The Coordination Tax: Why AI Is Burning Out Your Best People | Dan O'Connell, Front

Everyone assumes AI in customer operations means fewer people and lower costs. The data from 700 customer operations leaders is more complicated, and more honest. Dan O'Connell, CEO of Front, joins Craig Smith with findings from a survey of customer support, service, and account management leaders that reveal what's actually happening as enterprises deploy AI at scale, and the central finding is a problem nobody has named yet. For every hour of actual customer resolution, organizations spend roughly three hours on internal coordination: the handoffs, context transfers, escalation management, and back-channel communication that happens invisibly between customer interactions. More than 40% of companies don't measure this at all. AI is making it worse before it makes it better, because more agentic automation generates more coordination events, not fewer, and the companies reporting the highest AI technology satisfaction in the survey are also reporting higher coordination burdens. The survey also pushes back on a dominant narrative: that AI in customer operations is primarily about cutting headcount. O'Connell's respondents are overwhelmingly focused on improving net revenue retention and customer expansion. The conversation closes with O'Connell's most forward-looking observation: the coordination tax is a temporary problem that AI should eventually solve, but measuring it is the prerequisite that most enterprises are still skipping. Subscribe to Eye on A. I. for weekly conversations with the people building and deploying the future of AI.

• 0:00Speaker 00

  We get really excited that AI is going to solve all these problems. Well, there's plenty of things AI can't do and that needs to hand off to somebody. Then we have more things to hand off and more context to check.

• 0:09Speaker 01

  You can have a small team of direct employees in your head office or near your head office. That's powered by all of this AI tooling. You can do so much more with a smaller team.

• 0:22Speaker 00

  The vast majority of people are focused on not reducing costs, but how to actually drive improved NRR. It's like this notion in the market of you buy AI and it solves all your problems. And that means that you have people. Actually, it's not about that. It's actually about how do we deliver an exceptional customer experience.

• 0:40Speaker 00

  So I'm Tien O'Connell, CEO of Front. Thrilled to be here, Craig. I know we had a conversation many years ago when I was at DialPad. I joined Front, we're a customer operations platform. What that means is we want to be the single platform where both humans and agents can work together to solve customer problems for businesses. A hundred million dollar business, just north of that, nine thousand customers. We specifically focus on working with businesses that have some notion of complexity around how they serve customers. And really focus on getting multiple teams to leverage our platform that are customer facing. So usually that starts with a support organization, also includes client services, customer operations, account management, customer success, and even sales.

• 1:25Speaker 00

  And so thrilled to be here and get into our conversation.

• 1:29Speaker 01

  What is customer operations? Because I've spoken to a lot of people about customer service.

• 1:35Speaker 00

  Yep.

• 1:36Speaker 01

  Using voice, a hybrid of voice and call center. But it sounds like you're talking about something broader. Can you break down that phrase, customer operations platform?

• 1:48Speaker 00

  Yeah, so we use it intentionally because our platform is really focused on how we get multiple teams that are serving the customer today and how do we solve multiple problems for those teams? So we do compete directly with traditional help desks or ticketing platforms. So your listeners may be aware of, you know, when you send an email into a, into a business or you chat with them on the web, there is a platform that is powering the initiation of that conversation. And then how people are actually solving that problem and all of the analytics. So we compete with the traditional help desks and chat bot businesses of today. And then we also, as I said, when we focus on having multiple teams that are customer facing, we also compete with the account management and client services platforms that are out there.

• 2:33Speaker 00

  So the gainsites of the world, the vitals of the world and the churn zeros. And so again, we really believe in this notion that, look, we want to be where your, your customer conversations initiate, whether they are on the web, whether they're on voice, whether they're on email, whether on SMS. We want to try to automate as much of those conversations as we can. So again, if we can provide an AI-first experience, we will deliver that. And then we recognize that there's high value relationships and even conversations that you don't want to start off by AI tackling. You want to have a human first. And so we're that single platform where both the humans and the agents are delivering that service and solving those problems for those customers.

• 3:18Speaker 01

  Say I'm shoe manufacturer or a big shoe retail operation. I get Front, the platform, and what do I, do I have like different dependencies? Like customer service, receptionist or call answering? And then there's, that passes it to an LLM that answers. And that would be one thing. And then another thing would be customer experience, people that are working with customers. Can you describe what the platform looks like schematically or?

• 3:56Speaker 00

  Yeah, the, the best part is it's very familiar to what you use today in terms of like Gmail and Outlook, right? It feels like, or even Slack. So it feels like a workplace. And so that workplace is, it's an inbox. Conversations across all channels come into that inbox. So you will see messages that appear from customers. Again, those messages may be SMS messages. They may be voice conversations, telephony conversations through integrations that we provide. They might be emails. And then what is happening within those conversations is there are either agentic workflows that have tried to solve those problems and then been passed for an escalation. Or again, perhaps they are passed because they're a high value relationship where you want to have a human actually tackle them first.

• 4:40Speaker 00

  But we're really, think of us as the workspace or the inbox. And then we're helping people both categorize, triage, and then respond to those conversations, assuming those conversations are customer first. And then we would sit alongside any contact center on voice.

• 4:56Speaker 01

  I see, but integrated with?

