In this episode
Kevin Yang, Director of AI at Front, argues that the operational cost most customer-facing organizations never measure — the coordination tax — is the reason AI investments improve individual productivity without improving business outcomes. Front’s research across more than 700 customer service operations and account management leaders found that teams spend roughly three hours coordinating for every hour they spend actually solving a customer’s problem, and that the organizations with the most advanced AI adoption report both the highest satisfaction with their technology and the highest number of coordination issues. Yang and host Greg Kihlström work through where that hidden cost shows up, why layering automation on top of a broken workflow amplifies the problem, how to split work between AI and humans by task type rather than by function, and what governance a multi-agent environment actually requires.
Key takeaways
- Teams spend three hours coordinating for every hour spent solving customer problems. Front surveyed more than 700 customer service operations and account management leaders; coordination consumes the vast majority of the time, and almost none of it is measured.
- The coordination tax is the hidden cost of work that spans people, teams, and systems. Organizations measure response time, resolution time, and CSAT — the outcomes — but not the coordination required to reach them.
- The most automated organizations report the most coordination friction. AI-first organizations simultaneously report the highest satisfaction with their technology and the highest incidence of coordination problems.
- AI accelerates the individual, not the system. Yang’s analogy: you can accelerate hard in traffic, but if the car ahead is still stuck, your progress toward the destination hasn’t improved at all.
- Automation alone does not solve coordination. AI performs best embedded in workflows with access to full customer context, the right systems, and clear ownership across teams — throwing AI into an unchanged workflow does not produce impact.
- Split work by task type, not end to end. Research and analysis tasks go to AI, which does them more thoroughly in virtually no time; empathy, context, and relationships stay with people; judgment tasks sit in the middle, where AI drafts and a human decides.
- Over 40% of organizations do not measure coordination at all, and only 5% track all three core measures. The three buckets are handoffs, coordination time, and duplicate work — and the 5% that track all three are the organizations that spend more time solving customer problems than coordinating.
- Coordination is hard to measure because it happens across systems. A second window or a walk down the hall leaves no trace; coordination becomes measurable only when the handoff happens inside the system where the work lives.
- Production-ready agents require two distinct kinds of trust. Trust that the agent behaves as expected in situations nobody explicitly tested for, and trust that the interactions it produces are high quality enough to put in front of a customer.
- AI agents make coordination failures worse because they propagate mistakes system-wide instantly. Multiple humans in a workflow already create coordination challenges; adding one or more agents raises both the speed and the blast radius.
- Accountability does not transfer to the agent. In account management, AI can prepare a quarterly business review by analyzing product usage, open tickets, sentiment, and champion turnover — but the human account manager still owns the renewal and the expansion.
Chapters
- 0:00 — The question: is the AI you bought making the team harder to coordinate?
- 2:05 — From Idiomatic to Front: Kevin Yang’s path into AI product and strategy
- 3:21 — What Front is, and why it’s built for B2B complexity
- 3:41 — Customer conversations as the richest source of business intelligence
- 4:57 — What Front learned analyzing its own sales conversations
- 5:37 — The coordination tax, defined
- 6:06 — The research: three hours coordinating for every hour solving
- 7:21 — Why the most automated organizations report the most friction
- 9:34 — Mapping the workflow: which tasks go to AI, which stay with people
- 11:45 — Why more than 40% of organizations never measure coordination
- 12:30 — Handoffs, coordination time, duplicate work — and the 5% who track all three
- 13:44 — Bring your own agent: governance in a multi-agent environment
- 15:20 — The two types of trust between prototype and production
- 16:52 — Smart CSAT and evaluating every agent conversation
- 17:51 — What stays human: accountability, not skill set
- 20:03 — One year out: agent proliferation as a live operational problem
- 20:50 — How Kevin Yang stays agile
Why customer conversations outperform surveys as business intelligence
Yang’s argument starts from the front line: support, success, and sales conversations capture customer pain at the exact moment someone is trying to buy, use, or troubleshoot a product. That richness is not recoverable from an after-the-fact survey of the kind sent after a hotel stay or a flight. Those conversations tell an organization what customers need, where operations break down, and what the business should do next — but most organizations treat them as tickets to close rather than as intelligence to mine. Front analyzes its own sales conversations and found that close rates are several times higher when AI automation comes up in the conversation, which shapes both discovery questions and which companies the team should be targeting.
