER_03: So, what are large organizations looking for?
• 4:10
  SPEAKER_03: They're looking for efficiency, right?
• 4:13
  SPEAKER_03: You know, most large organizations have centralized services, and people struggle with centralization versus decentralization because the the loss of quality when you bring things to the center, and they also want revenue.
• 4:27
  SPEAKER_03: And so we they want revenue, right?
• 4:29
  SPEAKER_03: They want to build revenue.
• 4:30
  SPEAKER_01: Very important stuff.
• 4:30
  SPEAKER_01: Money is handy.
• 4:31
  SPEAKER_03: Very important stuff.
• 4:32
  SPEAKER_01: Yeah.
• 4:33
  SPEAKER_03: And so we try to help do those three things by leveraging the review ecosystem.
• 4:39
  SPEAKER_02: Right.
• 4:40
  SPEAKER_03: And so, and I can talk more about how we do that, but that's essentially what we're what we're doing for our larger organizations.
• 4:46
  SPEAKER_01: And so what is uh what is different?
• 4:48
  SPEAKER_01: Like because people have been doing sentiment analysis for a very long time.
• 4:52
  SPEAKER_01: Uh, they've used text analytics, they've you know used uh different mechanisms to kind of harvest uh, you know, data, social media data, and listen to customers in one way, shape, or form.
• 5:02
  SPEAKER_01: And you know, lots of different ways of getting that voice of the customer.
• 5:06
  SPEAKER_01: So, how is it different?
• 5:07
  SPEAKER_01: What is different now?
• 5:08
  unknown: Right.
• 5:09
  SPEAKER_03: Okay, so focusing it on sentiment analysis, AI in general has upped the game with sentiment analysis by allowing the phrasing to be looked at in context.
• 5:20
  SPEAKER_03: So, you know, when you have sarcasm or when you have various things like that, AI does a much better job of understanding that.
• 5:27
  SPEAKER_03: In a review, for example, if somebody said, Well, I went to, I went to this one restaurant and I was unhappy, and then I came here and was happy.
• 5:35
  SPEAKER_03: Typical natural engine processing doesn't do a great job of saying, well, that first part is really not about the company evaluating.
• 5:43
  SPEAKER_03: AI does a really good job of that.
• 5:45
  SPEAKER_03: So when we use AI to look at sentiments, we pull in the review, we break it down into phrases, we determine, and this is through a series of agentic steps, we break it down into phrases, we determine which of those phrases have to do with the customer's experience with the business itself, and then we map it to a set of topics that the customer uses as their KPIs.
• 6:08
  SPEAKER_03: So it's a much more granular approach.
• 6:10
  SPEAKER_03: And not to say that other people couldn't do that, but we do it as part of a larger review ecosystem.
• 6:16
  SPEAKER_03: And so that we then leverage those findings in other places in our process.
• 6:22
  SPEAKER_03: So we don't just use it in sentiment analysis, but we use it when we're generating a response.
• 6:27
  SPEAKER_03: We use it when we're doing a monthly summary, we we use it in a lot of different ways.
• 6:31
  SPEAKER_01: And one of the things that um we talked about uh is that, you know, a lot of a lot of times uh if we're asking AI or generative AI, you know, to answer a question, it's answering that question based on its model of the world, its understanding of, you know, uh interactions and relationships and you know, all the data that it's been programmed on.
• 6:54
  SPEAKER_01: And it's very, very impressive and it's very uh important, but it doesn't necessarily know the details behind what you sell or your value proposition or your challenges and so on, or what the answers are, right?
• 7:06
  SPEAKER_01: So it can't necessarily contextualize a very specific response.
• 7:10
  SPEAKER_01: So, how do you do this?
• 7:11
  SPEAKER_01: And we talked a little bit about retrieval augmented generation, and you were saying you didn't necessarily use RAG, but you did use some mechanism for building repositories of facts and data and information about the organization.
• 7:22
  SPEAKER_01: So you want to talk a little bit about how you contextualize the AI in company-specific information.
• 7:29
  SPEAKER_03: Right.
• 7:30
  SPEAKER_03: So if you if your objective is to sound human, then what you want is the best, most human-like sounding model.
• 7:37
  SPEAKER_03: And as you said, what you see today in review responses that people that that consumers are reacting negatively to is essentially just the repetition of the review and the response.
• 7:50
  SPEAKER_03: And in a sense, what we're hearing from consumers now is that this is really irritating because you're just taking up space on the page.
• 7:58
  SPEAKER_03: You're adding no value to their understanding of the business, you're just taking up space.
• 8:02
  SPEAKER_03: And so what we do is is we so when somebody on boards with us, we do a series of things automatically.
• 8:13
  SPEAKER_03: We read their entire history of reviews and we determine what people talk about.
• 8:19
  SPEAKER_03: What are the things that people talk about that are unique to this business?
• 8:22
  SPEAKER_03: And we kind of set those aside.
• 8:24
  SPEAKER_03: Then we read the responses to those reviews and we read the website and we determine are there things in the responses on the website that would be helpful and useful to a customer to know if they were reading that review and listening for a response.
• 8:42
  SPEAKER_03: And we create this fact, we call them facts, we create this library of facts that's a very rich view of that company's kind of environment.