• 4:59Speaker 00

  Correct.

• 5:00Speaker 01

  Yeah, yeah.

• 5:01Speaker 00

  So we will be like the single workspace where businesses would manage, triage, and respond to those customer conversations.

• 5:09Speaker 01

  Right. And is there an agentic system that's watching that inbox and, and making decisions about who? Because I can imagine that would be pretty onerous for a large organization to have people.

• 5:25Speaker 00

  Yes.

• 5:25Speaker 01

  You know, watching.

• 5:27Speaker 00

  Yes, we have, we have, we take two approaches to it. One is how do we provide? We call it autopilot, which is autopilot is an autonomous agent that you can set up on the web or over email, which will automatically be the initiation point for conversations. So again, if somebody is trying to talk to a business to get support, they're on the web, talking, talking through chat. Autopilot would go and be able to respond to that question. If it can't solve the question itself, it would escalate or route the question to a human for that human to then go be able to engage with the customer. And again, it's all in that single platform. And then within the platform, we believe there's massive opportunities to help partially automate workflows

• 6:09Speaker 00

  or help people just drive productivity and efficiency. And so we have a copilot. And that copilot again is an autonomous teammate that is there to help do tasks for you. Again, those tasks are really focused on how do we, how do we help you, the person that is serving the customer, do things faster and better for them.

• 6:28Speaker 01

  Yeah. And how different is this from Talk IQ, which I think was the DialPad? One of the DialPad products, is that right?

• 6:41Speaker 00

  Yeah, speech rec, yeah. So Talk IQ was a speech recognition company that, that I built.

• 6:47Speaker 01

  Oh, that's great. That's what you sold to DialPad, that's right.

• 6:50Speaker 00

  Yep. And then, and so if you think of speech rec was how do we go and understand the customer conversation? So how do we do real time transcription? And then that was acquired by DialPad. DialPad is a contact center business, phone system as well. Took that technology and said, look, if we're powering phone conversations, whether those are sales or support conversations, we want to have transcription and understanding of what is happening within those conversations.

• 7:15Speaker 01

  Yeah. Is the engineering behind those two systems very different?

• 7:22Speaker 00

  Yeah, I think, you know, what I learned in building speech recognition a decade ago. It's funny when I look back at that journey, we were really building almost four startups in one. We were doing our own speech recognition models. We were building and doing our own machine learning. We had to do our own tooling around AI. So at that time, not all of the platforms and tools that developers can use today existed. And so it's funny that I look back and say it was this one startup focused on speech recognition. There were probably four or five different startups within that startup, simply because that was a moment in time when tooling and LMMs and frontier labs didn't exist.

• 8:04Speaker 00

  Well, they might've existed, they were very early in their journeys today. And then I think, you know, at that time, like we would take open source models and we would fine tune them. We would also learn to go and build our own models. And so, you know, I think it's been really interesting and informative when we fast forward a decade and then see the world that we're in today, where we suddenly see, even in today's market today, people are excited about open source as a way to control costs and perhaps have

• 8:37Speaker 00

  similar types of quality in terms of output. We see people start to think about, should we build our own models? Do those need to be smaller models? Again, because you don't always need the most powerful, large frontier model for basic tasks like summarization. And so, so this is trying to see where I was like 10 years ago to where we are today and how that informs like product and engineering and really R and D decisions around investment.

• 9:00Speaker 01

  Yeah, so you're you're building components uh the the do different things, right? You'll have smaller models to take care of simple tasks. This is all behind the platform. And and then you you'll have frontier models or or like, are you experimenting with Kimi three? Uh, which is the most powerful? Open source or open weight model at the moment.

• 9:28Speaker 00

  At the moment, yeah, I laugh at like the moment in time of like every every there is a there is a a different change on on the on the L L M battleground. Um, yeah, and I think what's interesting on this is we take a couple of different approaches. Um, one, we have like very, we have deterministic rules. Um, and so I think like, uh, the market today gets really excited about how do you push? Um, how do you leverage the biggest best models for almost every task? And you don't always need to, you don't always need the largest model on um, kind of simple, basic tasks. And so we have deterministic rules. Um, we have um, uh, traditional machine learning, we have traditional N L P models that will go and do uh categorization.

• 10:12Speaker 00

  And then we're really thoughtful about like, when do we introduce uh frontier models? Uh, and again, uh, these cutting edge models to go and do sophisticated things. And I think like, that's really the journey that I think most companies are taking. I think vast majority of companies are are run by really smart, competent people, and they recognize that you're not always going to need. It's not one model for all things. Um, and that means that you're going to be focused on, you know, what do we do in house? Um, what do we leverage a smaller model from a frontier lab? When do we go and use the biggest and best model? And so those are all things that our engineering product and design team thinks about openly debates and discusses. And I think there's always like three constraints on everything, which is, how do you control cost? How do you deliver an exceptional experience?

• 10:57Speaker 00

  So how do you think about quality? And then how do you control latency? Um, and it's right, those are like the biggest three problems. And you usually are optimizing for as much as you want to optimize for all three, you usually can get two of the three.

• 11:10Speaker 01

  Yeah, yeah. And and you're using open source to to lower the cost, not only for yourself, but for your customers, right? To be able to provide.

