How can employee experience drive better CX
About the show
Hosted by Nikola Mrkšić, Co-founder and CEO of PolyAI, the Deep Learning with PolyAI podcast is the window into AI for CX leaders. We cut through hype in customer experience, support, and contact center AI — helping decision-makers understand what really matters.
Summary
Most companies think about customer experience from the outside in. Sam Stern thinks about it from the inside out.
In this episode of Deep Learning with PolyAI, Michelle Schroeder speaks with Sam Stern, who leads service design at LinkedIn, about why improving employee experience is one of the most effective ways to improve customer experience.
Sam’s team focuses on the human side of LinkedIn’s CX — sales, success, and support — and the systems those teams depend on. That includes the tools they use, the workflows they follow, and the data they have access to in real time.
What they’ve found is simple: when employees are slowed down by fragmented systems, missing data, or constant context switching, customers feel it immediately.
In this conversation, they discuss:
- Why employee experience is often the root cause of poor customer experience How “toggle tax” across systems breaks focus and degrades interactions
- Why employees need the same real-time data as customers to build trust How AI can surface insights and suggest next best actions in the moment
- Why fewer human interactions actually make each one more important
- How LinkedIn is democratizing research and insights across product teams
- What service design looks like when it’s embedded inside product
Watch the full episode to hear how LinkedIn is approaching it in practice.
Key takeaways
- Employees need the same data customers already have: At LinkedIn, service reps were often working from data that lagged behind what customers could see themselves, which undermined trust before the conversation even started. Closing that gap freed reps to focus on what actually adds value: helping customers interpret the data and decide what to do next.
- AI works best as a suggestion engine, not a replacement: Sam Stern's team found that AI often catches nuance and real-time detail a human can't, but customers still trust the human relaying it more than the AI itself. The winning pairing: AI surfaces the insight, and the employee delivers it.
- Fewer human interactions make each one matter more: As self-service absorbs more routine contact, the human moments that remain carry more weight in shaping how a customer feels about the relationship. LinkedIn is designing those moments to be memorable, not just efficient.
- Democratizing research turns insight into action: LinkedIn built an internal insights agent so product teams can query customer research directly, without waiting on service design to run a new study. It's a shift toward embedding institutional knowledge into daily workflows rather than gatekeeping it.
-
[00:00:32] Michelle Schroeder: Welcome to deep learning with PolyAI. We're here to help CX leaders get a window into the latest and greatest developments in AI. Uh, before we get started, just a quick reminder to subscribe, give us a like on your favorite podcast app, or and on YouTube. Alright. So today, I'm joined by Sam Stern, who leads service design at LinkedIn. Sam spent more than two decades in CX and service transformation. Early in his career, he was an analyst at Forrester, a world we understand very well on the vendor side, where he helped define how companies think about CX strategy and design. And today, his team focuses on something many companies overlook, the behind the scenes systems and workflows that actually enable great experience. And increasingly, that includes figuring out how AI can help employees serve customers more effectively. So, Sam, thanks so much for joining me. Sam, so to start, tell us a little bit about your role. You lead service design at LinkedIn. Uh, but what does that mean in practice? What kinds of problems is your team focused on solving?
[00:01:35] Sam Stern: Yeah. Uh, Michelle, thanks for having me. Um, so I would say first and foremost, we are looking at the, um, employee experience. So the human powered parts of our customer experience, which is sales, success, and support. And they're interfacing with customers of LinkedIn constantly, all the time, driving renewals, driving enrichment, providing service. And we're looking at the tools they use. We're looking at the processes they follow, the policies, and really trying to take efforts out of their work, but also give them better information at the right times about our customers so that they have better interactions with our customers. So it's both removing work and giving them hours back in their week, which is one of the biggest pain points that they've told us about over and over again in research. Um, but then also thinking through how can we make each of those interactions they have with a customer more impactful, more engaging to amplify what is primarily a self serve platform to human experience for a LinkedIn customer. But we do have all of these key moments within that where you if you need support, you're chatting with one of our support consultants, it better go well, or it's gonna be a problem.
[00:02:46] Michelle Schroeder: Yeah. Interestingly, your team evolved from what used to be a customer experience team. What changed when LinkedIn moved to a more service design model?
