Kylah Field on AI Customer Experience: Why It Takes Real Work, Not a Quick Fix

Meet Kylah Field: From Cost Center to Business Driver

Kylah (0:02) My name is Kylah. I work for Spanx: women’s shapewear, intimates, and apparel business. I’ve been there about three and a half years. I’m our Vice President of Global Customer Experience and Consumer Insights, and recently have expanded my scope to wholesale optimization.

We’ve been working together for, I would say, almost a year and a half. Honestly, I think we signed on maybe in 2023, the end of 2023.

Lisa (0:32) Yeah, you were one of the AI pioneers, and I love talking to AI pioneers and getting their insights into how their AI journey has been, so we can educate other interested brands about AI. But I’m curious, as a VP you’re responsible for both the team and the business outcomes. So how do you think AI has changed your mindset about building and structuring a CX organization?

Kylah (1:01) Yeah, I think the CX that folks knew, call it three years ago, is not the CX that we have now. I feel like CX has gone from a place that’s kind of just answering questions, solving customer problems, to really an organization that’s a business driver, responsible for results, responsible for generating revenue. And I think AI has really allowed us to level up in that regard.

For us at Spanx in particular, it’s given us almost immediate insight into who our customer is, what they want, what they need. They could be shopping on a PDP, and we’re able to stop and engage there and help close a sale right then and there. So I think that’s one thing. The other thing I’d say is just the breadth of insights it gives us in real time about who our customer is. For me in particular, it’s been super helpful to partner with our brand team, to partner with our product and engineering team, because we’ve been able to use that data to essentially tell the business what customers want and need.

As I mentioned, I also own consumer insights. In the past, you might have to go out and conduct a survey, do a huge focus group to get this information. Now it’s accessible right away.

Why Every AI Agent Needs a Dedicated Owner

Lisa (2:15) We were talking about this earlier, about how you’ve structured the team and how you assign responsibilities with the new AI projects. But I’m curious how you also change the way you’re hiring or leading the team.

Kylah (2:29) Yeah. So from a hiring perspective, I think the biggest thing with having an AI agent is it’s really important to appoint a person to own the product end to end. A lot of my peers, and a lot of people out there, think you onboard a tool like Siena and then it just takes care of all your problems. While Siena is awesome, I think to get the most out of it, you need a person completely dedicated to it: someone in the weeds with Siena, understanding what the new product releases are, QAing it, figuring out what the tone and voice should be.

So for me, we hired a person to do just that. Call it three years ago, I wouldn’t have looked for a person and said, “You’re going to sit on the CX team and your only job is going to be to focus on our AI agent.” That’s one of the big things that have changed, and I think the future of CX hiring is just that: you need AI strategists, super-skilled individuals who know how to leverage these tools and manage AI agents.

The Business Case: Personalization at Scale Without Tripling the Team

Lisa (3:33) So I think what you’ve done is really important, and it’s like what makes or breaks the success of an implementation. So when you decided, “Okay, we need AI in our customer experience,” what were you going through? Was there a tipping point, and how did you go through that decision?

Kylah (3:48) Yeah. So for us, obviously Spanx has a ton of competitors, and we want to have that competitive edge, to be able to give them the best experience possible. So for us it was really about personalization: to get as personal and as intimate with our customers as we wanted, we’d have to hire a ton more people to do that. AI actually enabled us to cut our existing team’s workload in half and also provided that personalization at scale. So for us, there needed to be a way to have these really one-to-one, intimate, personalized connections with customers without tripling the size of our team, and AI was the way for us.

Lisa (4:28) And did the decision come from the leadership team, or how did you decide that, okay, maybe we should start? When it comes to AI, the first use case that makes the most sense is customer experience, and then maybe across other departments.

Kylah (4:42) Yeah, on our end it was really the CX team just deciding this is something we want to do, and Spanx being a place where employees are free to tinker and figure out what it is that they want for their individual businesses. So that’s kind of how it happened. We started small with Siena. We were very cautious and careful. We implemented first, like, “let’s try to test out where’s my order,” okay, they’ve done a good job with that, “let’s try and now give Siena product recommendations.” And we ramped very slowly to make sure it gave us everything we needed, refining things with you guys until now. At this point, Siena is handling 50% of our conversations, but it was a very slow, deliberate, and methodical process to make sure it gave us exactly what we needed.