• 11:21Speaker 00

  Do we we have part, we have partnerships today with both Anthropic and ChatGPT and OpenAI. Um, we don't use open source uh today. Those are things that we do experiment and uh uh test. Uh, that might be a journey that we go on. We're really happy with the partnerships that we have. And I think the frontier labs are on the cutting edge of uh continually to impress us in terms of building fantastic models that can do things. But again, as things scale, being a a CEO and a leader, um, the cost comes into it. And so you always want to deliver the thing I'm really focused on for our product is how do we deliver the very best product? How do we, how do our customers have an exceptional experience with our product that they ultimately want to pay for?

• 12:03Speaker 00

  Um, and so I think there are the questions as you scale a business is, how do you think about your gross margins? How do you build a durable, efficient business and scalable business? And so I think those are always conversations that we have. Um, I think there's the argument in the market that, look, pricing comes down from the frontier labs over time because of scale. Um, T V D on whether that um, that is the case long term or not. And so if that, if we don't have conviction that that will play out that way, then it does have us think about, do you go the open source route? And do you fine tune models or do you build your own models themselves? And I think we're in a very interesting moment in time when a lot of those questions are coming at pretty much every business on the planet to think through.

• 12:47Speaker 01

  Yeah, yeah, I mean, this uh, this battle between Chinese open source and U S proprietary models is fascinating to watch. And I, I know that the the the government uh is and and OpenAI and Anthropic are warning about, you know, don't let these Chinese open models uh dominate the market because then you become beholden to Chinese companies at some level, uh, even though they're they're open source. But but it's pretty hard to argue against the the cost differential.

• 13:29Speaker 01

  Uh, uh, you you you argue that uh, early enterprise AI use cases were mostly about cutting costs and speeding up tasks. Is that still the case? Or have we moved on to uh?

• 13:48Speaker 00

  Yeah, I think I think it shifted. And like, I think what's really interesting in these markets is like, things continue to involve. I think when you look two years ago, like people got really excited about, look, we can automate away the vast majority of support. That means that we can cut the support teams. And I think what you suddenly realize is like, look, these L L L M's and a genetic experience can provide these wonderful experiences that really, I think, elevate the types of work that people then need to engage with. And I think it's also clear, like, our belief is the humans don't go away. And again, that's why we've we're really focused on how do we build a platform where both humans and agents help serve the customer? Um, is I don't think it's all about cutting costs.

• 14:27Speaker 00

  I think that is for sure top of mind for some businesses to say, can we build a more durable business or a more efficient business? What can we automate? But I don't think that always necessarily translates into that means that we have to cut people. Um, I do think the conversation shifts into, how do people think about what hiring looks like in the in the future? As opposed to, do I need to cut people that I have today? And so I think that's been an evolution that we see play out in the market. And there's a sentiment even this week where we've seen that like, the hype that AI was going to replace, you know, mass labor and disrupt people is is probably incorrect, that these are tools that are going to drive efficiency. And if we drive efficiency, that likely means that we should have more people and and growth comes from that.

• 15:12Speaker 00

  And I think that's what we see playing out across the board is, there's definitely opportunities where you need fewer people in specific functions. But that means that you get to reorient those people on higher value work, uh, specifically for businesses that continue to grow.

• 15:27Speaker 01

  Are are companies using I AI to to increase retention? Or expand their workforce rather than uh, cutting headcount?

• 15:39Speaker 00

  Yeah, I think a lot of it comes down to, you know, specifically as we play in the the C X market. So we sell to, again, to um, support buyers and account management buyers. Um, I actually ran a pretty, I'm in the middle of running a pretty interesting survey where um, you would see the vast majority of people that responded to this survey um, actually are focused on not reducing costs, but how to actually drive improved N R R, so uh, net revenue retention and expansion. And so it's really interesting that, again, there's like this notion in the market of you buy AI and it solves all your problems, and that means that you cut people. And what you see when you actually talk to like the operators, whether it's the CEOs or the support buyers or even the account management buyers, is they'll tell you, actually, it's not about that.

• 16:23Speaker 00

  It's actually about how do we deliver an exceptional customer experience? And if we deliver that exceptional customer experience, that means that we're going to have stronger retention of that customer and likely opportunities to grow that customer. Um, and so I think it's been really interesting. I have about twenty replies to this survey that I'm doing that's focused on founder CEOs and support leaders. But it's been really interesting to see that dynamic, because I think it's slightly different than the narrative that is out there at times, whether on Bloomberg or LinkedIn.

• 16:49Speaker 01

  People are focusing on customer retention and expansion rather than reducing headcount. But does that mean that they're they're not reducing headcount? That that they're just uh shifting responsibilities? Uh, or or the the human agents are just becoming faster and better with AI tools?

• 17:15Speaker 00

  Yeah, I think it's a, it's a combination of both. Like, there's obviously some businesses that are very cost centric and and think about, you know, can, I'm not, again, I've um, uh, I don't think there's a world where people aren't thinking about costs and um, and thinking about, um, does that mean that you might need fewer people in certain functions? Um, I think what we're generally finding, though, is that people want to automate their remedial tasks and then repurpose people that they have on higher value tasks. Um, and so I think that is the dynamic that we see in playing out for the vast majority of businesses, as opposed to every business saying, like, I can go plug AI and that means that I need far fewer people to go and do all of these pieces.

