Bringing Your Own AI Agents Into Customer Service With Front

[00:00:00] 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. Simply visit denodo.com to learn more.

[00:00:28] What happens in your organisation when somebody wants to use their own AI agents inside the customer service platform where people are already working? Well, in today's episode of Tech Talks Daily, I'm going to be speaking with Dan O'Connell. He's the CEO of a company called Front.

[00:00:52] And he's going to be joining me on the podcast today to talk about bring your own AI agent. A big announcement there. And we're also going to explore together why flexibility is becoming a very serious buying consideration. Because Dan will explain why businesses are now combining native agents, third party systems and tools that they build themselves.

[00:01:17] As well as the context that each agent needs before it can act. So I want to discuss today where people should remain involved. How managers can turn autonomy up or down. And why audit logs and customer outcomes, how all these things matter equally as much as the resolution rates. And Dan will also share how Front is approaching coordination across agents and employees.

[00:01:44] So customers don't have to repeat their history whenever work changes hands. So much of today's conversation will resonate with you wherever you're listening in the world. But enough spoilers and scene setting for me. Let me introduce you to Dan right now. Well, thank you for joining me on the show today, Dan.

[00:02:06] For everyone listening, can you tell them a little about who you are, your role at Front, and also the customer problems you seem to spend most of your time thinking about? Sure. Most of my time. All of my time. So thrilled to be here. So thank you. I'm Dan. I'm the CEO of Front. I've been in role for almost three years now. Time has gone by really fast. And for those that may not know Front, we're a customer operations platform. And so what that means is we unify your support and customer success teams. We solve customer problems.

[00:02:36] So that might be how do we go automate away a first digital touchpoint where somebody has an issue to handling all of the escalations. So we do that all seamlessly through one beautiful platform and help businesses ultimately deliver just really exceptional experiences for their customers. And I have to confess, one of the reasons I was excited to get you on today was after you maybe unintentionally set off my tech spidey sensors when I read that Front is preparing to launch bring your own agent.

[00:03:05] So tell me more about that and what business problem ultimately prompted the decision to let companies bring all these different agents into the same customer service environment. I've got a feeling there's got to be a story there, right? Yeah, I think it's obviously a really interesting moment in time. You know, I say a lot of cliched things for us. But I think like the CX market, when we talk about support and servicing customers, probably one of the – look, it's a massive market, obviously one that's ripe for change and ripe for AI transformation.

[00:03:34] And so what we have clearly seen a trend over the past really nine months is people are not just picking native agents that are being built by platforms, but they're building their own agents. And so I think there's a couple of trends that are playing out, not just in the CX market, but many markets, which is people are consolidating to single platforms to go and unify teams. We happen to be unifying teams around customer problems.

[00:03:59] And then what's very clear is people want to have flexibility to building their own agents to do particular tasks. And they're going to build those agents, whether through open source models or Anthropic or OpenAI, pick a different platform. And they want to be able to power them and bring them to a platform that they have selected. And so we've been really amazed. We did an announcement a few weeks ago. So we had over a thousand signups just in that first announcement.

[00:04:24] And so you can actually see there's a lot of gravitational pull and a lot of interest from customers to say, we love this platform. It's where we start our customer journey. And then we also have all of these capabilities that we're building and we want to bring them to this platform. And so it's been really enlightening and really exciting to see what people are building. And I think it just demonstrates that there's an immense opportunity and having closed walls really wins, I think, in markets. And so we're super excited about that.

[00:04:49] Yeah, well, I must admit, one of the reasons I was excited having read this was most enterprises now, it's not like the old days where everybody's in a locked in Microsoft ecosystem or something like that. Many companies now use several different models and agents all for different tasks at hand. And that will change per department as well, I would expect. So how should leaders, though, how should they decide which agent belongs in which part of the customer journey when there are so many different models at play here?

[00:05:17] Yeah, I think it ultimately, and to give the business school answer, like it all depends, unfortunately. And I think like what's very clear is, look, like we have our, you heard me talk about it at the beginning, like we have our own native agent. We have two of them. One is autopilot, which can go fully resolve tasks end to end. And then we have copilot, which is there to be assist the support and service teams around escalations and handoffs and help them, you know, drive efficiency and productivity.

