How Front is Helping Companies Cut the Hidden Coordination Costs Slowing Customer Service.

[00:00:04] - [Speaker 0]
What if your customer service conversations are actually one of the most valuable sources of business intelligence inside your company? And despite that fact, nobody is paying attention. Does that sound familiar? Every complaint, support ticket, lost deal, customer question, all these things contain clues about what's working and what needs fixing. And my guest today is Kevin Yang.

[00:00:34] - [Speaker 0]
He's the head of AI at Front, and he's gonna be explaining how AI can turn all these conversations that every business has into better products, better sales, and indeed marketing decisions. And we'll also explore why companies are spending 3 hours coordinating work for every hour spent solving customer problems. It's a shocking stat, that one. So we'll dig a little bit deeper on that and some of their research, and also why the smartest use of AI might be knowing which job should remain firmly in human hands. We got a lot to get through today, so buckle up and hold on tight as I beam your ears all the way stateside so you can join myself and Kevin in a conversation right now.

[00:01:25] - [Speaker 0]
So thank you for joining me on the podcast today, Kevin. Can you tell everyone listening a little about who you are and what you do?

[00:01:33] - [Speaker 1]
Yeah. Absolutely. So I'm Kevin Yang. I'm the head of AI here at Front. For those of you that aren't familiar with Front, customer facing teams, like support, sales, success, use Front to talk to their customers in a unified way.

[00:01:47] - [Speaker 1]
And Front is unique in that it is built for b to b complexity. Right? So, like, the real needs of passing off to another team or talking to another system. And so so that that's that's what's really unique about Front.

[00:02:02] - [Speaker 0]
Well, thank you for taking the time to sit down with me today. And before we get into the technology side of things, I know this is a tech podcast, but tell me a little bit more about your origin story, your journey into AI, and what attracted you to Front, and what can what it was that convinced you that customer operations would become one of the most exciting areas for enterprise AI. In many ways, it feels like you were ahead of your time there. But what was that origin story? I feel there's gotta be one there.

[00:02:29] - [Speaker 1]
Yeah. Absolutely. So applying AI to customer operations actually combines two core themes of my career. So prior to Front, I was a serial entrepreneur for fifteen years. I started my first company in 2010.

[00:02:43] - [Speaker 1]
It was a office lunch delivery business called Eat Club, which got pretty big, delivering something like 20,000 meals a day, $85,000,000 a year in revenue. But in the early days, I took customer service calls directly. Actually, my phone number personal cell phone number was on everyone's credit card statements for four years, so that was fun. But I took calls by day, and I was our software engineer by night. And so I have a deep appreciation for how valuable this customer feedback is.

[00:03:12] - [Speaker 1]
Then about ten years ago, I started a AI voice of customer company called Idomatic, which was way before, you know, AI was cool, right, at the time. You know, we were mostly focused on trying to convince people that the technology could actually do the work that we all take for granted now. But we analyze customer feedback for who's who of tech companies, so, like, Slack, Pinterest, Intercom, HubSpot, Instacart, Instagram, like, Upwork. So, like, before before AI was broadly available, we were classifying what their customers were telling them. And I joined Front because in 2024, Front acquired Ideomatic.

[00:03:51] - [Speaker 1]
And at that point, I joined Front to lead AI product and strategy. And so it's been really fun to use AI to solve the problems that I experienced personally at various points in my career when I was actually, like, on the front lines.

[00:04:07] - [Speaker 0]
Wow. Listening to your story there reminds me a lot of the Steve Jobs quotes, whereas I suspect at the beginning, you wouldn't be able to know how those dots would join up. But looking back, you can see how everything joins up and and joins up for a reason. Right? Does it does it feel that way?

[00:04:22] - [Speaker 1]
Yeah. I mean, I think it I've I'm not the type of person that plans too far in the future, but Yeah. Seeing this opportunity to kind of bring these things I really care about together was was a really unique opportunity.