• 17:57Speaker 00

  Um, and so again, it's a combination of like, cost is top of mind for everybody, especially if they're buying software, they want to see value from it. And then they want to think about building a durable, efficient business. Sometimes that means shifting some costs around. But again, we see it as like repurposing people as opposed to necessarily just blanket replacing them.

• 18:15Speaker 01

  One of the advantages of using a platform like this is you collect tremendous amount of data and you can use that data for different predictions, one of which is uh, customer churn. Uh, what signals in the data do you look at uh, to predict customer churn?

• 18:40Speaker 00

  I'm going back to like why I fell in love with speech recognition a decade ago, which is like, look, like if you're powering customer conversations, but you don't understand them in real time. You can understand pretty much every problem that a business has. You can figure out what problem you're solving. You can figure out whether the marketing message is resonating. You can figure out whether your sales team is or is not talking about specific features. And then you have all of the signal coming from the customer around what they like or what they don't like, right? To better inform your your roadmap. I had this notion of like, look, like, if you don't understand those conversations, then you can not only identify sentiment, positive or negative sentiment, but you can actually start to then think about highlighting potential churn or potential purchase intent.

• 19:21Speaker 00

  And so those are things that we naturally think about here at Front is, like, if we're powering all of these conversations with customers and they're telling us what is right, what is wrong, what is they're frustrated with for businesses that they're engaging with, we should be able to help the businesses package that information and understand it. Again, it comes back to like those roots of of speech recognition of understanding in real time these customer conversations. There's a wealth of data that is in them. And now we have fantastic capabilities thanks to L L M's to help us take unstructured data, make it structured, make predictions on that data to then, again, help a business make better business decisions.

• 19:57Speaker 01

  So for churn, for example, that presumably is one of the main things that you're looking to prevent.

• 20:06Speaker 01

  How, how early in the customer journey can you see that this customer is gonna have to take, is gonna need special attention to prevent them from leaving? And how often is that prediction wrong? I mean, how accurate is that prediction?

• 20:25Speaker 00

  Yeah, I think anytime you build, you know, my experience, like anytime you build prediction models, they're never going to be absolutely perfect, right? And I think anytime we do like, for example, we do inferred customer satisfaction. So we infer whether somebody has had a great experience or a less than great experience through their interaction with, whether that's a human or with an AI agent. And so like what you learn in that is you want to be directionally right as opposed to perfect. And the reason I say that is like, no model is going to be universally accepted by everybody within a company or within the person that built it and the data scientists within the company. But what you need it to be is directionally accurate.

• 21:10Speaker 00

  And I think that's the piece that I really try to help educate people around, which is like, you want it to be directionally right as opposed to perfect. And obviously you can always plug in more and more signals. We're focused on how do we understand the customer conversation? So what is the customer saying in terms of their experience? We haven't yet pulled in product level detail from a customer. Because again, if you're trying to identify potential churn or upsell, you want to be able to look at product usage combined with what is this person saying when they engage with customer support around that product, right? Those two things I think are really, really important for prediction models, specifically around churn.

• 21:53Speaker 00

  Do you understand the usage patterns of that customer? And then what is that customer saying to you about their experience with the product? Or how they engage with support? And then you want to cut across not just one interaction or one person within a company. It is how do you, you know, what is, if you and I are in the same business and we're on the same platform, we're writing in, we would want to know what does Dan say and what does Craig say? And start to have correlation across that.

• 22:19Speaker 01

  How much proactive outreach does the platform trigger or, or push for or advise or whatever? Or does, is there an agentic layer that does proactive outreach?

• 22:38Speaker 00

  Yeah, so we have an agentic layer. We call that co-pilot. And so again, if you're in like your inbox or your workspace, like we have, you know, our co-pilot, our assistant that is there to help get work done. And really what we're focused, and I think like the first iterations of co-pilots that you see across pretty much every piece of software is those co-pilots are a destination that you go to, right? You're in a platform, there's a right-hand bar that user navigates and we engage with our agentic experience to ask us to do things for it. And what we're really focused on right now is how do we bring that agentic layer to be completely woven into the platform and to act like a teammate on your behalf? And so we've got a really interesting release coming up here next month.

• 23:20Speaker 00

  We'll introduce this notion of AI teammates that you can at mention in a conversation and ask them to do things for you. And it's a much more natural way to engage co-pilot as opposed to opening a sidebar, engaging with co-pilot there. And I think, you know, this is a pretty novel concept. I think we'll see this proliferate across pretty much every, every piece of software, which is like, they want to have agents that are interwoven into the software and act like a teammate as opposed to being like a single destination that somebody goes to.

• 23:53Speaker 01

  Yeah, yeah, Slack is, is doing that. Or has just done that, yeah.

• 24:00Speaker 01

  Talk, tell me what you mean by coordination tax and define it for someone who's hearing it for the first time.