Choosing a Vendor: Why Brand Voice Won

Lisa (5:27) That’s amazing, thank you for sharing. And how did you go about choosing the right vendor? What are some of the criteria you looked for, because the market is so noisy?

Kylah (5:38) Yeah, transparently, we talked to a ton of different vendors. Some had really great technology. A good chunk of people had really great CSMs, things of that nature. But I think the thing for us that really tipped the scale towards Siena was its ability to get our customer and brand voice and tone.

Our customers are coming to us for a wide array of reasons, but typically it’s for a moment that matters. That could be a woman who’s getting married: that’s a moment that matters. They could be postpartum, their body is changing, and they’re coming to us: that moment matters to them. They want to look good and feel good and feel confident in what they’re wearing. So when you think about that, customers are divulging really important, personal, intimate details about themselves. You don’t want them engaging with an AI tool that’s impersonal.

So for me, brand voice and tone, and the ability to understand our customer and tweak things to make it better understand our customer, was the most important thing. When I went out to figure out what we were going to do, it was: brand voice and tone is the most important, full stop. We can figure out the technology and what works, because in 2023 everybody was starting in this very, very crowded space. We were talking to tons of people, but Siena, for me, really felt like a girlfriend, a trusted person that I’d be comfortable divulging personal information to. And when it didn’t feel that way, the ease of being able to tweak the voice, tweak the tone, tweak the disposition so that it felt that way was seamless.

The Surprise: AI Raised the Bar for Human Agents

Lisa (7:10) That’s a very detailed way of describing it. I appreciate that. And it’s so true: when I think about Spanx, even when I have the logo in my head, I think immediately “confidence” is the word, and that’s kind of what you also want to portray when you have customer conversations across any channel. That’s so cool. Is there anything that maybe surprised you about Siena, or the agents you have working because of Siena, that you might not have known you’d be able to do or achieve?

Kylah (7:40) Yeah, I would say what surprised me most was how Siena became almost like a standard for our human team. Obviously, when we implemented Siena, I assumed we’d get efficiency, ease, things like that. What I didn’t think about in the moment was how our human agents would come to emulate some of the responses that Siena came up with.

We use a number of different guidelines for human agents and Siena to respond to different questions, comments, what have you. And as we implemented Siena, I noticed the responses, the thoughtfulness, and the personalization tied to our human agents also got better, because they had other examples of what good work might look like through the responses that Siena gave us. So not only did CSAT increase overall from Siena, it also jumped significantly on the human-agent end, because they had that experience of Siena and now the ability to think differently: like, “oh, that might be an interesting way to answer that question,” or “that might be a different way to tweak the tone,” or how to make that product recommendation, where I hadn’t thought about that before. I felt like that’s just something I literally hadn’t considered, and I was like, wow, it’s really helped elevate us overall.

When Customers Ask for Their AI Agent by Name

Lisa (8:58) I’ve heard that before. I think Sarah from MUD\WTR, the mushroom coffee brand, shared that she used the brand persona to train her human team, like, “hey, these are the guidelines, here’s how we want to communicate.” That’s so cool. And are there any customer interactions or reviews or interesting feedback that you saw after the AI deployment?

Kylah (9:23) Yeah, I mean, I think some of the moments that stand out to me are how our customers are mentioning our AI agent by name regularly, and how they started responding to it as if it was a trusted friend. We’ve had folks want to invite our AI agent to weddings. We’ve had them want to take her out to places, ask if she could be the personal stylist for their daughter. There’s been a number of things where I was just like, “whoa, they really love our AI agent.” I just didn’t expect that. I expected it to be efficient, I expected it to be good enough, I felt like customers would feel like, “okay, she answered my questions with empathy,” etc. But I didn’t expect the connection that customers would have to it.

Lisa (10:12) Wow. And do you disclose that it’s AI or not?

Kylah (10:15) We do. But I guess once folks start talking, they forget or don’t care.

Lisa (10:22) In terms of channel distribution, you have a lot of ways someone can get in touch with Spanx: email, chat, social. How do you feel about that, because recently we’re working through adding a shopping agent to Siena. I’m curious why you think having a single AI layer across all these communication channels, across the entire CX, is better versus maybe using different solutions. How do you think about that as you scale?