• 8:53
  SPEAKER_03: So in many cases, the company doesn't have that information.
• 8:58
  SPEAKER_03: There's some things on the website, there's some things in the responses, but most people today don't respond with any kind of depth to reviews.
• 9:06
  SPEAKER_03: And so we craft a response that we think might be relevant.
• 9:10
  SPEAKER_03: Now, then what the company has to do is go in and review those facts and make sure that they're actually valid.
• 9:16
  SPEAKER_03: But we did the hard work of determining and kind of creating a draft.
• 9:21
  SPEAKER_03: From from that point forward, every review that comes in gets analyzed in the context of those facts.
• 9:28
  SPEAKER_03: And there may be one fact, three facts, zero facts, any number of facts that are relevant to that review.
• 9:34
  SPEAKER_03: And when the response is generated, those facts are incorporated into the response itself.
• 9:40
  SPEAKER_03: So they so they become, rather than a recitation of the review, they become contextually relevant.
• 9:46
  SPEAKER_01: It's a more meaningful answer response to a question that someone has.
• 9:50
  SPEAKER_01: And you're not just repeating back their question and then some uh potentially uh less relevant response.
• 9:57
  SPEAKER_01: You're really trying to fine-tune that response to this.
• 10:00
  SPEAKER_01: So you're really doing a personalized presentation of information that's contextually relevant to them and the particular question and the particular problem.
• 10:08
  SPEAKER_01: So you're able to now you're able to, and of course, a good customer service rep would do that, right?
• 10:13
  SPEAKER_01: But uh give us some examples of kind of the scale or the volume at which organizations are facing this.
• 10:20
  SPEAKER_01: You know, you could have smaller businesses that are that are, you know, regional or local and they have their challenges.
• 10:26
  SPEAKER_01: You can have businesses with multiple locations, you can have large businesses that deal with a huge amount of responses.
• 10:32
  SPEAKER_01: Give us some sense of um what that landscape looks like in terms of the number of responses that organizations are dealing with, how they dealt with it before this type of approach, and what some of the other attempts at automation have had kind of where they've kind of had gaps or did not necessarily uh meet the expectations of the customer or the organization.
• 10:53
  SPEAKER_01: So give us the landscape that works.
• 10:55
  SPEAKER_03: Yeah, so let's focus on large organizations.
• 10:57
  SPEAKER_03: So large organizations, the problem with large organizations, I used to run a large organization, the problem with large organizations is the bigger you are, the less money you actually want to, you know, as a CEO, you're sitting there going, I don't want 10 people sitting in a room responding to reviews.
• 11:10
  SPEAKER_03: It's not value add.
• 11:12
  SPEAKER_03: And they're 100% right.
• 11:13
  SPEAKER_03: And so the problem is that when you when you when you have a large organization, because whether it's a templative response or generic AI isn't helpful, what you have is people having to look at each of those responses, modify them, and that's time that's not being spent helping customers who actually had a negative experience and need hand holding.
• 11:37
  SPEAKER_03: So what we find is that larger organizations that we work with end up redeploying about two-thirds of their spend.
• 11:46
  SPEAKER_03: So they're able to take people that were, so they might have a thousand reviews a week and they're spending their time responding to those thousand.
• 11:55
  SPEAKER_03: And so they have three FTEs, and they may be spending, you know, a half of an FTE focused on working with customers who have had a negative experience.
• 12:05
  SPEAKER_03: What we find is that companies are able to redeploy so that they're spending two and a half FTEs working with the customers that need help, and only a half of an FTE, in that case, kind of you know, working on reviews or response for a negative client.
• 12:23
  SPEAKER_03: Because very you know, often the negative reviews require some knowledge about the business.
• 12:29
  SPEAKER_03: No AI is going to be able to deal with those.
• 12:32
  SPEAKER_03: You don't want AI to deal with those.
• 12:33
  SPEAKER_03: And so someone has to focus on that, write the response, figure out who that customer is, call them up, and try to solve the problem.
• 12:40
  SPEAKER_03: And if you can have two and a half out of three versus a half out of three working on that, you end up with a much better customer experience.
• 12:47
  SPEAKER_03: And that's where people really find the benefit.
• 12:50
  SPEAKER_01: So one could argue that the purpose of grounding this in facts and your facts about the organization and using this type of approach to really understand what that customer interaction is, what problem they're trying to solve would be in the realm of the of the AI.
• 13:05
  SPEAKER_01: But you're saying that there's a lot of edge cases that really defy that automation and that AI approach because they do have a lot of complexity to them.
• 13:14
  SPEAKER_01: There are multiple factors.
• 13:16
  SPEAKER_01: You really do need a human being involved in that, right?
• 13:19
  SPEAKER_01: And what you're suggesting you're saying is that we're going to take away the things that can be automated, the lower hanging fruit, and we're going to focus on that higher value interactions and those high and those more complex problems and more of those edge conditions.
• 13:32
  SPEAKER_01: But can you talk about uh kind of an anecdote of what a come what an organization, and again, I'm not trying to make this about your company, but I'm trying to make it about what's a tangible kind of business case or where are people seeing value?
• 13:44
  SPEAKER_01: So can you talk about kind of a a before and after with a large enterprise?