• 24:10Speaker 00

  Yeah, so we did a survey across 700 respondents across customer support, customer service, account management leaders. And what we found out was there's a lot of coordination that doesn't get measured in between handoffs for teams or how they communicate to solve a problem. And so again, it's this messy part, we call it this coordination tax. It's this messy part of like coordinating work and sharing the proper context. And so we wanted to measure that and we wanted to do a survey to best understand like, do you companies measure this? Do they even attempt to measure it? Is it something they're even aware of? And again, if you get into AI, we get really excited that AI is going to solve all these problems. Well, there's plenty of things AI can't do and then it needs to hand off to somebody.

• 24:54Speaker 00

  And so if we have more AI agents doing more work and at times having to hand things off, then we have more things to hand off and more context to share. And so we wanted to really do a survey to best understand, you know, how do businesses think about that? What's the impact on burn, a burnout for, for the employees that are engaging with it? And it was pretty, as I said, it was a wealth of information and super interesting to see. There is some of the respondents and things that people share.

• 25:23Speaker 01

  Yeah, and, and you say that there's roughly from the survey, three hours of coordination for every hour of resolution. That was stunning to me is, is, and, and you know, then you begin to wonder, are you adding work to, to these employees who presumably are using AI in their job, but it's not their main job to manage AI? Or maybe that's what's developing.

• 25:55Speaker 00

  But yeah, I think, I think that is the question. We were, you see, we smiled a lot. We were equally shocked and surprised. And again, it's interesting to see just that reaction and insight come from the mark, from market leaders, where I do think it highlights that when you implement AI, you want to be really, really clear on what is the problem that it's solving? And then you want to be able to clearly understand the measurable impact that it is providing. I think what's really nice about customer support is many of those businesses, and what I love about this market are super measured. They have every metric under the sun. They know the cost per dollar, per hour of a ticket that comes in.

• 26:40Speaker 00

  They know how quickly they respond to it. And so I think it is a perfect market for AI to have very clear output and measurable impact. And I think that gets slightly harder where we think about how do we partially automate workflows? I think every business says, hey, I want to drive productivity and efficiency increases. And we do this ourselves internally. We play around with a lot of tools. Right now, we're in the experiment, the experimental phase. So we will happily try different things and we believe that there's a different way to work. At some point though, there will be the reckoning of costs to say, what does this cost me to go and use this tool? And then what is the measurable impact that it's having?

• 27:22Speaker 00

  And if you can't demonstrate a measurable impact, then I think there's the question that comes back to the vendor. So part of what I love about our business and the product we build is like we're in a market that's very measurable. We solve a very clear customer problem. There's a existing budget and then we've got a platform that can clearly go and show this is the ROI and the impact that we're having. And I think that's not always true for, as I said, like every different piece of software that's out there.

• 27:47Speaker 01

  Yeah, this three hours of coordination for one hour of a resolution, is that self-reported or is that from the data?

• 27:57Speaker 00

  That's from the, well, from the, self-reported from survey respondents.

• 28:03Speaker 01

  Yeah.

• 28:03Speaker 00

  And I think what's, what's interesting is you get into the businesses that are tracking, you know, their coordination tax versus not. And I think what you highlight is like people traditionally haven't been focused on it. And I think what AI is clearly highlighting is like these opportunities to truly understand like, what is the impact of these tools? Where is this tax? Where can AI even have a positive impact on that tax and that coordination to make things easier for people?

• 28:31Speaker 01

  The surveys showed the little over 40% don't measure coordination. Is that because it's so hard to measure or because no one's accountable for it? I mean, that, that to me, it just seems like. A massive opportunity to streamline if it's if it's if you're spending three hours to get AI to give you, you know, four hours, maybe you could have done it with a team of people.

• 29:01Speaker 00

  For sure. I think it's a combination of both. I think people don't know what they don't know. And if you don't measure it, then there's no accountability to it. And so I think what was really interesting and great about that survey is it highlighted to a lot of customers that hey, here's a problem that's going on. Within pretty much every business on the, as I said, like pretty much every business on the planet says like, we don't, we're aware of it or we're unaware of it. We don't really then measure it. And again, I think that presents like a really, I think it presents a really great opportunity for the leaders to think about how to drive efficiency within their business. And then highlights a very clear problem that exists. Because I think if you don't, if you don't think about solving it, you're going to throw AI at a problem that AI, what, what the survey says is

• 29:46Speaker 00

  I have to go to quote the great Bob Barker for Price Right, or not Price Right, right? Different game show. Is it's going to say like, look, you're going to throw ages at this problem. They're politely going to create more work. And then you can't figure out how to solve that and better coordinate it.

• 30:03Speaker 01

  Yeah. You mentioned earlier that, that there's burnout among people in the coordination side. And, and companies have lost, you know, good employees to that burnout. Is there a, a common thread in that, in that burnout story?

• 30:26Speaker 00

  Yeah, I think, yeah, I think like common threats is like, um, look, the, the best employees want to have tools that solve clear problems for them and drive efficiency. And I think what you clearly see playing out is like, there's a lot of change management that's also happening with every business. And I know we get excited by every, by many new tools that can show up. But that does create like very real complexity for, for all, not just the best employees, for all employees. They have to learn new things. They have to learn new ways of working. They may have more increased requests come to them because the notion and belief is, hey, AI gives you superpowers. You should be able to handle more.