[00:04:37] - [Speaker 0]
Yeah. Customer conversations, of course, generate an enormous amount of information or data every single day. Yet many organizations still see them as simply as just support interactions, which feels like a real big wasted opportunity there. But tell me more about why you think these conversations are actually one of the richest sources of business intelligence. And, again, how is AI helping organizations trying to uncover insights that a few years ago, they will they would have completely missed.

[00:05:08] - [Speaker 1]
Yeah. So, like, when I was answering customer service calls, it was obvious that support is the moment when your customers are trying to use, configure, troubleshoot your product. Right? It is, like, so much more valuable data than, like, the after action, like, Net Promoter Score surveys. Right?

[00:05:28] - [Speaker 1]
Where people are like, oh, it was good. It was bad. But, like, in support, that is their moment of pain. Right? And so that signal is so valuable.

[00:05:37] - [Speaker 1]
So the the reason I got into this in the first place was that at my food delivery business, it turned out that food quality or food ratings was perfectly correlated with every business metric that we cared about. Right? Retention in account growth, referrals. And so in my mind, I was like, all I wanna do is make the food better, and then everything else will follow. And we were getting a million reviews a year.

[00:06:05] - [Speaker 1]
Right? Customers were were telling us oh, you know, people care love of food, so they have a lot to say about it. And so they were just telling us what we needed to do, but it was all unstructured text. Right? There was just conversations.

[00:06:18] - [Speaker 1]
We didn't really know what to do with it. And that's actually why I started my my AI voice of customer company because I I knew that this was such a visceral need, but there was just no way to do it at scale. Whereas when I was answering all the phone calls and also writing the code, I was able to really have that loop. Right? And so, you know, just just a concrete example here is question is, what is driving food scores?

[00:06:48] - [Speaker 1]
Right? And, anecdotally, it was actually very hard to tell. It's so noisy that we had the whole team that was just running around, handling one offs. But if you're actually able to crunch the data and through all the conversations, we'd actually discover that, there was this massive concentration around, and this is gonna sound gross, but, like, organic greens having bugs in them. And it turns out that this is actually a relatively simple problem to solve, and you basically get I mean, this is way too much detail, but ultrasonic washer, right, that that you run all the greens through, and then you will be able to make sure that there are no foreign objects in the greens.

[00:07:30] - [Speaker 1]
And, you know, that had a larger impact on customer satisfaction for a lower cost than, by orders by, like, order of magnitude compared to anything else that we were trying to do, right, at the food delivery company. And so, like, this is just an example where if you can take all this feedback and quantify it, there's, like, huge business impacts. Right? It'll be gives you clarity that you can do certain things on the business that improve the customer experience.

[00:07:59] - [Speaker 0]
And the insights that we're talking about here often have value that stretches far beyond the support team. And to bring this to life and you don't have to mention any names here, but are you able to share any examples of how information from those customer conversations that we're talking about here is influencing everything from product development, sales, marketing, or even a wider business strategy? Because I think this will really bring it to life.

[00:08:23] - [Speaker 1]
Yeah. So two examples. I mean, I'll even give you examples from within front.

[00:08:27] - [Speaker 0]
Yeah.

[00:08:28] - [Speaker 1]
So one of the things that we care most about is how like, why do we lose deals that we're trying to win? And, also, if customers unfortunately churn, like, why that's the case. Right? And, from a product development perspective, we want to understand those reasons and then, address the make the product better so that, we keep our customers and win more customers. And, we are able to basically analyze every single conversation we're having with customers, whether it's voice or text, and then that is actually driving our product development strategy.

[00:09:08] - [Speaker 1]
Right? And that gives us much better confidence that the expensive technology resources that we're we're deploying to build product are, deployed on problems our customers actually care about. You know, on the sales and marketing side of things, we have done the similar analysis, and we know that if our customers are talking to our team about AI automation use cases, our win rates are literally several fold than if we did not have that conversation. And, you know, that tells us, you know, what is the discovery questions that our sales team needs to be having with customers, right, to understand their pain points and to have this conversation. And then also to tell our marketing teams what is the persona that they have to be targeting so that we have these conversations.