• 13:49
  SPEAKER_01: What was this, the volume they were receiving, what were the challenges, what were they spending?
• 13:53
  SPEAKER_01: And then, you know, after they were able to successfully automate this, build their fact base, which sounds like a knowledge base to me, uh, and it does sound like a retrieval, but that's okay.
• 14:02
  SPEAKER_01: We won't have to get into too many uh uh semantics.
• 14:06
  SPEAKER_01: But tell me a little bit about what that before and after looked like.
• 14:08
  SPEAKER_01: How how was that impacting the business?
• 14:10
  SPEAKER_03: Yeah, so the what I just talked about is kind of the outlines of that, a thousand reviews a week, right?
• 14:16
  SPEAKER_03: Is kind of a kind of a what was the scale of it, with essentially three people focused on this full time, focused mostly on dealing with the reviews and the responses as opposed to helping customers.
• 14:28
  SPEAKER_03: In that particular case, they were able to redeploy a half a person, focused two people on dealing with customer, solving customer problems, which led to customers changing their ratings, right?
• 14:41
  SPEAKER_03: So one of the benefits is if you have a negative review and you solve the customer's problem, they go in and they change their rating.
• 14:48
  SPEAKER_03: So that now improves your overall rating as well as having a happy customer, customer attention.
• 14:54
  SPEAKER_01: So you improved your your CSAT scores, and uh, did you actually reduce the costs or the expense, or you simply redeployed people to higher value uh work?
• 15:05
  SPEAKER_01: Right.
• 15:05
  SPEAKER_03: So there was a small cost reduction.
• 15:08
  SPEAKER_03: In most of the company, most of the large organizations that we deal with, there's so few resources in this area anyway that they're happy to redeploy them on actual customer issues as opposed to capturing a cost savings.
• 15:20
  SPEAKER_03: But the flip side to that, the flip side to that is the revenue growth.
• 15:24
  SPEAKER_03: And so when you make responses contextually relevant, right?
• 15:29
  SPEAKER_03: So let me step back.
• 15:30
  SPEAKER_03: No one's closer to the bottom of the funnel than someone who's reading reviews.
• 15:35
  SPEAKER_03: They are making a decision, they're reading and they're trying to decide what to do.
• 15:39
  SPEAKER_03: And and I I know, you know, and your listeners will have experienced this themselves when they read reviews.
• 15:46
  SPEAKER_03: You read down the page, you're looking at the reviews, you're glancing at the response from the company.
• 15:51
  SPEAKER_03: And if you see something in that response that's relevant, you actually stop.
• 15:56
  SPEAKER_03: You're not expecting it.
• 15:57
  SPEAKER_03: You stop and you read it and you go, that's interesting.
• 16:00
  SPEAKER_03: And so what happens is you start to think this company actually cares.
• 16:04
  SPEAKER_03: They're they're providing something other than garbage in this response.
• 16:08
  SPEAKER_03: I'm gonna pay more attention.
• 16:10
  SPEAKER_03: So they start reading down the page and they see it.
• 16:12
  SPEAKER_03: And now you have a higher likelihood of being selected by that customer.
• 16:18
  SPEAKER_03: So we've seen revenue.
• 16:19
  SPEAKER_03: Now it's always hard to see, but you can see it in terms of kind of click-through rates to call click to calls and things.
• 16:26
  SPEAKER_03: When those percentages go up, it's essentially kind of confirming that that that what you used in your response is attracting more people.
• 16:35
  SPEAKER_03: So there are kind of, you know, there are things that you can look at as a proxy for revenue growth.
• 16:41
  SPEAKER_01: And uh any any data on that that you can share in terms of an increase in the click-throughs or the increase.
• 16:47
  SPEAKER_01: I mean, what what is what are what are you, or even just talk about the framework.
• 16:51
  SPEAKER_01: So what are you baselining and then what are you trying to measure against again for these proxies?
• 16:55
  SPEAKER_03: Yeah, yeah.
• 16:56
  SPEAKER_03: So the baseline is kind of their click-through rate of four that you used us because you know, generally speaking, you know, not a lot of people are trying to be contextually relevant in their responses.
• 17:05
  SPEAKER_03: And what we see is increases of around 15 to 20 percent in the click-through rate.
• 17:10
  SPEAKER_03: And so that's you know, that's pretty substantial because, you know, and it reflects the engagement in that in that review.
• 17:17
  SPEAKER_01: Okay.
• 17:18
  SPEAKER_01: And and so is there are there ways to kind of measure the the quality of these interactions?
• 17:25
  SPEAKER_01: Is there any kind of objective measure to say, and again, this is a difficult thing, you're trying to put a quantitative measure to a qualitative problem, right?
• 17:33
  SPEAKER_01: Uh, are there ways to measure the quality of those responses?
• 17:37
  SPEAKER_01: I guess, uh I guess the people, their actions, their click-throughs would do that.
• 17:41
  SPEAKER_01: Are there other objective measures uh to look at those?
• 17:44
  SPEAKER_01: If you had a thousand uh reviews and and comments uh for company A and a thousand reviews and comments of company B, are there uh objective ways to measure uh the relevance?
• 17:57
  SPEAKER_01: Not looking at click-through reads, but just looking at the at the text.
