Michael: Welcome to the disambiguation podcast, where each week we try to remove some of the confusion around AI and business automation by talking to experts across a broad spectrum of business use cases and the supporting technology. I'm your host, Michael Fauscette. If you're new to the show, we release a new episode every Friday as a podcast on all the major podcast channels, on YouTube as a video.

And we also post a transcript on the area on research blog in case you want to stop by and read it in our show. Today, we take a look at AI for customer experience, and I'm excited to be joined by Vanessa Thompson, who's the vice president of product marketing at Twilio. Vanessa. Welcome.

Vanessa: Hey, thank you. So nice to be here.

Yeah, love that you were able to join in. And if you could just give us a little bit of your background and what you do at Twilio and, whatever stuff you want to brag about.

Vanessa: Great. I'll give you my brag sheet, Mike. Thank you so much for having me. I'm Vanessa Thompson, vice president of product marketing at Twilio. My role covers the product marketing teams, includes competitive intelligence, our analyst relations function, as well as our developer network team, which includes all of our evangelists that develop the documentation and as part of my career, I grew up being an industry analyst. I actually worked with Mike at IDC for many years. I was formerly a senior VP of customer experience at a Salesforce consulting firm called BlueWolf. And I really started my career in banking and government back in my home country of New Zealand. So I did spend a lot of time like in the industry, really getting to know some of those hardcore IT problems.

Michael: That's great. And I think it does give you an interesting background because you've worked in IT in a business, you've worked as an analyst and now doing product marketing. So different perspectives. So I really do appreciate it. I'm excited about the conversation because obviously this is, generative AI in customer experience is one of the hottest areas. And, I spent all week last week in San Francisco at Dreamforce. I heard an awful lot about generative AI in customer experience. So anyway, let's just jump into it. So, I'm just curious. I'd like to start. Maybe let's just talk a little bit about what you're seeing today with customer experience and how AI is being used and maybe some of the more impactful use cases that you've seen.

Vanessa: Yeah, Mike, I think you just said it right. Generally speaking, AI is here and it burst onto the scene really around the ChatGPT kind of era when that kind of got kicked off. But I think prior to that point in time, AI wasn't really well understood, or at least the generative AI elements, right? And so I think a lot of our developer network crew were playing around with GPT 3 back then and like really curious and thinking through what kind of use cases can we support, but it wasn't really until it became a consumer household name that we really started to get the groundswell of interest in it across the broad spectrum of buyers that we, we usually talk to.

And so, if almost every CX buyer that we're talking to right now is asking about how AI can enhance their customer. And I'd actually really like to share an example about how we show up in some of our customers. And so for our Twilio Flex, which is our digital engagement experience for sales and support, we're able to bring in real time data from multiple enterprise systems and have that show up inside that the agent experience.

And so one of our customers is HealthFirst in New York. largest nonprofit health insurer, and they're able to take data from offline and online sources, including visits to health clinics and interactions on their website or on the app, as well as calls made into the call center. And that all feeds into a customer profile that's available inside our contacts. Each experience actually reflects. And so, the data that's available from inside that agent experience help us can actually use all of those insights to enable them to guide members around offerings, provide them with personalized experiences. They can automate some of it and they can also have a human in the loop, like having a conversation with somebody if they have exception based questions or something like that.

And the coolest thing about this, though, is as a nonprofit, automation and AI to be able to help make them more efficient is just a huge win for them. So, prior to using flicks and flicks unified generally, they were only able to enroll two or three new members per day per agent, and now they have scaled that up to around seven new members per day. So, think about that as a 250 percent productivity increase. That's just huge. And so, when we think about like, how does this impact the experience, especially in customer experience, I think that the numbers are just phenomenal.

Michael: Yeah, that's an impressive return. And I see that a lot of use cases are around force multiplier. And I think that's an interesting one. That sounds like they're using this both as a way to interact directly with the patients, but then also to assist the agent. Is that? Yeah. Okay. So that makes sense. It's like a, like an agent assist sort of function to.

Yeah, that's great. Is there anything, that, the only challenge I wonder there is around the data, is that is that anything that they've done special there to meet all their HIPAA requirements, that sort of thing? And you may not know, I'm just curious.  

Vanessa: I think I won't go into these sort of specific requirements, but I will just say generally the idea of getting your data in order so that you can take advantage of AI is just something that everyone we talk to right now is reckoning with and it's really important just to recognize that everybody's going to be struggling with the same problem that you are, and it's okay to ask questions. It's okay to ask for help. And just generally like every company has their own sort of skeletons or like messiness or spaghetti of data. And so, they have to work through what is most important. I would just say for the customer experience side and especially for the contact centers, they've historically relied on static sources, which don't change.