[00:09:56] - [Speaker 1]
Right? Because our sales and marketing efficiency is just much better if we have these conversations. Right? And so those are just like, yeah, practical examples from both product from product marketing sales on how how this customer intelligence is helpful.

[00:10:12] - [Speaker 0]
And many of the AI conversations looking back, what, two to three years ago, they seem to mainly focus on just making individuals more productive. And I think we're finally starting to think bigger than that now and seem to be entering a new phase where the real opportunity is actually helping entire teams work together more effectively. But what does that look like in practice for what you're seeing here?

[00:10:36] - [Speaker 1]
So, yeah, the version that you're you're describing is people have Chatuchi BT or Claude open on the side, and, you know, you're able to, you know, use it as a personal assistant. And, you know, I think that has a ceiling of improving productivity by 30%. And in real life, probably way lower than that because that assumes people are using it all the time every time that they they should be. Right? And it's hard to change people's patterns.

[00:11:05] - [Speaker 1]
Where you companies get a lot more gains is is actually a kind of restructuring of their work so that you you can bucket off all the tasks that are, basically, researcher analysis, which AI is very, good at. Right? And then you basically change the workflows so that the research and analysis happens automatically. And so to to give you an example, you know, for in the customer operations perspective, there are frequently questions like, where is my order? Right?

[00:11:37] - [Speaker 1]
Or and there are complex versions of this. Like, we have customers that, are shipping, shipping containers from one side of the world to another, and that's, like, a much more complicated question than if you are, you know, sending something through FedEx. But the point is that what's happening in the human version of this is someone's going into that third party system, looking it up, and then basically coming back with the answer. And we, like, what AI can do is automate, like, the interpretation of the question, figure out how what is the research you would do, go to that third party system, find that information, and respond to that. And we're basically taking research and analysis tasks and automating them, and then making it so that the humans are able to really focus on the the conversations that require judgment and empathy and relationships.

[00:12:30] - [Speaker 1]
Right? And, by kind of taking the different tasks and putting them in, hey. Is this research analysis, or is it, like, a a more fundamentally human interaction? That's where you really get a lot of gains.

[00:12:45] - [Speaker 0]
And before you join me on the podcast today, I came across a breathtaking stat in your coordination tax research because it found that teams were spending three hours coordinating work for every hour spent solving a customer problem, which just completely blew my mind. So why is this hidden operational cost become so widespread, and why are organizations not measuring it? Help me see sense in this.

[00:13:12] - [Speaker 1]
Yeah. Yeah. So this is particularly true in b to b companies, but a lot of the work crosses teams. Right? Like, recently dealing with a situation where we, you know, had a commercial deal, and then it had to touch finance, and it has to touch sales.

[00:13:30] - [Speaker 1]
And then, you know, the analogy that I would draw is, like, if you're, you know, driving on the highway in a traffic jam, like, you could really accelerate. Right? And but if the car in front of you is stopped, then, like, you have not sped up the thing at all. Right? And this is basically what's happening when there's lots and lots of teams involved.

[00:13:48] - [Speaker 1]
This is just the nature of things. The the core research like, the stat that you're referencing, right, three hours coordinating per hour of solving, that's really the baseline of what's happening. In our research, like, nearly 70% of companies reported spending more than twice as much time coordinating than than solving. And, you know, I think the the reason this happens is that many platforms do are not built for this multiteam, setup. Right?

[00:14:22] - [Speaker 1]
And so they assume that there's gonna be one customer and one agent and one answer. And when the the platform, doesn't when that's not actually the case, then folks take the the work off platform. Right? And so you have all these, like, off platform emails and Slacks and meetings and, you know, researching other systems that, where where all this work is, happening outside the customer operations system. And, you know, our view, of course, is that you want all of that in one place, and you wanna make it so that the steps that, like I was mentioning earlier, are just research and analysis just happen automatically at the right time so that you don't have to wait, right, for yet another party.