• 18:01
  SPEAKER_01: Have you tried using uh uh large language models to do that analysis or to kind of have that baseline objective measure of quality response?
• 18:11
  SPEAKER_03: Yeah.
• 18:11
  SPEAKER_03: So in a sense, we get at that when we do that analysis that I was talking about, when we read the thousand reviews and the thousand responses to see whether the existing responses had any information in them that would be relevant to the review.
• 18:24
  SPEAKER_03: And more often than not, there's nothing.
• 18:26
  SPEAKER_03: The AI that we're that we're using, the the models that they're using are not identifying useful information in responses.
• 18:36
  SPEAKER_01: And talk a little bit about these fact uh repositories, these knowledge bases, or whatever you refer to them as, it's a repository fact.
• 18:45
  SPEAKER_01: So what you're looking for is what are the things people are asking about in their reviews, or what are they complaining about?
• 18:50
  SPEAKER_01: What are the problems?
• 18:51
  SPEAKER_01: And then where are the solutions?
• 18:53
  SPEAKER_01: Where are the answers to that?
• 18:54
  SPEAKER_01: If somebody says, Well, I couldn't find any parking, and on the website, there's a place, well, here's our our overflow parking, it's across the street, but they didn't see that, they didn't know that, right?
• 19:04
  SPEAKER_01: So you're in some ways, you're surfacing the information that may already be present somewhere.
• 19:09
  SPEAKER_01: In other cases, you're backstoring that information, right?
• 19:12
  SPEAKER_01: You're saying, oh, wow, here's a real knowledge gap or information gap.
• 19:15
  SPEAKER_01: People don't know this piece of information about our organization.
• 19:19
  SPEAKER_01: Now we're going to capture that, codify it, and make it readily available to respond to that, but also to go back into other touch points.
• 19:26
  SPEAKER_01: So uh so how large are those fact bases?
• 19:29
  SPEAKER_01: Like what do they consist of?
• 19:31
  SPEAKER_01: Can you give us a sense of of um what it takes to create them and what they look like and the number of facts that you would have in something like this?
• 19:39
  SPEAKER_03: Yeah, I mean, a typical large organization will have somewhere between 20 and 30 facts.
• 19:44
  SPEAKER_03: People tend to talk about the same thing.
• 19:49
  SPEAKER_01: 20 to 30, 20 or 30,000.
• 19:51
  SPEAKER_03: No, no, no, no, no.
• 19:52
  SPEAKER_03: I mean, you know, when you think about reading reviews, people don't talk about 30,000 different things.
• 19:57
  SPEAKER_03: They tend to talk about, you know, the the most of the reviews tend to talk about, you know, 20 to 30 different topics, whether it be in a restaurant or, you know, a you know, automobile business.
• 20:10
  SPEAKER_03: And that's an interesting point.
• 20:11
  SPEAKER_03: And we'll talk about we can talk about, we can switch in a second about review requests and kind of where AI can be used in review requests as well.
• 20:20
  SPEAKER_03: But that's what you see.
• 20:21
  SPEAKER_03: You see, you tend to see these things.
• 20:23
  SPEAKER_03: You also tend to see, unlike in review responses, you do tend to see kind of content or tip for those things in websites.
• 20:32
  SPEAKER_03: People spend a lot of money on their websites, they spend a lot of money kind of building those out and making them useful.
• 20:38
  SPEAKER_03: What's happening, and it's even more so with kind of AI overviews and things like that, fewer and fewer people are going to websites.
• 20:44
  SPEAKER_03: So you have to be where they are.
• 20:48
  SPEAKER_03: And people are reading reviews.
• 20:49
  SPEAKER_03: And even today, AI overviews are reading reviews and responses.
• 20:53
  SPEAKER_03: And so you have to be where the world's going.
• 20:57
  SPEAKER_03: And the world is not necessarily going to the website.
• 21:00
  SPEAKER_03: And so that's why we we try to, you know, we analyze the website because of that reason.
• 21:06
  SPEAKER_03: We're trying to make the point, we're trying to kind of do that light bulb thing where you say, you have all of this tremendous relevant stuff on the website, none of it's making it into this review response.
• 21:17
  SPEAKER_03: Why not?
• 21:17
  SPEAKER_03: And the reason is if you're a large organization, you have a small team of people responding to reviews, could you really expect them to incorporate all this website content into the review response?
• 21:27
  SPEAKER_03: I can tell you as a former, you know, CEO of a business, I would never expect that.
• 21:31
  SPEAKER_03: It's just too difficult to expect that with turnover and all these kinds of things that are happening in centralized services to expect that.
• 21:39
  SPEAKER_03: And AI is really good at it.
• 21:41
  SPEAKER_01: You had mentioned a minute ago just uh talking about review requests.
• 21:44
  SPEAKER_01: So talk about AI and review requests.
• 21:47
  SPEAKER_01: Right.
• 21:47
  SPEAKER_03: So the problem with review requests, it's kind of the it's a other side of the coin.
• 21:52
  SPEAKER_03: When you you know lots of organizations send out review requests, everybody gets them, everybody gets a lot of them.
• 21:58
  SPEAKER_03: Anytime you engage With someone and it's the same problem, right?