• 31:09Speaker 00

  And I think those are very real dynamics that are playing out. I say in like the labor market, but those are the very real dynamics that I think every business and every person is, is faced with, which is we all recognize AI presents a massive opportunity. There's always going to be costs of, I've, um, questions around like the impact of it and the cost of it. And then there's the third is there's just the human element, which is how does that impact the expectation that perhaps my employer has of me? And am I able to, to handle that burden? Am I able to learn new things? Can I change the way that I'm going to work? How does that align with the impacts or the expectations of like the things I need to get done? And I think the survey like highlights that that's a very real threat to a business, right?

• 31:53Speaker 00

  Is you may lose not just the best employees, but employees that are doing solid work because the expectations aren't necessarily aligned.

• 32:03Speaker 01

  Yeah, it seems the, this in particular, this coordination tax.

• 32:10Speaker 01

  It seems like that's a, a phenomenon at the early part of, of a technology's development and integration into the economy that should surely get solved. And is AI going to solve that? Are you guys working on, because as I said, that's, that's a huge opportunity. If you can solve that coordination, then it opens the logjam. Are, are you guys working on it? Or is that something that, that AI systems will solve over time, just as they develop?

• 32:47Speaker 00

  Yeah, I think, I think two things, like we're very focused on it. So we want to have clear measurability and observability for any of our AI related features. And so people should clearly know down to like, why did AI make a recommendation or do a certain task? And then again, getting into like the ROI of being able to prove that, hey, here's the value that we delivered on that. I think what you see play in the market is like many, many businesses understand that that is an opportunity and also an expectation of a buyer in the market. That if they're delivering AI, there needs to be this observability governance, a measurability around those features. I think a year ago or two years ago, many of us didn't have those features, right?

• 33:32Speaker 00

  We were quickly trying to solve a problem of put AI in the hands. You like sell the dream to somebody and the impact, which I think is like very real. But you can suddenly see the market evolve around the expectations to say, I see the opportunity. I believe, I believe what it's capable of, but now you have to go and show me, why is it making decisions? What is the cost of it? Right? What is the impact that it is having in terms of the bottom line to my business? And so again, I think that's an evolution of the market. I think it's a very valid and appropriate expectation that customers have.

• 34:06Speaker 01

  Yeah. The survey indicated that, that AI first organizations

• 34:15Speaker 01

  rate their satisfaction, rate the satisfaction with their technology very highly. I think it was four point five or four point six out of five. While reporting more coordination issues, how, how do you reconcile those two? Is it because they see the coordination issues as temporary? Or, or, or they're confident that'll get sorted out?

• 34:43Speaker 00

  Yeah, I, I think what it highlights is, um, you know, we again, get really excited about agentic experiences. And I think what we're finding today is some, sometimes like those agentic experience is can actually solve more things. But also create more work. And so if they're creating more work, that means more handoffs, more contact, more context sharing on those pieces. And so again, I think it's like part of the evolution of the market that we're seeing that is, I, I, I think like pretty interesting. Over time, we have to get to a place where that gets solved, right? You want to have great coordinated work. It needs to be efficient. It shouldn't create more work to solve that, right for the, for the business. And again, I think it's like an interesting way to go and do that survey and hear from customers to say, hey,

• 35:29Speaker 00

  we're leveraging AI, we're putting it in different places. What we're finding is at times it's creating additional handoffs. Partly because it's perhaps doing more work to begin with. And so it's having the impact that people want, but then it's having this underserved impact, right? Or this, this coordination tax that's showing up because of the benefits from it. Which is more handoffs, more things that are playing out. But again, we have to deliver value and, and figuring out how to solve that, not just, not just us at front, like every business that's out there.

• 36:01Speaker 01

  Yeah. And when you talk about coordination, just

• 36:07Speaker 01

  to make it more concrete, what kind of coordination are you talking about? Is it people meeting to define a problem to give to AI? Is it people?

• 36:20Speaker 01

  Yeah. What, what are the coordination tasks?

• 36:23Speaker 00

  Yes. If you, if, if you think of like a, a customer request comes in, like a support ticket comes into an inbox. You know, ideally, AI has handled that and resolved it and, and, and not, you know, a human doesn't need to get involved. Let's just say like the human needs to get involved. And let's say that like that first person doesn't know how to answer the question. What they then need to go and do is they probably need to find somebody else within the business to help answer that. And that person then needs to be, have a platform that they can communicate back and forth to solve the customer problem. They also need to have all of the context and history around the request that's coming from the customer. And it's probably even beneficial to have all of the customer history and understand what's the size of this customer? How often have they reached in? Have they reached in before around this?

• 37:07Speaker 00

  So I think when we think about like tasks and handoffs, those are all specific examples that show up every day for a business. Who works on this problem? Who do I escalate it to? Do they have all of the contacts? Do they have the account history? When's the, is this a customer or a renewals coming up? And so this all proliferates in this coordination tax and these handoffs, all of those types of questions that show up.

• 37:31Speaker 01

  Yeah. Although again, that seems like that's.