[00:02:55] Sam Stern: Yeah. Well so the biggest change for us as service design is we moved from being on the side of the business that was aligned, organized with sales success, um, and support to being in product. And product is really the beating heart of LinkedIn. And so now we're working directly with the teams responsible for creating the tools and platforms that employees use, but also that our customers and our members use. And so for us now, we only do research when a PM or a product leader asks us and comes to us with a set of questions that they don't have answers to, that they need answers to to build with conviction the right thing, whether it's for an employer or a customer. And that's really changed in terms of the alignment between what we're looking at and what they're building. Because now we've got these trusted relationships with product. They come to us with lots of questions, which is great. It keeps us busy. But we're also positioned to go to them and say, hey. You're not asking us this question about this part of the experience. This is a problem, and we want to bring this to you. We wanna surface this to you, um, because we think that we need to fix this. We need to address this for either our employees or our customers or both.
[00:04:06] Michelle Schroeder: From what you're describing and, uh, the all of the masters that you're servicing and that you're you're you're you're serving in service design, we have a very similar practice here at PolyAI, uh, that we call agent design, where we are, you know, like you, sort of supporting the employee experience, but also the end user, like the customer experience. Um, so I'm wondering with how we both think about workflows and the flow of information, context, and how interactions actually unfold between users and your product. Um, do you see the same similarities?
[00:04:38] Sam Stern: Yeah. For sure. So I think one of the biggest challenges that we saw, just to give a specific example of this, in the employee experience was they didn't have access to the same data that the customer had. So they had data that was maybe lagging or that wasn't as specific or wasn't as, um, accurate even. We would have data accuracy issues. And so they'd get on a call with a customer, and the customer would say, well, that doesn't mirror what I have. And the customer's data that we had shared with the customer, which is great, at least if the customer had the right data, would be real time, uh, would be accurate, would be down to specific, you know, impressions that they've created with a marketing campaign as an example. And so the employee said, I'm embarrassed. Right? And I and now what I'm doing is I'm doing all these workarounds to get the same picture my customer has. We have the data, and we're giving it to the customer. Can you please give it to me? Because what the customer needs from me is not accurate data. Right? They get that. Why are they having the conversation then? The conversation is to help them understand the data and contextualize it and think about what to do next. What's the next best action? And that is where the expertise of the salesperson and the customer success manager is so valuable to that customer, but that value can't be realized if the sales rep is fighting all of these barriers to getting the same data and being on the same page as the customer. So, absolutely, that's a huge one to have a shared view of things that then facilitates that value adding conversation.
[00:06:04] Michelle Schroeder: So important. We have a a a very similar thing in our world too where, you know, these conversations that you're having with your customer, with your end user every single day should add up to insight. It should add up to very rich data to make decisions off of and figure out what that next best action is, but sometimes takes design and thinking through and prioritization and guidance to actually make the most of that and to interpret it and all of that. So to to your your point, it's not about data accuracy or any of that. It's like the the the coaching and the, you know, being able to actually have people with the right expertise to glean what they should be out of that data. And speaking of which, it sounds like one of the projects you're working is building this internal interface for support and success teams. So, essentially, a pane of glass that sits between Yep. You know, that gives employees everything they need to know about a customer. What problem are you trying to solve?
[00:06:56] Sam Stern: Two problems, I would say, bundled together in that. One is what we've been we've taken to calling the toggle tax, where a customer asks a question that they didn't have an answer to or they haven't answered before, and they're hopping all over between all these different systems to quickly find an answer. And you can imagine when it's supposed to be a human to human experience, even if it's hidden through a chat interface, you still, as the customer, feel the lag, feel the lack of focus and engagement from that employee. And the employee feels that they know they're not as present as they could be. And what is the point that I'm having in a human interaction, right, if you're you're not getting a human interacting? So removing the toggle text, bringing the data to them. But then also, there's so much that the AI can build on top of the data if it's coming to that employee. So it's in that pane of glass, uh, view for them to suggest things they can say, to suggest next best actions. And then the employees still suggest them. And we're finding this still, that the trust factor is more so with a human. But the AI is more accurate often or has more nuance or has a better understanding of the data in real time than the human can. So you want that pairing of give me suggestions, me, the human employee, but let me communicate them to the customer so that the customer will actually take them and do something with them. We want them to make, uh, better decisions because then they have better results with LinkedIn. They're more likely to renew and enrich. So it it works for everyone's benefit. So that's why the trust factor with the human is important, but the actual correct and nuanced and detailed and in the moment answer from AI pulling in all that integrated data and learning from it and suggesting the right thing for the human to say, all of that comes together to, one, remove all that extra effort that distracts them, and then, two, make them sound even smarter to the customer.