[00:05:42] But what's clear is like customers are not going to go all in on just a native agent from a platform, at least at least in our experience in the markets that we plan. And so people are building these different agents. Some of that is around cost control. Some of it is around flexibility. Some of it is around like, hey, they want to when we talk about flexibility, hey, maybe it's a set of tasks that they want to go and choose a less complex model, a cheaper model, a faster model, whatever it might be for that set of tasks.

[00:06:09] But what I think is really clear is people want to have flexibility in today's market when they're picking platforms. And you alluded to this. We think that's a winning strategy. We think there's like very clear market pull around that. It's what our customers are asking for. And so as opposed to saying, look, you can only use our agent. That's the only way to get work done. We want to go and say, look, we think we can deliver a great native agent experience. We would love for people to choose the native agent to go and do, you know, as many tasks as possible.

[00:06:38] But we also recognize that there's going to be things our native agent doesn't do. And you should have that flexibility to plug in another agent from a different platform based on a different model and have it help you ultimately get work done across those humans and agents. And 2026 has been the year of agents and agentic AI. I think we've all been on a journey.

[00:06:59] And if we have one eye on 2027, I think we've all learned that an agent can only make a useful decision if it understands the customer, the conversation and the wider business context. And the emphasis on that word context there is a word I've heard just about every tech conference this year. But what information must be available before an agent should be allowed to act? Yeah, I think like the biggest piece is it's got to have the full customer history and relationship. Yeah.

[00:07:28] And so I think number one, like agents can do really compelling things. That all comes back to your point of like they've got to have the right context. So that is, do they understand the customer? Do they understand the relationship? The next piece is do they understand everything that has happened so far? And so, again, I think the biggest frustration for customers when you're talking about customer support ultimately comes into like the handoffs or they have to constantly re-explain themselves and start over.

[00:07:54] And so, again, when you play that into like, well, how does the agent experience play out? Then you want to make sure that any agent, when they get involved, understand who this customer is, right? Understands what has happened so far. And then also it starts to understand what needs to happen next and has an understanding of like, hey, is that something that the agent has the skills and capabilities to go and handle? Or is it something that they're not capable of and they need to then go hand it off to another person within the chain or another agent within the chain?

[00:08:22] And you've got to provide all of the context and sharing around that handoff as well. And I suspect that every customer or client or person that speaks to you at a tech conference will ask you this question. But where should their company draw the boundary between work an agent can complete independently and the moments that require human judgment? I suspect you've said so often it's almost a podcast episode on its own. But what's the happy medium here?

[00:08:49] Yeah, I keep giving the depends answers non-intentionally. Hopefully there's some opinions that are there. But I think it's really interesting. Look, like we a little bit more color around front and maybe perhaps to provide some context there. We support 9,000 customers globally. Many of those customers are large customers, the likes of Navon and Uber Freight and Equipment Share. And then we also work with some really small businesses as well. And so you see like very different dynamics play out here.

[00:09:16] And you also see dynamics play out across different categories and verticals. The reason I say it depends is there's very much businesses that want to have an agent first touchpoint. A lot of those businesses can be e-commerce businesses. Why? A lot of those needs and questions that come up tend to be relatively, I say, simple. Meaning, you know, a shipping request, a tracking request. Maybe it's a refund. Maybe there's an issue with a product that they need to get a return or a shipment done.

[00:09:43] And so a lot of times those businesses will say, look, I'm going to have an agent first touchpoint. The customer's coming in. They want an immediate result. We know it's something that's fairly repeatable in terms of the problem. Like we can go and solve that. And then I think you have other businesses. I can talk about online pharmaceutical businesses, people that are providing, you know, hair supplements and things to that nature. They, while they can provide an agent first experience, recognize that, look, there is some empathy around what they are dealing with. And there's some human judgment.

[00:10:13] Perhaps there's some sensitive nature around the medicine that they're prescribing. And so they actually, while they could go and say, hey, we can go automate all of this away. We actually want to have the human that's introduced there. And so I think one, it stems from what is the business and what is their take of the customer experience that they want to provide. I personally believe like agents can do some really, really, AI agents can go do some really, really amazing things. There's also times when you want to have the human involved.

[00:10:40] And then separately, I think, look, there are very real moments where a human is always going to be going to be needed. That might be require human judgment. It might require just the natural empathy. Something's gone really wrong. It's a really sensitive moment. And so I think a lot of this is this transformation in this moment that we're seeing where I think almost every business is rethinking, you know, what's the first touch point that we want to have? Where do we want to support customers? Right.