They're not real time. And they can quickly become outdated and irrelevant. And so, one of the things that we can now do with some of these generative capabilities, as well as bringing in multiple data sources from like data warehouses or being able to synchronize profiles, those kinds of things to bring them into the experience we're able to solve some of those issues that, previously maybe an agent had to bring up five or 10 different screens so that they could understand like what's going on here and there with different parts of the customer experience. And so, knowing that these data behind every source or every screen that agent has got in the moment, the knowledge of that in and of itself is helpful and then really trying to figure out like okay What problem am I trying to solve first and what's the most meaningful to me? To help me show that I can produce those kinds of productivity gains we talked about earlier.

Michael: Yeah, that's cool. And I would imagine from an employee experience that has to be a real enhancement for their jobs to be able to have information that they can trust that's at their fingertips all the time rather than having to dig around on five or six different screens to pull it together. Yeah. So, in my adoption survey that I did last month data quality actually was the number one concern when you ask about data, which doesn't necessarily, surprise me because again, that is something that I've heard quite a bit. Is there anything else that you think from a best practice standpoint around data quality that you know that you would be willing to share that you think would be interesting for us to know?

Vanessa: I think the way that we think about it is really in the context of the agent experience. And so if you think about the two main components, it's like first we think about the customer profile, like what information do we need to know about the customer, what they're doing, the key information, making sure that that's up to date. It's also identity resolved and all of that information is correct. And then it's connecting with all the rest of the data from the data warehouse, like what other data sources and other things can we connect that profile to make sure that. It's all kind of showing up in the right way.

The other part about it though is like all of the other information, let's call it, that you can bring into the experience. In our experience, we think about it as inferred traits, predictive traits, and computed traits. Think of inferred trait like... Something you can collect from a customer conversation. Like for example, I'm at a hotel, I call down to the desk. Hey, I would like a different pillow. That's technically something that you could infer as a trait about me and something that we could then add to my profile, and it would show up inside my experience. And then like some of the predictive or computer traits, the things around what is a customer's propensity to buy? Like how many orders have they made previously? That can be computed from my data source of orders. And then other things like lifetime value as a prediction in terms of like how, based on the things that I have previously ordered, what else would I or should I be buying from you in future?

And what is my future potential like lifetime value as a customer? So, there is, there are some ways to think about, like, how can we compute and predict like the likelihood of a customer doing something else versus what is everything that I know about the customer in this moment?

Michael: So, you almost get like an assessment from a loyalty perspective too then, because you're yeah, that's interesting.

Vanessa: Yeah.

Michael: So, it's a part of what you're talking about. It is really this human in the loop idea, or I wrote something not long ago human machine collaboration. Is that something that you guys are working with companies on because obviously agents, you've got systems that can deal with customers directly without a human in the loop and then with, so talk a little bit about that.

Vanessa: I think mostly the way that we like to think about it is like this, you should be focused on having a human centric experience anyway, right? Like the human in the loop is really only to make sure that the experience is going to be, magical for your customer, but at the same time, it doesn't mean that like you can't make it human centric.

And I want to share another story, actually, if you don't mind, right? Yeah, that'd be great about another one of our customers. They're called Travel Perk, and so they are a Barcelona based corporate travel management company. And part of the secret sauce is the experience that people have with them and their platform. And they really wanted to more deeply understand their users before they began booking trips on the platform. And so, the issues that they were running into though what we talked about earlier, they didn't understand what was going on with the customers. They didn't have that full profile.

They also didn't have a bunch of standard procedures. Around how to deal with data governance and so they had to think through all of that and put the structure in place to be able to take advantage of all that data and build the insights so that they could use it. And so, what they do now is they collect and standardize data across all the different business units through one platform on our Twilio Segment CDP. And then, because they use the CDP so extensively, they're able to build out an extremely robust profile that they bring into the Twilio Flex contact center platform. And so now, they can really get to that next level engagement experience that they were striving for and really deeply understands their customers.

And so, from a metrics perspective they are looking at around a 95 percent NPS score, which is pretty solid, right? Like the travel industry generally is like diabolical. When you think about some of the kinds of customer experiences, it's not as bad as some of the wildest carriers. They're getting better. Everyone's getting better, but I think generally speaking, that is an excellent NPS score.