• 22:03
  SPEAKER_03: The problem is that those reviews don't connect with you as a person, they're very robotic requests.
• 22:09
  SPEAKER_03: And so to the extent that you can do a couple of things, one is that you can leverage information that you have in your company's databases about that person and bring that information to bear in that review request.
• 22:24
  SPEAKER_03: Make it more personal, make it more emotional.
• 22:28
  SPEAKER_03: The re the request conversion to review rate goes up pretty dramatically, 30, 40%.
• 22:35
  SPEAKER_03: And it can be, it can be because of the personalized text, it can be the inclusion of a photograph.
• 22:40
  SPEAKER_03: It can be the inclusion, and this is a different point, of SEO-related questions.
• 22:46
  SPEAKER_03: So based on what you know, what's a question that that customer could answer and write a better review for the benefit of all the review readers.
• 22:58
  SPEAKER_01: Can you give us an example like what a what a um kind of a generic request would be and then what one of these enhanced requests would be?
• 23:07
  SPEAKER_03: Absolutely.
• 23:07
  SPEAKER_03: So it's a personal injury law firm, great example.
• 23:10
  SPEAKER_03: They know a lot about their customers.
• 23:12
  SPEAKER_03: They know what type of accident that person came to them about.
• 23:15
  SPEAKER_03: They know whether they got a settlement or whether they won a court case.
• 23:19
  SPEAKER_03: They know when they did this work for the customer.
• 23:22
  SPEAKER_03: So a typical review request is going to say, hey, Bob, you know, it was our honor to work with you.
• 23:29
  SPEAKER_03: It would be very helpful for you if you get for us if you left her a response or a review.
• 23:33
  SPEAKER_03: With that information, the request can be, hey, Bob, it was it was an honor working with you on your motorcycle accident case.
• 23:40
  SPEAKER_03: And it was wonderful that we're able to get the negotiated settlement that we did.
• 23:45
  SPEAKER_03: It's been two months since we worked together.
• 23:48
  SPEAKER_03: And I thought I would ask for a review.
• 23:50
  SPEAKER_03: That helps us get the word out and would be very meaningful to a lot of other people.
• 23:54
  SPEAKER_03: That connects with you in a way that another that a generic request wouldn't.
• 24:01
  SPEAKER_01: And so how is that automated at scale?
• 24:05
  SPEAKER_01: Because I'd imagine that there could be uh some uh you know uh borderline conditions where you you don't want to use too much personal information, right?
• 24:15
  SPEAKER_01: It's that that creepy factor where you're being too personal versus not personal at all.
• 24:21
  SPEAKER_01: And and uh so your systems would have to interact with the customer relationship management systems that would have to gather some of that customer intelligence and that relationship knowledge in order to do that.
• 24:33
  SPEAKER_01: So, how is that actually done?
• 24:35
  SPEAKER_01: And then what are the guardrails around something like that?
• 24:38
  SPEAKER_03: Yeah, so generally there's a human in the loop in those kinds of things.
• 24:41
  SPEAKER_03: It's typically, you know, it's it's someone, you know, it's not a paralegal that's working on cases and doing those kind of administrative tasks, will, you know, in their in their database, you know, they've decided that these three or four fields are okay.
• 24:58
  SPEAKER_03: So they've made the decision that those fields are okay.
• 25:00
  SPEAKER_03: And then the paralegal will, they know the interaction with the client, they know whether it's relevant, you know, whether that's something that they want to actually ask that client.
• 25:08
  SPEAKER_03: And so when they're closing out the case, they'll either mark that as a yes or a no, and that request will either will or will not go out.
• 25:14
  SPEAKER_01: Yeah, even if you're simply beginning with the you're you're you're you're you're eliminating a lot of steps, manual steps to the process when you're giving someone to react to rather than say, write this from from scratch, right?
• 25:26
  SPEAKER_01: It's like using it, it's like a template on steroids, right?
• 25:29
  SPEAKER_01: You're saying, right, here's what the structure is, here are the facts that I'm gonna be pulling from our other repositories.
• 25:34
  SPEAKER_01: Now, a reviewer, just give this an okay or give it some edits.
• 25:38
  SPEAKER_01: And then so, so how would that translate to, you know, a very large organization that is dealing with, I mean, you know, I I'm not this is not my area of expertise, so I'm not familiar with how often larger enterprises are asking for reviews.
• 25:53
  SPEAKER_01: Is that more of a small to mid-sized business dynamic, or are you seeing that with uh enterprise customers as well?
• 26:00
  SPEAKER_03: We see it with enterprise customers as well.
• 26:02
  SPEAKER_03: They tend to stay a little bit further away from things that could be at the boundaries.
• 26:07
  SPEAKER_03: But again, if it's a large organization, they know what product was purchased, they know whether the customer's been a loyal customer, things that are not sensitive at all, that can be incorporated in a more personalized message.
• 26:21
  SPEAKER_03: And in addition, in those cases, they may lean more heavily on the questions that are asked that help the customer figure out what to write.
• 26:31
  SPEAKER_03: So that notion of connecting the questions with the customer allows that review to have more, you know, to have more resonance to readers, but also to AI.