[00:15:09] - [Speaker 1]
And, you know, I think the last point you made was, you know, our our team's tracking this. Right? And what teams are doing right now is they're tracking the outcome metrics. Right? You know, resolution time or CSAT or first response time.

[00:15:25] - [Speaker 1]
But, yeah, they they really do not have a good system of tracking this coordination because all of that is scattered all over the place. And for us at front, like, obviously, we wanna drive that number down, but we also wanna put that all that coordination in one place, which we we do certainly within our own system.

[00:15:46] - [Speaker 0]
And I've been to, I think, 16 tech conferences this year so far. Predictably, every single one of them is themed around agentic AI and agents. And as organizations will inevitably introduce more and more AI agents into their customer operations, I would imagine governance will become just as important as capability. So on that side of things, what should leaders be thinking about when it comes to visibility, oversight, and ensuring multiple AI agents can work together safely and most importantly responsibly?

[00:16:19] - [Speaker 1]
Yeah. One thing about AI that's remarkable I mean, for folks, your listeners that are playing with AI on the weekend, it is remarkable what you can do in, you know, fifteen minutes or two hours. Right? You can build something that kind of does the thing that you're asking it to do. And what companies find out though as they're trying to operationalize these prototypes is that you actually have trouble doing things like measuring okay.

[00:16:51] - [Speaker 1]
How often is it actually correct? Right? Or, okay. I see that there's this behavior that I don't like, so I go, you know, into Claude or Chetraputy, and it's like, hey. Make it not do this anymore.

[00:17:03] - [Speaker 1]
But you don't know what other ramifications it has. Right? Is it making it perform worse in other contexts? And so that's, like, what is required. Those types of things are what's required for governance of of AI agents.

[00:17:20] - [Speaker 1]
And, you know, that's that's if you have one that's if you have a single agent, this really compounds when you have multiple agents. Right? And so our our view here is the right level of governance requires you to be able to measure a few things. First of all, you wanna measure, is it is it following the processes that you you have? Right?

[00:17:48] - [Speaker 1]
And so we think of that traditionally in the human context as QA. The second thing is, is the customer happy? Right? And so you can kind of look at what the customer's reactions. And then, you know, traditionally, the human analog of this is CSAT.

[00:18:03] - [Speaker 1]
And then finally, which I think is the most interesting part, is trying to look at all the signals where maybe something went wrong with the agent and then diagnosing, okay. Why did it go wrong? Is it because the agent didn't have a certain piece of knowledge, or, did it not have guidance of how to handle a certain situation? And then, basically, providing, you know, the the ongoing analysis of, conversations and when it went wrong to say, how do we improve the agent? Right?

[00:18:35] - [Speaker 1]
And we think that that constitutes the governance layer that you really need in order to operate agents.

[00:18:42] - [Speaker 0]
And I'd love to give people listening even more valuable takeaways before we sign off for the day. So for any business leader listening who might want AI to improve how their organization works together rather than just adding another tool into the mix, practical steps that you'd advise that they take over the coming months to maybe reduce coordination, friction, and build workflows that genuinely improve both employee and customer experiences? Anything you'd you'd advise there?

[00:19:11] - [Speaker 1]
Yeah. Yeah. So one of the failure modes is just trying to automate the human process. Right? But there are some things AI is good at, research and analysis, and some things that is not good at, which is, you know, relationships, empathy, judgment.

[00:19:26] - [Speaker 1]
And so what you actually wanna do is take whatever function, you know, you lead and understand where the work happens. And, you know, this is an exercise you can do in, you know, thirty minutes, but you, like, create multiple columns. Right? Well, first column is work that research and analysis should be fully automated. Next column is, okay.

[00:19:48] - [Speaker 1]
Like, AI can do a lot of it, but requires human review. And the third column is things that should just be human. Right? So, like, building relationships. And then where companies will get the most gains is focusing their energies on automating that first column and then, you know, some of the second.

[00:20:06] - [Speaker 1]
And when company like, we we have experience working with many customers that when things fall short, it's because they're trying to do the third column. Right? The relationship stuff with AI, and it just, like, is not there yet. Right? And it's not it it, like, just doesn't have the context or the judgment.