What the coordination tax actually is
The coordination tax is the hidden operational cost organizations pay when resolving a single customer issue requires touching multiple people, teams, and systems. Companies measure the outcomes — response time, resolution time, CSAT — but not the coordination required to produce them, which leaves the largest consumer of time invisible. Teams carrying a heavy coordination tax describe their day as finding the right person or system, managing handoffs, re-explaining context, and closing loops between teams, rather than fixing the customer’s problem.
Why more automation can produce more friction, not less
Research shows AI is genuinely good at improving individual productivity — an all-knowing assistant to think against and to hand knowledge work to. Organizational improvement is much harder. Yang’s traffic analogy makes the mechanism concrete: accelerating is useless when the car ahead hasn’t moved. That is what happens when AI speeds up one participant inside a workflow that still routes through several teams. Front’s research bears it out — the organizations with the deepest AI adoption report both the highest technology satisfaction and the most coordination issues. The conclusion Yang draws is not that AI fails; it’s that automation alone doesn’t solve coordination.
How to map a workflow between AI and people
The trap is taking a human workflow in which a person bounces between research, analysis, and judgment, and trying to automate it end to end. The alternative is to break the workflow into task types. Research and analysis go to AI, along with access to the third-party systems it needs to run those tasks autonomously. Empathy, context, and relationships stay with people. In between sit judgment tasks, where AI can produce a draft and a human still decides. Restructuring the workflow this way is what lets an organization capture the gains AI is actually capable of, instead of being frustrated that it isn’t good at the relationship work everyone already knew it wasn’t good at.
What “measuring coordination” means in practice
The measurable buckets are the number of handoffs, coordination time, and duplicate work. They’re hard to capture because coordination usually happens across systems — the second window, the walk down the hall — and leaves no record in any of them. When the virtual tap on the shoulder happens inside the same system where the customer work lives, it becomes measurable. Only 5% of the companies Front surveyed track all three measures together, and those are the companies that have brought the coordination tax under control.
Governance for a bring-your-own-agent environment
It has never been easier to get AI to do approximately what you want, which is a real advantage for operators who understand business processes but aren’t technical. The risk is that agents propagate mistakes system-wide instantly. Front’s bring-your-own-agent approach targets the two governance gaps between a prototype and a production agent: capabilities that keep agents following guidelines and prevent updates from introducing unexpected behavior elsewhere in the system (analogous to unit tests in software), and Smart CSAT, which evaluates customer sentiment on every interaction rather than the small sample a human supervisor could review. Front invested in its own agent, Autopilot, but extends the same governance layer to agents customers build elsewhere.
What stays with the human
Yang makes the division concrete through account management. AI takes the research and analysis: preparing for a quarterly business review by analyzing product usage, open support tickets, the sentiment of communications, and whether champions have left or changed jobs. The human account manager remains responsible for the renewal and the expansion. The line is drawn on accountability, not skill.
What Yang expects a year from now
Agent proliferation moves from an emerging pattern to a problem companies are actively grappling with — headlines about organizations carrying real battle scars from multiple AI agents that each do something useful independently while collectively creating a mess for customers.
FAQ
What is the coordination tax? It’s the hidden operational cost organizations pay when customer work spans multiple people, teams, and systems, requiring handoffs and context re-explanation before an issue can be resolved. Most companies measure the outcome metrics it affects — response time, resolution time, CSAT — without measuring the coordination itself.
How much time do customer-facing teams actually lose to coordination? Front’s survey of more than 700 customer service operations and account management leaders found teams spend an average of three hours coordinating for every hour spent actually solving customer problems.
Why do organizations with the most AI automation report the most coordination problems? Because AI improves individual productivity without changing the workflow structure around it. AI-first organizations report the highest satisfaction with their technology and the highest incidence of coordination issues at the same time — accelerating one participant doesn’t help when the work still routes through several teams and systems.