• 37:36Speaker 01

  Something that will be solved in short order as you know, as it becomes more glaring. Uh, because a lot of those things you're describing could be taken care of by an agentic system or uh, you know, a system that that communicates to stakeholders with a full reports that they could quickly make a decision. Uh, you say AI works best with all of this customer context. Uh, but most companies don't have that. Uh, what's the sequence?

• 38:18Speaker 01

  Uh, how do you fix the org chart before you buy the software so that you can cut through some of this?

• 38:27Speaker 00

  Yeah, I think I think it starts. Yeah, I think it starts with like having a very clear wedge. Um, and so one is like, we really focus on, as I said, like support conversations or our, our uh, conversations around the customer, right? Whether that is your operations team, your client services that these are requests that come into a team every day. Every business deals with these requests that come in. They need to have a single destination and platform to go and manage those requests and reply to customers uh, on those pieces. Um, and so I think it really like starts there, which is you've got to have like a clear problem.

• 39:01Speaker 01

  So how do you fix that for companies that don't have that?

• 39:05Speaker 00

  Yep. Uh, so I think what's great on our platform, um, you know, is we're a single platform where multiple customer facing teams. Assuming you buy into our vision, right? Which is our, our vision would be like, you want one platform to have your customer, multiple customer facing teams, um, to have all of the necessary contacts when they engage with a customer. Um, so I think like the challenge is always like, if somebody doesn't run their business that way, then they get to deal with all of these coordination problems that show up. You've got four different pieces of software trying to solve a customer problem. You got a chat bot that you're managing. You've got a ticketing system that you're managing. You're probably on Slack. Um, who knows what your account management team is on, your sales team's on something else. And then you're like, I believe that I can go and get these six pieces of, six pieces of software

• 39:51Speaker 00

  to actually work together. And I would roll my eyes and say, that's a dream that I think, um, somebody thinks that they can go get an LLM and some MCP servers and APIs and like tie it all together. And I'd be like, this is not going to work. Um, and so we start with like that shared vision of the world of we think there's a better way to go and share that context around the customer for multiple teams that are likely to engage with a customer. Um, and even, and, and it's not just multiple teams. It's multiple agents in the future, right? If we think about in, in the future, it is likely that an agent is doing some work. And there's an agent that's sitting next to you, helping you, Craig, do work faster and better. And even that agent needs to have all of the context.

• 40:33Speaker 00

  And so again, like our notion is like, that's why we're so focused on being a platform for customer facing teams is how do we give them all of the necessary information? Here's all of the current conversations or past conversations that have happened for that customer. Here's the account history. Here's the renewal data. Here's when there, here's what the product usage even potentially looks like. Um, and so that's our vision of like the platforms, because I think there is such an immense opportunity to help drive better coordination around customer facing problems.

• 41:05Speaker 01

  Yeah. And the survey, I mean, you're, you're trying to solve all these problems. Or, or was there anything in the survey that uh, that you, you hadn't thought of that you're, you realize that's a problem we need to solve? I mean, was coordination one of those or were you aware of that? And then how do you go about tackling that?

• 41:27Speaker 00

  Yeah, I think that, that, you know, um, the highlight that you shared was most surprising to us. And it's usually like where we start the conversation on the coordination tax, which is this three hours, right? That people solve for everyone ever solve. Like, to me, that is the biggest takeaway from the survey. Um, and I think it highlights, um, how much of a problem there actually is sharing context and driving efficiency within a business. And I think again, that problem turns into an opportunity. Um, and so again, like not going back to us, but I think that's really why we think of my belief is there's fewer systems that people, um, coalesce around in the future to do things.

• 42:08Speaker 00

  I think there is very much this very real opportunity to have a customer facing platform. And that is the single platform that unifies teams and agents to solve and serve the customer. Um, and so I believe like that's how this world will play out. Um, and so I don't think that people are going to continue to buy like point solutions and hope that they can stitch everything together. I really much believe that like this universe will slowly start to consolidate and collapse into fewer pieces of software. Um, and again, perhaps those pieces of software become more agentic or more headless over time as well.

• 42:41Speaker 01

  Yeah. And, and is this, did that inform your roadmap, the, the survey or those coordination issue?

• 42:49Speaker 00

  Yeah, I think it highlights that look like you've got to have per, uh, anytime you release something, you've got to have clear measurability impact. You've got to make it easier for people to go and have that shared context. And so again, we started this business as a shared inbox business where um, you could have Slack like conversations on top of email. Uh, you never need to forward an email ever again. And so I think it highlights that there is an immense opportunity to both bring and unify customer facing teams together to collaborate around customer problems. Um, and again, providing all of the shared context, uh, and information for them to go and do that as efficiently as possible.

• 43:29Speaker 01

  Yeah. Uh, looking five years out, uh, does the support organization as a distinct department within a corporation still exist? Or has it, has it sort of diffused into the sales organization or, or engineering department or something? And everyone's sort of helping the AI agents, uh, manage the, the customer.

• 44:01Speaker 00

  Yeah, I think it, I think it continues to shift and blur, the lines blur. And I think we see that shift on engineering and product and design today as well. So as we see those roles shift and those responsibilities shift to, I would say like more of single teams, doing it smaller, single teams, um, or single roles, kind of doing multiple traditional functions within it. I think the same is true for support. Um, and again, I think that plays out in a couple of different things that I see in the market is people are consolidating to fewer platforms to align their teams. And then I think as that happens, it then brings the question of how do roles start to blur or evolve? Um, and again, you heard me talk about, hey, if AI can go and do all of the remedial stuff, what does that mean for your support people?