Kylah (10:52) Yeah, I mean, I’ve used different solutions in the past, and speaking from personal experience, one tool makes the experience more seamless. You’re leveraging who the customer is, you’re able to tap into their previous experiences, reference that as a context point for the interaction, and overall it just makes the experience better. And then, on the intelligence side, for your CX managers and supervisors, you’re able to look at the data and get an understanding of what’s happening across the entire customer journey, which I think is super important.

Lisa (11:14) Yeah, the memory and the context layer is what really makes a ton of difference, because you want your customers to feel like they’re speaking to the same person every time: like, oh, they know me so well. It’s like going to the coffee shop and they already say, “hey, here’s your order,” without you even ordering. That’s kind of the feeling.

Kylah (11:44) We actually have customers who ask for our AI agent by name, because of what you’re describing. They feel like she knows them so intimately. It’s like, “oh, let me get the AI agent, and I know they’ll remember that previous interaction we had, or that last purchase I made, or that fumbled return I had before, and she’ll know what to recommend.” It’s been really interesting to see.

From Inbox to Insights: How AI Freed the Team

Lisa (12:11) I want to talk a little bit about the team. I’m curious how your team’s roles have evolved, or do you feel like you can do so much more now?

Kylah (12:22) I mean, I think a couple of things have happened. On the actual agent side, I think by employing a tool like Siena, it’s really freed up their time to work on more complex cases. For us in particular, to cater to our most valuable customers, you might have a bit more time to engage in a really complex interaction, or send your VIP customer something for their birthday, a special gift, things like that. We’ve had a lot more time to put into surprise and delight, making customers feel special, than we had previously, when we had to be heads-down in the inbox. Having Siena, we realized the inbox will be taken care of, and taken care of well, to our expectation. So that’s one thing that’s happened.

But then I also think the rest of the team has been able to leverage Siena, the Siena Intelligence tools, to dig into gaps they may not have seen otherwise. A lot of my team’s time has transitioned from, “I’m going to be in the inbox, we have to do these things in the inbox,” to, “oh wow, I’ve noticed there’s a spike in defects on this particular product. Let me now talk to the product team to see if we can work this out,” which is exactly what you want them to be dealing with.

Lisa (13:34) Yeah, so they can spend more time in the data, accessing it faster, and then sending it to the right stakeholder. What are some of the other interesting data points you’ve started digging into now?

Kylah (13:46) Yeah, so I’d say shipping and delivery is something we want to get better at. A good example: we noticed there was a particular region of the country where delivery had been slow. We wouldn’t have noticed it if you looked overall, because delivery is generally what we’d expect, but in this particular region something was going on, and Siena helped uncover that. You don’t get a lot of context about this, but the context you do get is all saying the same thing. It prompted us to go back into the data, where we saw almost 100% of things going to this region were messed up, and it caused us to uncover some issues with our carrier that we wouldn’t have known before, because the data point wasn’t big enough for us to have seen it otherwise. But Siena spotted it and noticed right away: something looks a little off here, you should dig into it more. Things like that have been really helpful.

We’ve also been able to leverage it for ideas customers have about new products, what they’re surfacing, what defects we have with our existing products. And we’ve been able to leverage it from a brand perspective: when we think about positive comments customers make, we’ve been able to use that and surface it to brand for them to use in marketing. Whereas before it was a human going through a bunch of reviews, now it’s more like, “hey, we’re looking for comments where people talk about this particular product in a positive way, can you surface some for me?” And we get them.

Proving Revenue Impact and Advice for Getting Started

Lisa (15:13) Yeah, that’s amazing. I love how, because you can tap into this intelligence very fast, I feel like CX is now in the driver’s seat, and it’s being seen as a revenue-generating channel, not just a cost center.

Kylah (15:30) I think it’s also been easier for me to make the case for the customer to our ELT, to our board, because the data is right there, it’s clear what’s happening, and it’s easier to illustrate the story. Easier to tie metrics and revenue numbers to customer experience in a way that I could do before, but it just took a lot more work to do it.