Which tasks should go to AI and which should stay with people? Research and analysis tasks go to AI, which handles them thoroughly and fast when given access to the systems it needs. Empathy, context, and relationship work stays with humans. Judgment tasks sit in between, where AI drafts and a person decides.
What should a team start measuring to see its coordination tax? Handoffs, coordination time, and duplicate work. Only 5% of surveyed companies track all three together — and those are the companies spending more time solving customer problems than coordinating.
What governance does a multi-agent environment require? Two kinds of trust: that an agent behaves as expected even in untested situations, and that its interactions are high quality enough for customers. The first is addressed with guideline enforcement and regression-style checks that catch unexpected behavior after updates; the second with automated evaluation of customer sentiment on every interaction rather than a sampled review.
About Kevin Yang
Kevin Yang is the Director of AI at Front, the only customer operations platform built for B2B complexity, where he leads the company’s AI strategy across automation, analytics, and customer insight. His work is grounded in the belief that the most valuable business intelligence lives inside customer conversations. At Front, Kevin is building AI systems that operate within real workflows—helping over 9,000 businesses better understand what their customers are saying and translate those insights into action with full visibility
Kevin joined Front through the acquisition of Idiomatic, the AI-powered voice-of-the-customer platform he founded and led as CEO. Earlier in his career, Kevin worked in frontline support while building his first company, EAT Club, a leading virtual cafeteria serving over 1,000 companies, including Google, Netflix, and Tesla.
Kevin Yang on LinkedIn: https://www.linkedin.com/in/yangkevin/
———- Resources ———-
Front: https://www.front.com
Front is a customer operations platform that helps organizations manage customer communication across email, chat, voice, SMS, and other channels in one place. The platform is designed for teams handling complex customer relationships, where resolving an issue often requires coordination across multiple people, systems, and departments.
We’re proud to be a media partner for #MAICON26 – Oct. 13-15! Learn how AI can power your marketing and business and help you grow smarter. Use code AGILE150 to save! https://aglbrnd.co/r/7fe458ced0f04658
Reach your customers with Reddit. Spend $500 in ad spend, get $500 back in ad credit! Learn more: https://advertalize.com/r/491818c79fb1873f
Chaser is the only Slack-native project management platform that helps teams turn messages into tracked tasks, automate follow-ups, and maintain team-wide visibility, without adopting another tool. Now integrated with Claude and other GenAI tools. Learn more at trychaser.com and use code AGILEBRAND for a 3-month free trial (normal trial is 14 days).
The most influential minds in software, AI, and engineering leadership will be at WeAreDevelopers World Congress North America, September 23-25 in San Jose. Learn more: https://aglbrnd.co/r/60a7299222a7bcf1
Start building your own apps with Replit and get $20 off. Learn more: https://aglbrnd.co/r/93531742a7625a20
Enjoyed the show? Tell us more at and give us a rating so others can find the show at: https://aglbrnd.co/r/faaed112fc9887f3
Connect with Greg on LinkedIn: https://www.linkedin.com/in/gregkihlstrom
Don’t miss a thing: get the latest episodes, sign up for our newsletter and more: https://aglbrnd.co/r/35ded3ccfb6716ba
Check out The Agile Brand Guide website with articles, insights, and Martechipedia, the wiki for marketing technology: https://www.agilebrandguide.com
The Agile Brand is produced by Missing Link—a Latina-owned strategy-driven, creatively fueled production co-op. From ideation to creation, they craft human connections through intelligent, engaging and informative content. https://www.missinglink.company
Transcript
[00:00:00] Greg Kihlström: Hi, I’m Greg Kihlström, host of The Agile Brand, and here’s a question for you. What if the AI tools you bought to make your team faster are quietly making that team harder to coordinate? Agility depends less on how quickly you adopt new technology and more on how well you understand the work that technology is supposed to improve. Today we’re going to talk about the operational costs buried inside customer-facing work and why they’re so hard to see, let alone measure, why layering more automation onto a coordination problem often makes the problem worse, and what it takes to build AI into core workflows so that efficiency and customer experience improve together instead of trading off against each other. To help me discuss this topic, I’d like to welcome Kevin Yang, Director of AI at Front. Kevin, welcome to the show.