Michael: Yeah, that's good. It is a pretty amazing result. I can tell you after my last trip to San Francisco, the airline that I flew, I will, let them remain nameless would not be anywhere near the 95% travel. And as you hear people complain about it all the time. And so that is amazing in an industry where there are a lot of things that can go wrong.

Vanessa: Exactly. But I think that's also the thing that we're grappling with in the industry. Like balancing expectations with experience. And I think we've always had this gap between what a business thinks are a great experience and what a customer believes is the right experience for them. And so, the companies that are doing really well here are using data and context and all of these, some of these newer sort of AI based elements to really bring together. What does like a really good experience look like and how can we help engage those customers in a meaningful way? And yeah, I think the customers that are doing a good job here are really winning.

Michael: And I, we used to talk about like experiences in a way like, oh, this is over the top, how, it's amazing. That's where I think. But most of the time, the truth is people just want things to work. Like they don't really, it doesn't have to be some amazing new experience, but if you have the data and you can keep things flowing and you deliver where you're supposed to, to me, that seems like that is the ultimate experience, really, and once in a while you wow me cool, but in the meantime so one of the things you're talking about in there, I thought was interesting, though.

And this, I think a lot of companies, struggle with because they have a lot of data silos. They have data scattered everywhere and having a way to focus the data collection and build the profiles out in a CDP. Which is, something a lot of companies have worked on, but they're still Implementing. Is that a common sort of approach that you've seen with the customers you've worked with? They build out this kind of central collection point to manage the data.

Vanessa: Yeah, it really depends on the company. I think like from a customer experience perspective, and at least in the contact center, I think the CDP concept isn't something that they're used to, it potentially is owned by a marketing department or like some other department. It feels a little bit foreign to them. But I do think like the places where we've seen like the travel book example where they really focus in on like building that really nice, like data governance kind of experience. It does mean that it can disseminate out through the rest of the organization.

So as a customer experience leader, if you're really thinking about, how should I pay attention? We'll actually start asking questions, be curious about what data you do have and how can you use it to your best advantage is probably the first start. Everybody, we all have to start somewhere, and I think like just being curious and asking questions is the right place.

But I will say from our perspective and from how we're thinking about it in terms of like vision is really thinking about how can we bring all of those data elements. So, we want to the contact center or customer engagement experience and in kind of a low touch way, right? Like we don't want customer experience leaders to have to really think about the data necessarily. We just want them to be able to share like, okay here are the things that I need to bring into my profile. And here's how I would like to think about, how to move the agent experience forward.

If it's something like web page views or like I just want to see when this person went to the support pages or the returns page or something like that. If that's something that you feel like you need to bring into the agent experience, it's possible now for us to do some of those kinds of things and build integrations like, and actually just use AI to just bring it into the experience.

And, I think abstracting away from like the data layer itself, I think, we're really trying to think through, okay, what is the difference between an integration and how you use AI? They are quite different, like integration versus AI is quite different. Cause in the past, like the general, generally accepted, like concept was like, gosh, I've got to integrate this system and that system to be able to move data around and that's, we don't have to do that kind of thing anymore. A bunch of really cool things that help us leapfrog some of those old behaviors that get us to a much better viewYeah. That

Michael: Yeah. That makes sense. I think the one thing that I've seen over the last year finally is that CDP in general I think you're right. It was originally this marketing thing, right? And I always bothered me because that just creates another silo. Why would you, why does that make sense? The point is that everybody needs access to the customer profile and data, particularly in sales and marketing and customer service and finance. And if you have a way to consolidate it, why wouldn't you use it across all those different things? Yeah, I think that sounds, it sounds like it's creeping at least into the customer service part of the business, which is great.

Vanessa: Yeah, absolutely. And I think that the most successful, I'll just say like the most successful companies that we work with and have segment everywhere. They are really embracing that, the ability to bring in their total sort of organizational worldview in terms of what's going on in their environment and really coalescing around a segment to help them do that.

Michael: Yeah, that to me again, that's a very logical way to think of it is this, use cases exist all across the business. So that's, yeah, that's good. So just shifting back out again to AI in general and generative AI, particularly in in CX. What are some of the more promising use cases? What are people doing with it that you think are unique or are that the audience would be interested in? Yeah, I think what are the more promising use cases?

Vanessa: My, my commentary here is really around just the framing for the inferred traits, predictive traits, and computed traits. So, I think instead of just talking through use cases, because every company is a little bit different and they'll want to implement a little bit differently, but just structurally.