• 44:46Speaker 00

  There is a business that may say, I don't need the support people. There's another business that would say those support people have a wealth of customer understanding and industry knowledge. And they would be better for us to retain and evolve and put them in a different client facing role, right? Perhaps it's engaging on upsells, perhaps it's engaging on at-risk accounts, whatever it might be. Um, but I do think that there are perhaps fewer teams that are going to be, I think there are going to be fewer specific functions. And I think those functions evolve into like, we have to learn how to then manage and build our own agents to solve problems. And those agents need to work across functions. They need to work across teams. They need to work across people.

• 45:27Speaker 01

  Yeah. You know, somebody said something fascinating to me a, a month or two ago, we were talking about

• 45:38Speaker 01

  a, a piece of software or a company whose software helps design clinical trials for new drugs. And there's a lot of documentation that has to be prepared for the FDA. Uh, submit this, the, the trials for approval and uh, they were saying that, you know, you, you, you get this stack of documents and you email it over to the FDA. Uh, this, these documents that were AI generated, you email it over to the FDA and, and no one reads it. They have an AI that, that then sure, you know, parses it and, and you know, whatever.

• 46:22Speaker 01

  And they were saying, well, the, the day will come where the, you, you don't need the, the document anymore. It can just be, uh, AIs talking to AIs. And in the same way that I was talking, I think yesterday to a, a company called G2, which is a review site for, or software. And they were saying how software purchasing is increasingly automated. And there, you know, you can go into an LLM that's attached through your MCP, I'm sorry. Uh, to, uh, to one of the frontier models that's connected to, uh, G2's database.

• 47:05Speaker 01

  So the, the, you talk to the LLM about recommendations, it has access to the reviews. And at some point that purchase is going to be pretty much automated. Maybe it'll be teed up for a, an approval. In the same way, do you think it'll come to the day where uh, large enterprises will have agentic systems that are maybe running in trouble and they'll talk directly to, uh, front about, uh, you know, the problems they're having, not through a, a human agent, but just directly to your software, your software sorts out the problem.

• 47:49Speaker 01

  I mean, that there'll be this fabric below, uh, the humans and humans, as you say, are working out, upselling or more.

• 47:59Speaker 00

  Yeah.

• 47:59Speaker 01

  Uh, human touch or, uh, where the touches is necessary, but the AI will be talking to AI for a lot of this.

• 48:11Speaker 00

  Uh, I wholeheartedly believe that in our lifetime, I think the timeline, the time horizon is like the, the debate of that. I, I wholeheartedly believe that we will have agents that are reaching out to other agents to solve problems and then even take it a step further, go and deploy good to solve the problem. Uh, on those pieces. Um, I think the question is like, what's the time horizon for that? And again, I don't think that is the next couple of years. Um, that's my own personal opinion. I think all things are a little bit harder than, than, um, than how they, they turn out. Um, but I for sure believe that. Um, I also don't believe that like there is, um, Look, I personally don't believe that like every company just needs to be a one person

• 48:54Speaker 00

  ran business and it's run by a bunch of agents, or it's just a fully agent company. Every company in the world is an agent company, meaning it's just completely autonomous. I think what makes business really fascinating and fun is you get to do, you get to build things with people. I think that's what makes the human experience really unique and fun and hard. And I think that's the same for when we buy software is I think a lot of times being a past seller is it's not always buying.

• 49:25Speaker 00

  Yes, you want to buy a piece of software that solves your problem clearly. You also tend to be buying from the person that you're engaging with. And the journey that that person has brought you on, and I think that is a very unique, difficult to replicate human experience. And if you're writing checks for $50,000 pieces of software or a million dollars for pieces of software, whatever it is that you're selling, I think you want to know that you're talking and engaging with a human on those pieces.

• 49:51Speaker 01

  Yeah.

• 49:51Speaker 00

  And so every time I hear, you know, the, hey, can we get to the agent companies just building agents and talking to agents? I'm like, what do we then do? And I think that's a big part of the human experience.

• 50:04Speaker 01

  Front is, it's a, you know, obviously a common word. Is, is how do people find you? If you Google front, you're likely to come up with a lot of stuff that is not your front. So, so how, what's the easiest way for people to follow you?

• 50:25Speaker 00

  Yeah, you can find us on a, well, obviously you can go www upfront.com. If you put in front, we do buy our brand keywords, so you'll likely see an ad that is there for us on those pieces. You can find us on LinkedIn. We've got a fantastic YouTube channel. It shows off. We do some thought leadership and show the demos. You can find us at front HQ on YouTube and on any of the other normal socials. And so I'm quick to respond to requests, so I'm always happy if people want to have a direct conversation with me. Feel free to reach.

• 50:56Speaker 01

  Okay, great. And it'll be you, not an AI, right?

• 51:00Speaker 00

  I promise it will be me. Spelling, spelling's in bad grammar included.

• 51:05Speaker 01

  Yeah, okay.