[00:02:03] Kevin Yang: Hey, great to be here.
[00:02:05] Greg Kihlström: Yeah, looking forward to talking about this with you. Definitely a, a timely topic here, a lot- well, I’m sure on a lot of people’s minds. Before we dive in, though, why don’t you give a little background on yourself and your role at Front?
[00:02:17] Kevin Yang: Yeah, absolutely. at Front I lead AI product and strategy. I joined Front about two years ago when Front acquired my AI voice-of-customer startup called Idiomatic. And building products for customer-facing teams is really meaningful for me because in my 15 years as entrepreneur before joining Front, I’ve been both on the front lines and also built, customer support, success, and sales teams. So I have tremendous empathy for the problems that we’re solving. At the same time, I’ve also been building AI products for a decade, starting well before the current wave, so being able to apply this technology to, problems I’m passionate about, is a really great place to be.
[00:02:55] Greg Kihlström: Yeah. Yeah, yeah, you’ve been, you’ve been building with AI before it was AI, [laughs] right? So before it was ubiquitous, right? So-
[00:03:01] Kevin Yang: When it was machine learning.
[00:03:03] Greg Kihlström: Right, right, right, yeah. I mean, AI’s been around for decades, I know, but it’s like, you know, the, that, that recent wave just, yeah, I think it’s on everybody’s minds now, right? and, and for those that, that aren’t as familiar with Front, can you tell us a little bit about, you know, what, what it does, who it’s built for?
[00:03:21] Kevin Yang: So Front is the customer operations platform that’s referred by over 9,000 companies, for functions like support, success, and sales. And it’s really unique, in that it’s built for the daily realities of complex B2B work, whereas many other tools, kind of dodge the complexity or try to automate it away.
[00:03:41] Greg Kihlström: Great. Great. Perfect. Well, let’s, let’s dive in here, and we’re gonna talk about a few things here, but wanna start with this idea and this concept of conversations with customers as business intelligence. And so you’ve made the case that customer conversations are one of the richest sources of business intelligence, but most organizations treat them as tickets that need to get closed. What’s actually sitting in those conversations that a dashboard or a survey is never really gonna capture?
[00:04:12] Kevin Yang: one thing I learned from being, you know, taking customer service calls on my cellphone, at my first business, is that these conversations from support or success or sales capture the pain that customers have at the very moment they’re trying to buy or use or troubleshoot the product. Right? And the level of richness in these conversations is so much deeper than the after-the-fact surveys that, you probably get what, the last time you stayed at a hotel or flew on airplane. And, you know, this information tells you what customers need, where operations break down, and what the business should do next. So, you know, here at Front, you know, we have some examples of this. For example, we analyze all of our sales
[00:04:57] Kevin Yang: conversations, right, with, with prospects, and we have– we’re able to glean from this that if we talk to the customers about, of AI automation, our close rates are several times, [laughs] than when we don’t have that conversation, right?
[00:05:13] Kevin Yang: And having that intel is really helpful in both in terms of, trying to drive, you know, conversations and discovery questions to actually have those right conversations, but also, tells us something about the types of companies that we, should be targeting, right? And so that’s also helpful from a marketing perspective. And so, like, that’s just an example of, insight that we can get that really helps us change the trajectory of the business.
[00:05:37] Greg Kihlström: Yeah. Well, and, and so your– so- some of your research also introduced the concept of a coordination tax. you know, I think, I think we’re, we’re used to things like technical debt and, and other things like that, but this coordination tax is, it’s the hidden cost in customer operations. where does that cost actually show up day to day, and, and what does that tell, you know, what, what’s the tell that a, that a team is paying it?