[00:08:46] Michelle Schroeder: I'm definitely stealing toggle text. That is so smart. Um, and in in our industry, we almost have your reverse issue where it's like the toggling or the the the script that you get as a human being in a contact center can take away the humanity out of a conversation. So it's almost like, you know, if they could, you know, have that toggle tax removed, they connect more deeply with the problems that they're solving and the people they're solving them for. I I love that as a as a frame. For you, these customer interactions that you're having are fairly high stakes. You've mentioned that each one might be the only human interaction a customer has with LinkedIn all year. How does that change the way that you think about these moments, that you design these moments?
[00:09:26] Sam Stern: Yeah. Well so, again, just coming back to my previous answer, and I think it it connects through really clearly. If they're only going to have one or two human interactions you know, sometimes with clients with higher touch service, it's maybe per quarter, but it's still not frequent. We have to get those right because, you know, we evolved to be social beings. Right? Uh, humans have the largest social group of any species. We're always intuiting how other people feel, their emotions, their facial expressions. This is what our brains are consumed with, and that's not gonna change. So how other humans make us feel is something we remember, and we are much more attuned to than a digital interaction. It's just reality. So even though there are fewer and fewer human moments, each one of those remaining human moments takes on greater importance in the customer's overall impression of their relationship with LinkedIn. So that one chat interaction that a customer who doesn't have a big relationship with us, maybe doesn't even have a dedicated sales rep, if that one chat interaction goes well, they'll remember that. Service recovery is so powerful in this in this way to say, well, it didn't work. But then when it didn't work, LinkedIn really stood behind that and fixed it for me and explained how, you know, to avoid that problem the next time. That will live long in their memory. So these opportunities, as they become less frequent, which is good, we're automating, we're providing self serve, which our customers want, which all customers say they want, at least, we are then trying to maximize the positive emotions we evoke in the remaining human interactions and how memorable those interactions are so that when it's renewal time, they remember that, um, interaction with one of our employees fondly and wanna stay our customer.
[00:11:05] Michelle Schroeder: I love that. You know, it's it's almost a good analog for LinkedIn in and of itself. You know, they took something that used to be so static and cattle call like, which is just like the act of kind of getting a job and, you know, and networking to a certain degree, and they made it human. You know, your profile, your CV can talk. They it's like they gave him a mouth. And so you you see the same type of thing, uh, in what you're describing with service design. Out of curiosity, and I know this can be your opinion and not the opinion of LinkedIn's, do you see LinkedIn as as a social media platform in a way?
[00:11:37] Sam Stern: Yeah. It it is. I mean, if I don't I don't want it to be. Personally, this is my opinion now because I'm I'm off all other social media, and, you know, people lament on LinkedIn all the time to me. Actually, they'll reach out to me, and they'll say it's becoming more like the other places that I left because the conversation on LinkedIn was more professional and interesting.
[00:11:57] Michelle Schroeder: Totally.
[00:11:57] Sam Stern: So I get it. But it is a social media platform. Of course, it is. People are sharing. People are interacting. People are, uh, reacting to posts with likes and, you know, other reaction, uh, emojis. So we can't fight that. I would just say it's a reality of I think it's a reality that goes way back in time, but at one that we've maybe acknowledged more recently, which is you spend most of your waking hours at work. It is part of your relationships, your social world, whether you like it or not. And I don't like, again, speaking just personally, I did meet my wife at work. So and I think a lot of people have done that. It's it's a very common place to find, you know, romantic partners in addition to good lifelong friends. Right? And people you you spend a lot of time with even outside of the office, let alone that you're spending so much time with them in the office. So I think it's not the worst thing that LinkedIn's a social network, but it is a reality that we have to deal with.
[00:12:49] Michelle Schroeder: Totally. And then you take the the the positive with the negative. I think my my feed, for example and I know, like, to your point, a lot of people lament the types of things that people share there, but I've learned so much from my network that I would take 10 of those I would I would take, you know, 10 to one of those posts just for the nuggets and the the the value I get out of it. I guess that one of the things kinda coming out of what you were describing with these human intentionally human moments, do you think that they're going to get more important in this age of automation where a lot of this stuff gets, yeah, just taken over by AI? Um, are you guys leaning in harder on those human elements?
[00:13:25] Sam Stern: I wouldn't say we're leaning in harder necessarily because I do think the the direction of travel is clear that people want more self serve options. Um, and I do think more than ever with with AI, with the interactive and nondeterministic nature of it, it's easier to make digital nonhuman experiences feel more human. And, um, so I do think there's a lot that we can do there. But I I also think given that there are fewer and fewer human moments, the relationship will be more anchored by each one, which I I was saying that earlier. But, um, so we're acknowledging that the importance of the human moments does not go away even as the possibility to make them fewer and further between also exists. It's a sort of a duality.