[00:20:25] - [Speaker 1]
And so being able to focus AI on on the task that's really good at is really the place to start.

[00:20:33] - [Speaker 0]
And for yourselves moving forward, anything excites you about the road ahead? What makes you wanna jump out about in the morning? What excites you about everything you're working on?

[00:20:42] - [Speaker 1]
So I would say about eight months ago, AI coding got a lot better. Right? And you see this with the massive explosion of Anthropix revenue. And, fundamentally, what's going on here is certainly, the models are getting better, but there's this concept of a a loop, right, which is the AI agent that doesn't just, you know, follow set a plan and follow the plan linearly, but it kind of self corrects while it's going along. It will try to do some research, and then if it didn't have get everything it needed, it'll try harder.

[00:21:22] - [Speaker 1]
And you see that, like, this has had a massive impact on coding. Coding is way better. But this actually is helpful for every other piece of knowledge work, including, you know, customer communications. And so what I would say is that we are seeing just massive explosion or or, like, inflections in AI quality, in a way that I don't think we have seen since the introduction of Chatuchu BT in 2022. Right?

[00:21:50] - [Speaker 1]
And so there's, like, another moment of renaissance right now, empowered by the the latest models and also kind of this agentic loop that I just described. And, being able to work on this, when it just kind of works is, creates almost like a childish, like, a fascination and curiosity. You're like, I can't believe this works. And so, yeah, I mean, that is my experience right now building building in this arena.

[00:22:18] - [Speaker 0]
Love it. And for anybody listening that would like to read more about some of your musings, if you're publishing on LinkedIn, etcetera, or find out more about the announcements that could be coming out of Front this year. Where where would you like me to point everyone listening?

[00:22:31] - [Speaker 1]
Yeah. Front's website, front.com, but the coordination tax report is actually something that's really interesting to read. That's available at research.front.com.

[00:22:41] - [Speaker 0]
Awesome. Well, I will include a link to everything you mentioned there, including that report that we revealed earlier. I found that teams spent almost three hours coordinating work for every hour spent solving customer problems. I really do urge people to check that out and connect with you on LinkedIn too. But more than anything, thank you for shining a light on this today and sharing your story with me.

[00:23:01] - [Speaker 0]
Really appreciate your time.

[00:23:02] - [Speaker 1]
Absolutely. Great talking to you.

[00:23:04] - [Speaker 0]
I think today's episode offer practical way to think about AI at work. And Kevin explained why the biggest gains might not come from just giving everyone another chatbot, but from redesigning workflows around what AI does well and what people do better. And by that, I mean, automate the research and analysis, then use human review where it adds value, and finally, protect the work that depends on empathy, judgment, and relationships. Sounds simple. Right?

[00:23:38] - [Speaker 0]
But so many still get this completely wrong. And I'd to hear your thoughts. How much time is your business spending coordinating work instead of getting to the root cause and solving your customers' problems? And could AI help give some of that time back? Let me know.

[00:23:57] - [Speaker 0]
Techtalksnetwork.com. If you go over to the blog post associated to this episode, there'll be lots of things for that we talked about today, including how to connect with Kevin, learn more about Front, follow them on LinkedIn, all that kind of stuff. But I wanna hear from you. I wanna hear about your stories. What's working?

[00:24:15] - [Speaker 0]
What isn't? The leading issue of agentic AI in businesses right now is ensuring agents act with compliance guidelines. And Denodo applies guardrails across your entire data estate. By aligning your company's data infrastructure under one system, these guardrails perform consistently across your platform. So start scaling your business and start with Denodo.

[00:24:42] - [Speaker 0]
Simply visit denodo.com to learn more. And while you walk away and have a think about that, I'm gonna prepare for another guest tomorrow. I'll be back in your podcast feeds bright and early tomorrow. Hopefully, I'll get to speak with you again then. Thanks for listening.

[00:24:59] - [Speaker 0]
Bye for now.