• 26:44
  SPEAKER_03: So if AI is trying to understand what do you do, and reviews are talking, let's go back to the personal injury lawyer, and reviews are talking about motorcycle accidents and slips and falls, then now AI knows this firm does motorcycle accidents and slips and falls.
• 26:58
  SPEAKER_03: And somebody types in, I need an attorney for a slip and fall, they now know that.
• 27:02
  SPEAKER_01: Do you have other examples that would be more because again, I can see how that would work for that kind of a uh law firm which is fairly specialized.
• 27:11
  SPEAKER_01: Any other examples for other types of products and services that would that a large enterprise might yeah?
• 27:17
  SPEAKER_03: We have you know, like uh property managers, we have large property managers that you know are signing leases and you know are hosting events and all of these things can lead to review requests that hit an emotional trigger with that prospective tenant or that tenant around a kind of an emotional decision to rent in a particular facility.
• 27:39
  SPEAKER_01: Okay, gotcha.
• 27:40
  SPEAKER_01: Now, you would also, and we in our preparation uh discussion, you would also mention uh using this approach for a competitive uh analysis.
• 27:50
  SPEAKER_01: Do you want to just talk about uh how that uh dynamic works and and what the uh approaches are for uh understanding the competition?
• 27:58
  SPEAKER_01: Again, one of the things we had talked about is that there are going to always be regional differences.
• 28:03
  SPEAKER_01: So let's let's talk about sort of the competitive landscape and how that how how this kind of approach can be used to understand more about the competition and ways to be more uh uh you know competitive.
• 28:15
  SPEAKER_03: Right.
• 28:15
  SPEAKER_03: So as you just said, every location of a location-based business operates in its own unique competitive environment, right?
• 28:24
  SPEAKER_03: And and so what we find, what we see our larger organizations doing is focusing in on the 20% of their businesses or their locations they're having a problem with, because there's some, you know, there's some work that is involved in interpreting results, right?
• 28:39
  SPEAKER_03: So you can't do this kind of thing across every business, right?
• 28:42
  SPEAKER_03: You want to focus in on where you actually need or where you can get the greatest bang for the buck.
• 28:47
  SPEAKER_03: So they identify the problem children.
• 28:49
  SPEAKER_01: Those are usually the problem children.
• 28:51
  SPEAKER_03: So everybody has them, everybody has them.
• 28:53
  SPEAKER_03: And so that's what you want to focus in on.
• 28:55
  SPEAKER_03: And so you're trying to understand in that case, like what is it about this location that's causing a problem?
• 29:01
  SPEAKER_03: Is it us or is it something else?
• 29:04
  SPEAKER_03: Is it a competitor that's eating our lunch?
• 29:07
  SPEAKER_03: And so these are things you don't know.
• 29:08
  SPEAKER_03: So you can have a sense by looking at the sentiment analysis of your location as to whether or not those locations are underperforming the average, right, around your KPIs.
• 29:19
  SPEAKER_03: So that's one clue that you have just from doing your own competitive analysis.
• 29:23
  SPEAKER_03: Then what we have is the ability to go in and identify your competitors in that area and and and start a competitive analysis process where we go in and we download all the reviews and we analyze those reviews in the same way that we analyze the company's business.
• 29:43
  SPEAKER_03: So now we know a couple of things.
• 29:45
  SPEAKER_03: We know how many reviews are they getting a month.
• 29:48
  SPEAKER_03: When you go to Google and it says this company has a thousand reviews and an average rating of 4.8, well, are those thousand reviews over the last year or the last 10 years?
• 29:58
  SPEAKER_03: Like nobody knows.
• 29:59
  SPEAKER_03: So we go in and we kind of calculate that.
• 30:01
  SPEAKER_03: Why does that matter?
• 30:02
  SPEAKER_03: Well, if they're getting 20 a month, 20 reviews a month, and you're getting five reviews a month, you're gonna have a really hard time ranking in Google on a search.
• 30:14
  SPEAKER_03: They're gonna beat you every time.
• 30:16
  SPEAKER_03: So if you're spending money on advertising dollars and you're bringing people and they go then go to Google and they see somebody else who's doing better, you're just paying for somebody else to get the business.
• 30:26
  SPEAKER_03: So you got to understand, is it my review cadence that's a problem?
• 30:31
  SPEAKER_03: Or you can then go look at the sentiment analysis and you can look at your KPIs and you can say, are they outperforming me on service?
• 30:40
  SPEAKER_03: Are they outperforming me on product quality?
• 30:42
  SPEAKER_03: Are they outperforming where are they outperforming?
• 30:46
  SPEAKER_03: Where am I outperforming them?
• 30:48
  SPEAKER_03: Where's my problem?
• 30:49
  SPEAKER_03: And in some cases, it might be the review cadence, in other cases, it might just be how you're delivering your products and services versus a competitor in your marketplace.
• 31:01
  SPEAKER_03: But you you if you don't have that information, all you're doing is yelling at the manager, do better.
• 31:07
  SPEAKER_03: Why are you so terrible?
• 31:09
  SPEAKER_03: What are you doing about it?
• 31:10
  SPEAKER_03: As opposed to saying, look, there's two things you need to do.
• 31:13
  SPEAKER_03: You're processing 40 customers a month and you're getting five reviews.
• 31:17
  SPEAKER_03: You need to get 15 a month on reviews.