[00:14:13] Michelle Schroeder: Yeah. Makes sense. It's very, again, very similar. Lots of analogs in our space. Uh, last question for you. As AI gets more embedded into these internal tools, how do you see the role of service design specifically evolving?
[00:14:26] Sam Stern: That's a great question. So we are working really hard right now, um, to democratize our insights, democratize research. And so the first, um, thing we're doing, we've built a service design insights agent, and we've, you know, um, put all of our research into it. We've, you know, uh, primed it with questions that we we shared it with lots of our partners and primed it with questions that it's like, hey. Here's how you get started. Here's how you can start to ask it questions to get value from it. The idea being everyone should have access to our research at any time without having us as gatekeepers or having to initiate a new research project. And I think that the that can continue to evolve where it's integrated into you know, almost as a product manager, I imagine, is writing a PRD, it's in a side panel, almost like that pane of glass we were talking about earlier, to give them real time access to insights about the person that they're building this new product for or new tool for, right, that the PRD is referencing. And it can even then keep going up the stack and suggest things that should be in the PRD in the build that solve the pain points that the AI is helping them understand the user has today. Again, without having to talk to us, but they come from our research. So that's a way to say, the research has been done. It as long as we are and we're sort of staying in the research base to understand, is it still relevant? Is it still valid? If it is, then you should use it in that interactive way to in in all of your work, colleagues of ours. And then the second thing is we can make it easier for them to do their own research. The tools allowed us now where they can have inputted from us best practices about conducting interviews, um, gatekeeping in sort of context windows to not, you know, slip outside of of best practices there and make sure that they're getting the right people. It can potentially even conduct research for them. We haven't quite gotten there yet, but I I know other organizations have, actually, right, sort of synthetic research and synthetic interviews. And then it's because we constantly are getting all these little follow-up questions from our partners. They could ask them themselves. And it's not to say we don't do that for them. We'd hap we'd happily do it, but it goes faster, and they don't have that friction of having to stop their flow and ask us. They could potentially do that themselves directly. So we're trying to enable that and empower that because at the end of all of this, what we want is more of our colleagues to use research and insights and the problems that the employee or the customer or the member, whoever the user is, to guide their decision making, to guide their work. So if democratizing it makes it more likely that they'll do that, we are all in on making that happen.
[00:17:09] Michelle Schroeder: Huge. The insights agent you described reminds me one of our clients, um, very famously, uh, and they're massive, massive, massive clients, like, tens of millions of customers, they will not have a meeting without designating someone in the room to represent the customer, to be the voice of the customer, to act as the customer so that they can give feedback in the room when decisions are being made. And it sounds a little bit like that insights, uh, agent is playing that role. You get to have access to this always on customer in the room and ask, you know, ask it those those questions. I think that's super powerful, um, and it made me have another question, so I lied. Sorry. What would you tell CX leaders that are listening right now, um, if they wanted to improve their employee experience and their customer experience with Agentik AI?
[00:17:56] Sam Stern: I would say the with Agentik AI, the there's so many opportunities to for employees, starting with employees, to take effort out of their processes, we counted we stopped counting at some point. There were so many hours in a week that felt like not the best use of productive time for these employees. And there were so many ways that for example, um, logging activities in the CRM for a sales rep. They've gotta do it. They don't wanna do it. It's it's not really additive to them in the moment, but it's so additive to the overall understanding of our customers and our relationships with them. Of course, We have to do it. But they don't do a good job of it either because it's not that valuable to them. So it's it's these ways to sort of say, what are the things that get in their way? But then also, how can we, using agentic AI, go beyond just the functional automation of a low value task to making it done making, you know, making it be more valuable and better. And I think if you take those two things together, you can actually create a lot more value in the ecosystem. So it's like, we're gonna create a real time transcript from your conversation and log that in the activity, but we're also gonna pull out the key themes and potentially even tag people who need to know about them that you collaborate with internally. And suddenly, this has gone way beyond just activity tracking to something far more valuable. There are endless opportunities for that using agentic AI.
[00:19:16] Michelle Schroeder: Super powerful. Really appreciate you sharing your insights with us. Thank you so much for joining me today.
[00:19:22] Sam Stern: Yeah. My pleasure.
[00:19:23] Michelle Schroeder: How great CX doesn't just happen at the surface level. It's built on systems. It's built on cross functional collaboration. It's built on information and access to that information. Um, again, really appreciate you joining us. Please, if you enjoyed this conversation, review, subscribe, and we'll see you on the next one.