[00:11:07] Because I think you can deliver exceptional experiences across every channel. And then where do the humans fit in? Right. Do I want my humans to be the first touch point or do I want my humans to be the orchestration of these agents and ultimately be escalation in the handoffs? And the good news is, of course, we've come a long way from those frustrating chatbots a decade ago and the poorly implemented chatbots more recently as well. But there will be a slight nervousness in some people listening.

[00:11:36] So what visibility do managers need to understand whether an agent is actually helping customers, creating extra work or quietly making poor decisions or even upsetting people? How do they get that visibility? Yeah, I think what we've really focused on in our platform is making sure that any time an agent is involved, there's full of observability. So all of the audit logs around, you know, why did an agent take the action it did? Why did it make the recommendation that it did?

[00:12:04] But I think more importantly is ultimately being able to measure the customer impact. So we do things like inferring customer satisfaction, that inferring customer satisfaction works across both human agents and AI agents. And I think those are the metrics that are most important to a business is to say not to have fear of AI, but to say, look, regardless of whether it's a person or an agent, what is the experience that a customer ultimately has? And this ultimately gets into cost.

[00:12:32] We're going to have a different conversation around the cost of these things. But I think what's really important is the businesses are saying it's not just about a resolution rate. You can resolve a lot of things. You can go and deflect a lot of things. It is really about understanding, did that experience solve the customer problem? And did it do it in a way that ultimately created a delightful experience? And if you're not doing that, then I'm like, you're probably focused on the wrong metric. Yeah, completely agree.

[00:12:59] And as I said at the very beginning, one of the things I love about what you're doing here is the open approach. It certainly offers a lot more choice. And that is what users are looking for right now. But there will be some people listening, maybe IT people are a little bit more cautious that will say it can also create a fragmented collection of models, tools and vendors. So for the more cautious listening, how can their company gain flexibility without losing control? Yeah, I think it's about providing that control.

[00:13:27] You know, I do think there's like philosophical differences of, you know, there's definitely going to be a business that says like, I want the walled garden. I want one platform. I want to have one agent that does these things. And I think like what you'll find, what I would argue in that is like, look, there's going to be things that that single agent can't do. I personally don't believe in a world where there's one agent that does everything. And I think there's like a multitude of things we can point to as to like why that will play it, why multiple agents will show up.

[00:13:54] But I think like when you're dealing with that, ultimately what that person wants is they want to have governance and control and oversight. They want to know that they can trust the agent and that they can dial back the agent or dial up the agent when they need to. And so we think about automation in terms of a dial. You can always introduce, you can believe in automation, but also always have a human end off. Or you can always require a human to do the final sign off and to click the button to do send.

[00:14:21] And so I think it's about providing that type of flexibility and that type of governance to the people that might be a little bit more cautious is to say, you don't need to go and implement AI and suddenly say it's going to go automate full tasks end to end. And you should just trust it is we want to say, look, you can step into that belief. You can always require the human intervention. And then again, we'll give you the measurability, the controls and the oversight that you need to figure out like how and when you want to turn the dial, which I think ultimately is what that person is trying to solve for.

[00:14:50] And as an ex-IT guy who still gets flashbacks from things going wrong on a Friday, I don't know why things go wrong on a Friday, but they always seem to. But what should happen when an agent possibly makes the wrong decision? Who owns the outcome and what should teams learn from a failure like that? And it'd be great to tick that box because it will happen at some point, I suspect. For sure. And I think like at the end of the day, this comes back to like the people have to own the outcomes.

[00:15:17] We talk about this a lot of, you know, even rolling out AI internally up front for our own needs is just because you're using AI doesn't mean that you get to say like, I didn't own the outcome or I didn't own the results. So I think like number one, I think there's a values and principles approach to that, which is like, hey, if you believe in automation, you also need to own the outcome of what that is. Two is you have to have like clear understanding of the outcomes. And so, again, that goes back to like the audit logs of, hey, what happened? Why did the agent take the action that it did?

[00:15:46] And then back to our to the point we were just making, which is like, do you have insight as to what was the customer result in experience that they had? Because that ultimately tells you, did this work? Did it not? The models and the experiences that we provide, we obviously build our experiences and focus on accuracy. So, you know, you can go through scenarios and testing and go and say, hey, if I get this type of question, what would the output be? Does that match the accuracy that you're looking at? So that's always baked into the platform.