• 31:19
  SPEAKER_03: Set that as a target.
• 31:21
  SPEAKER_03: And you're losing against your competitors on their on their perception of service value.
• 31:28
  SPEAKER_03: So we need to talk about how our successful businesses are positioning that so you can do a better job.
• 31:34
  SPEAKER_01: Yeah, go ahead.
• 31:34
  SPEAKER_01: Yeah.
• 31:34
  SPEAKER_03: No, no.
• 31:35
  SPEAKER_03: So that's you know, as a CEO, that's what I used to struggle with.
• 31:40
  SPEAKER_03: How do I get the right information at the right time to the people running that business who care, they care about the business, but they don't, they're busy working with customers every day.
• 31:54
  SPEAKER_03: They don't have time to go figure this out.
• 31:56
  SPEAKER_02: Sure.
• 31:57
  SPEAKER_03: And so if we can figure it out for them and AI provides this amazing ability to do that, we're now providing a service, a valuable service to them.
• 32:07
  SPEAKER_01: So I think uh the the the kind of the takeaway is that you have um uh businesses that deal with customers on a local level, right?
• 32:17
  SPEAKER_01: I mean, again, there's lots of businesses that kind of deal with the uh uh the landscape of customers and they deal with them kind of more homogeneously.
• 32:26
  SPEAKER_01: But then many times the service delivery is on the ground, it's in a local area, you're dealing with people, you know, on a day-to-day basis, you're interacting with them, and that experience is so important to understand at a very granular level and to be able to quantify and to identify the topics and the issues that they're concerned with, that they're having problems with.
• 32:47
  SPEAKER_01: You're comparing that with the things that you're communicating uh to the client, uh to customers around, you know, via vis-a-vis the website or email communications or onboarding processes or whatever it might be.
• 33:00
  SPEAKER_01: You have this whole landscape of uh information that you're giving them uh throughout their lifetime, right?
• 33:06
  SPEAKER_01: So, and then what you're doing is you're saying, well, let's really understand uh the nuances of the problems and perceptions of those customers at a hyper-local level.
• 33:18
  SPEAKER_01: And then let's also look at the competitive landscape and say, well, how is this compared?
• 33:23
  SPEAKER_01: Are other customers dealing with the same problems with other vendors, right, or other suppliers?
• 33:29
  SPEAKER_01: Are they are other competitors solving these problems more effectively?
• 33:33
  SPEAKER_01: Are they are they addressing things that we're not addressing, right?
• 33:37
  SPEAKER_01: So you're really getting that kind of understanding of what the gaps are and that uh and the and the differentiators at that hyper-local level.
• 33:44
  SPEAKER_01: So that's interesting in terms of understanding that competitive landscape, understanding how you're being perceived in the market, and then doing that across, you know, hundreds or thousands of local uh uh instantiations, right, localizations of your customer interactions.
• 34:01
  SPEAKER_03: Right.
• 34:02
  SPEAKER_03: That's exactly right.
• 34:03
  SPEAKER_03: And it's all about, you know, and we we have customers that are you know, we were talking about local-based businesses, location-based businesses, and most of our business is location-based businesses.
• 34:12
  SPEAKER_03: But we have, for example, a ferry booking service in Europe.
• 34:15
  SPEAKER_03: That's one of our, you know, that's one of the largest in your, and and so it's this, you know, Google, trust pilot, the app store.
• 34:24
  SPEAKER_03: It's a much more vertical as opposed to horizontal business, thousands of reviews a month.
• 34:29
  SPEAKER_03: But it's this, it's essentially the same problem.
• 34:32
  SPEAKER_03: They have issues with the app, they have issues with the website, they have issues with these different things and how you break it down, understand it, get the right information back to the right people at the right time, and take care of your customers, right?
• 34:45
  SPEAKER_03: What they found, what the but using, so they have like 25 facts.
• 34:49
  SPEAKER_03: They spend a lot of time developing their facts, great facts, very rich.
• 34:53
  SPEAKER_03: So if somebody complains about something about the app, they have this very sophisticated response, very informative response that gets incorporated.
• 35:02
  SPEAKER_03: Customers love it, right?
• 35:04
  SPEAKER_03: Really helps change the impression of what's going on.
• 35:07
  SPEAKER_03: Because the problem is that if somebody says something wrong in a review and you don't rebut it, all the review readers think that's the truth to engage in that ecosystem.
• 35:18
  SPEAKER_03: And unfortunately, you know, most companies just don't.
• 35:22
  SPEAKER_03: They they just don't engage.
• 35:24
  SPEAKER_01: Yeah, yeah.
• 35:25
  SPEAKER_03: And primarily because they don't know they can.
• 35:28
  SPEAKER_01: Okay, yeah.
• 35:28
  SPEAKER_01: So, so are there bandwidth issues and logistic issues and tactical issues and so on, or you might say it's it's a dawn on their day, right?
• 35:36
  SPEAKER_01: It's not that important.
• 35:37
  SPEAKER_01: But uh, but you know, I I and I imagine that all this this intelligence, this market intelligence and voice of the customer obviously always goes back upstream to try to fix whatever problem you have, right?
• 35:50
  SPEAKER_01: So is it a lack of onboarding?