What can you know about what your customers are doing and how important is that information to bring it into the context of the conversation? So that's the infiltrate space. What do you need to compute to help you provide that, like in a moment point of view around what a customer has done. Is it the number of orders it is it the order value? Is it store visits? I don't know exactly what it might be, but is there something that you need to compute to help you figure through like how to give your agent or system the best point of view of what's going on? And then what else do you need to predict, right?

Is there a churn predictor? Is there a lifetime value predictor? And it could even be like on a scale or something, right? Like one of the use cases that we were looking at is, is this sort of low churn or medium churn or high churn risk? And at what scale is the customer at risk and can we provide some pre prompted You know, scripts for somebody to talk someone through a day to day automatically show up as a high churn risk. Do we want to add like another agent to the conversation even before they get to the first level conversation is there somebody else that we need to be headed? Is there a retention specialist or something that they needed to get routed to even before?

The first person picks up the call and so I think there's some really interesting ways to think through how to innovatively use AI to help manage some of that risk profile because a lot of that risk is hard to manage because we can't do it in real time, but now we can actually. And so being able to proactively go through some of those scenarios is pretty interesting.

Michael: Two, two things jump out there as I'm listening to that is that it's almost like saying just step back and take a broader lens across everything that you're doing, because you can use AI in very strategic ways. Particularly around the predictive capabilities. If you have the data to help you with whatever that problem might be, and it could be churn, it could be, increasing revenues or whatever, right? It depends on what it is, but it's but the advice of, step back and think of it in a strategic way, not just tactics that I think that's a great way for companies to think about it. So that sort of leads me to the ultimate question that I get asked all the time. And I'm sure you do, too. And it's the when should we get in? And what should we do? And should we do anything with AI now or, maybe it's just too crazy out there. Maybe we should wait for a year. What advice do you give companies that are thinking about and not haven't gotten in yet?  

Vanessa: I will give you my, my sort of starting advice, like starting is hard. I think starting anything in life is hard. And so, this is just such a huge conversation that, and sometimes it can feel intimidating. Maybe there's a lot of sort of barriers that you feel like, gosh, I've got to get myself fully educated on this topic before I put a proposal together for my superiors or something like that. But I think like getting comfortable with the fact that like this is new to everyone and you really just have to get started. We talked a little bit earlier about data being really messy. We know it's messy. And so that's something we also have to just accept and move through and get comfortable again with the messiness.

And so if you are just getting started, like personally educate yourself Oh, look at some of the tools that can help you. And go and do some things and your consumer life and see it, see how they work and get comfortable with the consumer experience, but then also propose some sort of lower risk or smaller proof of concept projects so that you can test the edges about your level of comfort, but also your team skills. And that way that helps you build kind of confidence on your ability to take on something of a larger scope. I think the folks that are moving fastest, I would say spent a ton of time being really thoughtful and intentional about the data strategy. So getting straight with how to organize your data in a meaningful way, make sure that, like where to look for things. I think like those are the pieces or the cases where we see like folks we had to move the fastest. .

Michael: Yeah. That, that makes sense. The other thing that sort of strikes me too, when I think about it, is considering everything that's happened since ChatGPT was introduced in November last year. November, December. Think about just from there to here. Like even some of the tools that I tried in January that were unusable that now I use every day because they've evolved so quickly. And I, in that survey that I did last month, one of the other things that came out was one of their biggest fears and problems. Challenge is internal skills are getting the right skills. And then the second one was finding partners that have the right skills. So, it's all really the same thing, right? It's the fact that skills are you know, still not necessarily available, but if you don't do it now and you try to do it a year from now, it's going to be completely different then. And things have happened along the way that you just didn't have a part in. So, to me, it seems like a risk, a bigger risk. To not do it than it is to jump in and at least do some pilots and figure out what you want to do with it.

Vanessa: Yeah, in some ways, it's the risk of doing nothing is high. And yeah, I would just advocate for being curious and really trying to meet the moment, right? This is a huge moment in technology. Our CEO Jeff Lawson really talks about this as like the latest big shift since. For the longest time, right? I think we truly haven't seen a technology shift this dramatic. And I think mobile was it, but actually it might not be as big this might be bigger than mobile even, right?

And it does feel like a moment that you can meet with the growth mindset or you can probably and maybe get a little bit left behind.

Michael: Yeah, I think about it like it's more transformational like the big platform shifts mainframe to client server to cloud those kinds of things. Because this is really that. The next generation of a lot of things, even to me, I even think, mobile was very disruptive in its own way. But I think I may be more disruptive than mobile, because it is changing the way we use those tools as well. Yeah, that's interesting. One of the things that companies need to do is measure you know, watch the metrics and understand, are they getting a return for what they're doing? And what, how are companies measuring the ROI? And, how are they, how do they determine if their metrics are on track using these new tools and technologies?