[00:06:06] Kevin Yang: so yeah, as you r- mentioned, coordination tax is kind of this hidden operational cost that co- organizations pay when they have work that spans, different people, different teams, and different systems, and you have to touch all these different pieces before you can resolve a customer issue. many companies or, like measure core metrics as, like, response time, resolution time, CSAT, but they don’t actually measure the coordination, required to achieve these outcomes. And so, and in our research, we’re finding that that is actually taking up the vast majority of the time, so having that be invisible is, problematic. So the way we did this research is we surveyed over seven hundred customer service
[00:06:51] Kevin Yang: operations and account management leaders. and what we found is that on average, teams are spending three hours coordinating for every hour they’re actually spending solving customer problems. And, y- you know, to your question of what it actually looks like, the teams that are afflicted with heavy coordination and taxes say that they spend most of their time finding the right person or system, managing handoffs, re-explaining context, and then closing loops be- between the teams rather than actually fixing the customer problem.
[00:07:21] Greg Kihlström: Yeah. Yeah. Well, and, and I think one, one of the surprising things there is some of the organizations, a- according to that research, some of the organizations with the most advanced automation are reporting the highest number of coordination issues. So, you know, usually you, you, you go into automation trying to, to solve everything, right? [chuckles] And, and seems like the coordination tax is, is particularly prevalent there. What’s happening in the, inside those organizations that makes more automation produce more friction rather than less?
[00:07:54] Kevin Yang: Yeah, it’s an interesting dichotomy, in the sense that at the individual level, research shows, and this is backed up by probably every- your audience’s individual experience with AI, is that AI’s really good at pr- improving individual productivity, right? I mean, you have this kind of almost all-knowing, assistant, to, to bounce ideas off of, to do some, some, knowledge work for you. but even despite the individual productivity, the organizational improvement is much more difficult. And, you know, the analogy that I would use is, like, you’re on the road, there’s a lot of traffic, and yes, you could accelerate and go, t- you know, really fast.
[00:08:35] Greg Kihlström: Right.
[00:08:35] Kevin Yang: But if the car in front of you is still stuck there, then, like, the overall, your progress towards your destination hasn’t really gotten any faster, right? And so I think this is what happens when you have AI accelerating individual productivity in a workflow that still has all these different teams. A- and, and so we see this in our research as well, like AI first organizations, the ones with the most adoption of AI, do simultaneously report the highest satisfaction with their technology, but also, reporting more coordination issues as a result, right? and you know, the takeaway isn’t that, hey, AI doesn’t work, it’s that automation alone doesn’t solve coordination. AI actually performs best when it’s embedded in the workflows with access to full customer contacts,
[00:09:20] Kevin Yang: the right systems, and clear ownership across teams, and you can’t, expect to just throw AI into the mix and then, without changing the, the workflows and expecting it to actually, have a, make an impact.
[00:09:34] Greg Kihlström: Yeah. Yeah. So for leaders that are trying to avoid all of the, you know, that, that coordination tax, what does… You know, I d- I would imagine mapping things out, trying to understand where some of those friction points are and those work- workflows would, would be part of that process. But what, what does a useful map of that look like, and how precise do you have to get about what each piece of AI, you know, technology is doing versus what a person is still deciding?
[00:10:07] Kevin Yang: So I think the trap pe- companies fall into is they take a current human workflow where humans are bouncing back and forth between doing research and analysis and judgment and back, to research, and they’re just trying to automate that whole workflow end to end.
[00:10:23] Kevin Yang: And what we think, you know, to, to your point about mapping, what companies should be doing, is saying, “All right, here are the research and analysis tasks that AI is just excellent at, right? And it will do, more thoroughly in virtually no time. And then over here we have another bucket that is empathy and contacts and especially relationships, right? That, it is st- at this point, AI, is not quite able to, grok all that. And, you know, somewhere in the middle you probably have things that requires judgment, like AI can take a draft of it and still require some human judgment. but yeah, I, I would say research and analysis is on one side, relationships is on the other side. And what companies really need to do is, you
[00:11:08] Kevin Yang: know, map their workflows and actually break the w- break it down so that they’re assigning AI the tasks that are research and analysis tasks, and giving it all the, you know, third-party systems required to, for it to do this autonomously. And then also preserving the, relationship, tasks for, for the humans, right? And, by kind of restructuring the workflows, they’re actually going to be able to reap the rewards and the gains of what AI is capable of without being frustrated that AI is not good at the relationship stuff, which, you know, we kind of knew in advance.