Lisa (15:56) Yeah. Because this was actually one of our talking points: CX was always pretty siloed from the rest of the organization. Now you can actually embed it into so many areas, and the information flow across the company is…

Kylah (16:11) Almost. I’d say, in the three and a half years I’ve been at Spanx, for a good chunk of that time I feel like it was siloed. It’s not anymore, and I think Siena has helped it not be that way, because it’s easy to socialize information: easy to give other teams access to the information so they can look and be empowered themselves to say, “oh wow, I found this trend myself,” which I think is really helpful.

Lisa (16:37) Amazing, super cool. And then, as a leader, what are the metrics you track day-to-day, that you need to always have in front of you and have your team report on?

Kylah (16:47) Yeah, so we track CSAT, we track NPS, average handle time, SLA, things of that nature. I think what’s been really cool about a tool like Siena is seeing how those metrics have changed over time. Our average handle time has been cut in half. Our CSAT has gone up five to seven points, things like that. It really helps, year-over-year, show the power of AI as far as efficiency goes, and even quality.

I think a lot of organizations are excited about AI because of the efficiency gains, but for us, even more than the efficiency gains, it’s been the quality gains that have come with it. And that’s been really exciting, in a time where customers can go to a million different places to get their stuff. I think CSAT is like the X factor, the differentiator. If you’re able to recover gracefully when a customer has a problem, they’ll come back to you over and over. If you have a bad experience, if your stuff doesn’t deliver on time, if the experience isn’t easy, I can go shop wherever I want, I’m going to try something else. So I think it’s really helped there.

Lisa (17:55) How are you thinking about scaling with AI? Any other interesting projects or initiatives you have?

Kylah (18:02) I think we talked a little bit about it, but we’re really excited about introducing the shopping agent with AI: in addition to the AI agents we have where you reach out to CX, they’ll also be available on the site, can help you shop, answer questions about whatever product you’re looking for, and connect the dots throughout the experience. They might recognize who you are and recommend products, for example, that they know you like based on your shopping behavior, return patterns, etc. We’re excited to see what that looks like at scale, and what potential revenue comes from that.

I think there’s a lot of revenue that CX is already driving, and has been driving, but it’s just hard to recognize and prove that. Now I feel like it’s going to be way easier: like, “hey, the data is here, it’s because we trained it this way, it’s placed here, and this is the part of the journey it covers.”

Lisa (18:58) Yeah, I think it’s been helpful for us, like if we’re wondering where we should spend that next dollar: is it on this marketing ad, is it on this new feature, or is it on CX? In the past it’s been hard to say it’s on CX, because it’s like, sort of, why? And it’s really hard to provide true numbers of what that will look like. I think with something like shopping agent, and just in general with our experience with Siena, I’ve been able to tell that story in a way I hadn’t before, and often we’re getting the dollar versus another department, which has been good. There’s still a lot of skepticism, and a lot of companies that might have had bad experiences with AI or other automation. What would you advise someone if they were to start now? What are maybe the top three things you think they should do?

Kylah (19:50) If you’re skeptical, I think my advice would be to start slow. Pick one or two pain points and work with AI to automate that, and if you have a good experience with that, I’d recommend moving further.

The other thing I’d say is I feel like a lot of individuals, a lot of companies, think, “I’m going to partner with this tool, we’re going to implement it, and that’s it, all will be well.” I think it’s really worthwhile to hire a person, maybe two, depending on the size of your organization, whose sole job is to think about AI for CX: auditing, QAing what the agent is saying, working in partnership with whatever organization is spearheading your AI program. You really need to invest resources and time into it to make it work for you. If you don’t do that, it’s going to be just like another tool. You have to put the time in, and when you do, it really pays dividends.

And I think a lot of people expect overnight success, or they want to see the ROI in the first 30 days, and they don’t realize you might need to overinvest maybe in the beginning, and then you’re going to see massive savings if you’re doing it the right way. I kind of think of it like learning a sport: the first time you go out there, you’re not going to be that good, your swing, you might not be able to hit the ball right, but after working at it, tweaking it over time, you’re going to get better. I think that’s the same attitude you should approach honestly anything with, but certainly AI. You’re probably not going to get everything you need within the first 30 days, maybe not even the first 60 or 90 days, but the more time you spend with it, the more you refine and tweak the tool, the better it’ll work for you.

Lisa (21:34) Well, thank you so much. I think that was it.