• 35:52
  SPEAKER_01: Is it a lack of uh the right help?
• 35:54
  SPEAKER_01: Does a does a screen need to change?
• 35:56
  SPEAKER_01: Does a customer service a response, whatever it might be, but you're going back upstream to the source of the problem and doing that remediation so that you don't get those kinds of responses in the future?
• 36:07
  SPEAKER_01: So that's highly valuable.
• 36:09
  SPEAKER_01: And then I do like the idea of understanding the gaps in the competitive landscape to say, you know, are we missing uh you know the information?
• 36:17
  SPEAKER_01: I was gonna say missing the boat, but I didn't want to do that because you're a ferry example.
• 36:21
  SPEAKER_01: Where are we missing information?
• 36:22
  SPEAKER_01: Where are we uh failing the customer?
• 36:24
  SPEAKER_01: Where are competitors, you know, outperforming us in these different areas?
• 36:28
  SPEAKER_01: So very, very valuable.
• 36:30
  SPEAKER_01: And I can see, you know, how it's a kind of a uh an evolution of uh typical sentiment analysis because you're going a step further and you're trying to understand more of these factors in a broader way.
• 36:41
  SPEAKER_01: Let's just talk about where where do you see the future of this going and what are your thoughts about what organizations need to do to kind of get the best use from AI in their customer experience?
• 36:53
  SPEAKER_01: So, where is it going and what do companies need to think about?
• 36:56
  SPEAKER_01: Yeah, I think I think just gonna get better and it's gonna go away.
• 37:00
  SPEAKER_03: It is what what I've seen, what I my my experience with AI has been that from where we were at the very beginning when we started this, we have many more agents now than we had before because we find that quality is higher the smaller the ask, right?
• 37:16
  SPEAKER_03: If you ask AI to do 50 things, I'll just give you a simple example.
• 37:21
  SPEAKER_03: If there are 25 facts and we would feed 25 facts in a prompt and say which of these are relevant, it would always get the first one right and it would always get the last one wrong, right?
• 37:30
  SPEAKER_03: It just gets lazy and tired.
• 37:32
  SPEAKER_03: But if you ask 25 different times, you get the right answer each time.
• 37:37
  SPEAKER_03: And it's a tiny bit more expensive, but not that much.
• 37:40
  SPEAKER_03: And so, you know, AI will get better at handling these kinds of things over time, and it'll make the system more robust, right?
• 37:49
  SPEAKER_03: That's kind of what I that's one area that I see AI changing.
• 37:54
  SPEAKER_03: I think the other thing that will make this easier is when we start onboarding people with AI, getting these things set up, you know, we have a team, right, that like onboards people because it's one thing just to go in and give a generic AI response.
• 38:08
  SPEAKER_03: It's like anybody can do that, right?
• 38:10
  SPEAKER_03: It's easy.
• 38:11
  SPEAKER_03: But when you start making it more, when you when you make the system more robust, it's harder to set up.
• 38:16
  SPEAKER_03: And so our larger enterprise customers, they'll have somebody and we have somebody and we help them.
• 38:21
  SPEAKER_03: But I think longer term, as AI gets even better, we're going to be able to do kind of AI onboarding.
• 38:26
  SPEAKER_03: And that will help people migrate to these systems much, much more, much, much more easily.
• 38:32
  SPEAKER_01: Okay, great.
• 38:33
  SPEAKER_01: Well, listen, uh, George, thank you so much for your time today.
• 38:36
  SPEAKER_01: I appreciate your thoughts.
• 38:37
  SPEAKER_01: We'll have your information in the show notes, but where can people find you?
• 38:41
  SPEAKER_03: So writesponseai dot com is our website.
• 38:46
  SPEAKER_03: We have a special place, write responseai slash podcasts slash early AI, where they can go and we have a special coupon for your listeners.
• 38:57
  SPEAKER_03: If they have any interest, they can book a call with me, they directly with me, and they can get a coupon code to get free credits to help them learn more about the system.
• 39:06
  SPEAKER_01: Okay, you're gonna measure our effectiveness.
• 39:08
  SPEAKER_01: Great.
• 39:09
  SPEAKER_01: Okay, exactly.
• 39:10
  SPEAKER_01: I believe I'm a big believer in in quantitative uh measures, so that's great.
• 39:15
  SPEAKER_01: Well, again, thank you so much for your time.
• 39:17
  SPEAKER_01: I appreciate it, and thanks to our listeners.
• 39:20
  SPEAKER_01: Uh, we will see you next time to the next Early AI podcast.
• 39:24
  SPEAKER_01: Thank you for joining us, and we'll see you next time.
• 39:26
  SPEAKER_03: Thank you, Seth.
• 39:28
  SPEAKER_00: Thank you for joining us on another deep dive into AI innovation.
• 39:32
  SPEAKER_00: Tune in next time when we introduce another industry expert and discuss how to maximize AI in your world.
• 39:38
  SPEAKER_00: The Early AI Podcast is sponsored by Early Information Science Invector.
• 39:44
  SPEAKER_00: To learn more, visit early.com.
• 39:46
  SPEAKER_00: That's E-A-R-L-E-Y.com.
• 39:49
  SPEAKER_00: Thank you for tuning in.