[00:11:45] Greg Kihlström: Right. [chuckles] Right. Yeah. Yeah. You know, another part in, you know, I d- I would even say diagnosing or improving this over time is, is the measurement component of this, and I think this is another place where, you know, the, the research showed that over forty percent of organizations aren’t even measuring coordination. So, you know, if, if you’re not measuring it, you know, what, what should a- … teams start counting as coordination and what makes it more difficult to, measure than things like volume or thing- resolution time, things like that?
[00:12:24] Kevin Yang: Yeah, like, I think you’re alluding to, like, the, the old saying, “You can’t improve what you don’t measure.”
[00:12:29] Greg Kihlström: Right. Right.
[00:12:30] Kevin Yang: And, you know, the, the buckets of coordination work that, we see companies measuring are, the number of handoffs, the coordination time and duplicate work. And the reason that these are hard to measure is that frequently these happen across systems. And so if they’re happening across systems, like if you, you know, have your, whatever your customer communications platform over here, but actually you’re opening up another window or, like, going down the hall to talk to someone, of course it’s hard to measure that, right? And so one of the nice things about Front, you know, not to plug it, is that it all happens within the same system. So, like, the virtual, like, tap on the shoulder happens inside, inside, the system
[00:13:15] Kevin Yang: and, you can actually just measure that, right? But if you’re able to measure the handoffs, coordination time and, and duplicate work, the companies that… You know, w- we see companies doing, some subset of this. Only 5% of all companies that we surveyed actually track all three of those together. but those are actually the companies that have been able to conquer the coordination tax and, are the companies that actually spend more time solving customer problems than coordinating.
[00:13:44] Greg Kihlström: Got it. Got it. And so as more teams are building or buying or sometimes both, these specialized AI agents, the risk of creating new silos seems pretty high. You know, again, there, already there’s, there’s coordination tax and things like that. You, you’ve talked about, bring your own agent initiative. How does that concept address some of the need for governance and shared source of truth in a multi-agent environment?
[00:14:17] Kevin Yang: to- it’s never been easier to get AI to do approximately what you want, right?,Or, you know, you, you can just fire up, you know, Claude or Codex and, give instructions, and it is, it is remarkable. And I think this is a great development for operators because, if you think of operators as the people that understand the business processes that, you know, power these organizations, they now get to work with building the AI automation without having to be that technical, right? And, being able to just talk to yourself makes… or in this case AI, is way faster than trying to mobilize like a technical team to, to get your ideas, into code. but as we talked about earlier about the coordination tax, just having multiple humans involved in a workflow creates,
[00:15:08] Kevin Yang: you know, coordination challenges, and this can get much worse when you introduce one or more AI agents into the mix, because they can propagate mistakes, system-wide, instantaneously, right?
[00:15:20] Greg Kihlström: Right. Yeah.
[00:15:20] Kevin Yang: And so, that’s the reason why it, it, yeah, it’s, it’s a big challenge. Now, the, there are two big governance gaps between a prototype of an AI agent and one that’s, production ready, and, it’s two types of trust, right? The first type of trust is that the AI agent’s behaving, as expected, even in situations that you didn’t explicitly test for. and then the second is trust that’s providing high-quality interactions and the end users are happy and they’re not going to get pissed at you, because you’re having them talk to AI. And so as part of Front’s bring your own agent initiative, we offer capabilities to make sure that agents are following guidelines, and we’re also working on capabilities
[00:16:06] Kevin Yang: to make sure that updates to agents don’t introduce unexpected behavior else- where in the system. All right? the way you can think about it is that in coding there are unit tests where you can say, “Hey, regardless of which code I change, I want this to happen.” You could basically do something analogous, for, for AI agents as well. and then so, so that’s how we address the first, requirement, which is being confident that it will behave as expected. And the second requirement, trusting that it’s providing high-quality interactions, we have, you know, I think last year, launched a product called Smart CSAT that evaluates customer sentiment on every interaction. And at that time we built that for, for human interactions because,
[00:16:52] Kevin Yang: you know, the, the norm was for human, you know, customer support supervisors to, like, listen to a s- small subset of conversations or read a s- small subset of conversations. now that they’re AI agents, we actually wanna do the same thing. We want, to have AI, reviewing every conversation that AI or that any agent’s involved in and be able to assess whether or not the, the end customer was happy with, with the interaction, right? So that’s the other part of the governance layer.