Vanessa: Part of it is just the metrics themselves. I was watching something recently, which was a professor in a classroom scolding, like, all of the students because 90 percent of them had used ChatGPT for their homework. And he was like, I'm gonna fail 90 percent of you because you all use this for your homework, and for those that didn't good on you. But you still, not everybody was getting a good grade here. But I think that's... Illustrative of the difference and the challenge now, right? Like it will be part of how a student operates and how a student creates their work, but what is the next piece on what they do with the information that they get like what is the additional creativity? What is the additional insight? What is the additional clarity that they bring on top of what the computer can generate for them? And from a metrics perspective, I think the concept, or the mindset should be very similar in that you could automate a lot of what we had to do before, but the ROI could really look based on what you want that kind of outcome to look like. So, we talked a little bit about some customers of ours and health versus a great example, right? Because they took a productivity approach to how they're thinking about ROI. If you can process two to three applications per day, now you can go to seven.

That's huge as a nonprofit for them. They don't have to hire more people to do the processing. They can get through a higher volume and ultimately, like they can do more good. And so, for them, that's just a huge bonus. I also think like to that point about. The university and like where you can. And creativity, clarity, color to like what the computer generates for you. And some ways there's also a view around like exception management. Like how do you review what that, what it provides you to make sure that it's correct. It is like the intent is correct. And that's the human in the loop piece.

And so, I think I suspect, and we don't really know how this is going to play out yet to some extent, but I think there'll probably be some sort of human in the loop. Exception management type metrics that show up over and above the ROI. And so the, if you think about it from an OKR perspective, like maybe there's a broader objective to use the generative AI and then the key results will be focused around how are we managing exceptions and how well, things like that, right?

Michael: So, it obviously varies a good bit by the use case and specific strategy too. So that, yeah, that makes more sense to me. I could keep going with this for another couple hours, but I suspect that we'd lose our audience if we try to ramble on that long. So that's really all the time we have. And I really appreciate you joining me today. Great information. And I think I learned a lot and I know the audience will too. So, thank you. Before I let you go though, one of the things that I always ask the guest is, could you, did you recommend someone, a thought leader, an author, mentor, somebody who's influenced your career? And you'd like to share that so that, other people could learn from them too.

Vanessa: Yeah. I had the privilege of working with you and for you, Mike, for five years of my career. And so that was fantastic. I learned a ton. You're one of my main mentors, but I think there is someone that I came across a few years ago. Her name is Shelly Archambault. And I first heard about her on a Reid Hoffman masters of scale podcast, and he did this fantastic interview with her where she talked about her career And she is just unapologetically ambitious and I just love that about how she operates she basically took a look at how she what her playbook to being successful should look like.

And she went and executed that playbook. And then she got a name for herself as a sort of a turnaround CEO and Silicon Valley here. And so she has this really great story about working at Blockbuster and pitching them on an online business model. And when they said no, she was like, okay, that sounds great. I don't think this is a good fit for me. Yeah, exactly. And so, I think, and that's one of those moments where sometimes as business leaders, you roll over a moment and you don't realize the significance of a moment like that. And similarly, like the AI, this big shift in AI is one of those moments that I just thought that was really inspirational that she was able to have the self-awareness to look through at that moment and go, okay, this is not for me.

I'm gonna move on to my next thing that I think is a meaty enough challenge for me. So yeah, she would be my pick.

Michael: Yeah, sometimes stepping back and moving on is hard, is harder than you think and certainly it's the right thing to do. I love Masters of Scale, that Reid Hoffman podcast is great. I've been listening to it for quite a while. It's some really interesting things there. So, thanks. I appreciate it. Appreciate that. Thanks for sharing. And to the audience, thanks for joining us this week. Remember to hit the subscribe button and for more on AI, if you go to the Arion Research site and click on the research tab, you'll see an AI adoption report that we published last month. It's based on a survey from August and in all things AI, the more current it is. The more important that is because believe me, things are changing every week. So, it's free research, free download, check that out and then check back next week. We're going to have a special edition from London. I'll be at sugar CRM user conference, and I have a few interesting interviews set up and they're making some interesting gen AI announcements that I can't tell you now, but it'll be exciting I know. So that's it. I'm Michael Fauscette and this is the disambiguation podcast.