[00:17:24] Greg Kihlström: Yeah.
[00:17:24] Kevin Yang: Now, at Front, we’ve certainly invested a lot in our agent called Autopilot that lets companies build their agents easily inside Front, but we also see a lot of comp- customers building agents themselves, Claude, Codex or other systems, and we wanna support them as well. So with the bring your own agent initiatives, we’re just saying that these governance features should apply to both our f- our conversations, our agent conversations, and external agent conversations as well.
[00:17:51] Greg Kihlström: Yeah. Yeah. Makes, makes a lot of sense. and you know, I know a lot of the conversations here about automation and, you know, agentic AI, all of this are, are really centered on what can we get AI to do and, and the, the technology to do. Wanna talk a little bit about the, the human part of this as well. And so, you know- Some, some of the coordination tasks and, and all that aside, you know, that, that’s going to get more sophisticated and, and smoother over time, and agents are gonna be taking over more of the coordination over time. when this happens or, or even now, you know, what, what, what does stay with the person and what should stay with the, the humans on the team? And,
[00:18:36] Greg Kihlström: you know, less so from a skill set standpoint than from an accountability standpoint.
[00:18:41] Kevin Yang: So earlier we had drawn these two buckets, right? Saying AI is good at research and analysis, humans are good at relationships, empathy, context. So let’s, like, make this real by talking about in the context of a specific function. Let’s call it account management, where, you have a book of business, of c- customers that you want to renew and, expand. And, you know, ultimately the humans are using, AI as a tool. The human’s re- responsible for the outcomes here. but the types of things that you make the AI responsible for in research and analysis for the account management context is preparing for meetings. so suppose you have a quarterly business review coming up. Being
[00:19:27] Kevin Yang: able to analyze all the product usage, all of the, you know, the usage, the communications to see what open support tickets there are, the sentiment of all the communications, whether champions have left or, or changed jobs. Like, all of this stuff, is things that you can delegate to the AI, right? But at the end of the day, the human account manager is responsible for whether they get the renewal, whether they get the expansion, right? And so I think, you know, that’s the type of thing that, th- th- that’s like an example of the division of labor that you can see.
[00:20:03] Greg Kihlström: Yeah. Yeah. Love it. Well, Kevin, thanks so much for joining today. A couple last questions as we wrap up here. First one, if we were having this interview one year from today, what is one thing that we would definitely be talking about?
[00:20:18] Kevin Yang: So right now we’re talking about this, we’re starting to see the beginning of this, but agent proliferation will be a full-blown problem, that companies are grappling with actively. you will see, you know, headlines, I’m sure, of companies with real battle scars from multiple A- AI agents that independently do useful things, but create a mess for, for their customers. and, I think it’s just a matter of time, before, before this, this happens given how quickly, you know, agents are proliferating.
[00:20:50] Greg Kihlström: Yeah. Yeah. Definitely. Well, we’ll have to have you back on the show to, [laughs] to talk about the, the aftermath of some of those and, and what to do. Definitely, definitely agree. And, last question for you, what do you do to stay agile in your role, and how do you find a way to do it consistently?
[00:21:06] Kevin Yang: Yeah. So this, comes a little bit from my roots as entrepreneur, but, interestingly enough, we actually have a small team of former entrepreneurs and small company, operators dedicated to tackling risky but high return bets, right? And, they’re, they’re able to kind of do the research in the moment to see how, folks are, are solving these, like, cutting edge problems. And b- by kind of separating this group out from our main product development process, we’re able to kind of see what is possible within the context of our company at a very rapid pace and, that’s something that has worked super well and, allows us to separate the hype, you know, the, the social media hype from what actually works for our business.