[00:16:16] I don't mean to like skip over those pieces. But I think ultimately person wants to have all of the government testing and then the end result in a single place. And there has been a lot of noise over the last 18 months about return on investment from expensive AI projects, mostly after that famous MIT study, of course.

[00:16:35] But are you able to share a practical example that would just make people feeling maybe a little bit easier about how combining the right agent with the right human expertise can actually produce a better customer or business result? Because there's the old mantra in IT, if you can only improve what you measure, it'd be great to hear some real measurable results that you've helped with here. It is. And I think like what's clear for us, too, is like our own internal AI spend goes up nearly every month for like the different tools that we use.

[00:17:04] And I always have that same question of like, well, how do you prove ROI? How do you understand the value that they're delivering? The nice part about playing and the CX space and the customer support space is every metric is measured. I think if you go and talk to support leaders, they will tell you, here's what our cost to serve is, right? Here's what our resolution rate is. Here's what our CX space rate. They will know every metric by hand.

[00:17:25] And I think what's really important and what we share with businesses that think about how to implement AI agents is thinking about like, what is the cost to the outcome? So it's not necessarily the cost to serve, but it's really you should obviously care about your total cost to serve. But at the end of the day, like you care about providing the best experience in an economic way that fits within the realm of your business.

[00:17:48] And so we spend a lot of time making sure that people can see the ROI and the impact that understand the expense that they're making. Different businesses have different philosophies. You know, some businesses are looking at AI as a way to reduce either existing headcount or future headcount. Other businesses are thinking about it in terms of ways of like, hey, how do we up level the people that we have to go work on like more strategic relationships or have a broader impact within the business?

[00:18:16] But again, like what's very clear, I think, is people want to make sure that they can say, hey, if I implement this, I know what my spend is and I know the ROI that I'm getting. And that's what I love about this space is like we can very, very clearly go and point to say, here's the impact that it's having on your customers from a satisfaction. Here's what the cost is, right? We know what the cost is for those people, assuming like we have an open relationship with the business. And we can help talk about like, hey, well, what's the right framing and the right experience that you want to deliver?

[00:18:46] I love it. And I'm curious, we're what, two, three months away from 2027. It is literally around the corner now. What excites you about the workplace in 2027 and what will change and how quickly things will improve there? What makes you want to jump out of bed in the morning? Anything that's really got your attention? Yeah, I do.

[00:19:05] You know, like the nice part about being the CEO of a business and I think, you know, a really interesting space is I get to make changes and help steer the direction of the company to things that excite you. You know, obviously, you got to make sure they like make sense for customers. I really am excited about our open platform strategy. And I know, you know, this is not just me here talking about the greatness at front. But I really think and I'm and I think this is like a moment where people are going to pick fewer platforms. I think we see that.

[00:19:33] I think we see a lot of different teams starting to blur lines. We see that in the CX space. We see support and customer success account management blurring. You see that in EPD where design and product management and engineering are very much blurring. And so I'm super excited that I think there's fewer platforms that win. I think we consolidate to those platforms to do different pieces of work. We're focused on the pieces of work that are around customer that are customer facing. And then two is I think more agents are going to show up and that shouldn't be a surprise.

[00:20:03] I don't think there's the proliferation that sometimes we read if like the company of 100 people is going to have 3000 agents. I think you will start to consolidate the number of agents that we have. But I do think the openness that people are going to say, look, sometimes I'm going to pick a first party agent. Sometimes I'm going to pick an agent from a different vendor. There's going to be another moment where I build my own agent because you want additional flexibility.

[00:20:27] And you're going to say, I want to bring all of those things to the platform where my humans and my agents are working to get whatever work it is done. And so I think that's really going to play out over the next year, if not the next six months, the way the world moves right now. It's really, really fast. And then I think the other piece that plays out is people are looking for exceptional customer experiences truly across every channel.

[00:20:48] And I think there's naturally been some friction that has showed up over the past decade of how do you provide a seamless customer experience that is consistent across every channel? And I think agents allow us to go and do that, to be honest and frank. And so I'm super excited about building a fantastic experience across every channel for customers. Yeah, it is an incredibly exciting environment at the moment. And different teams use different tools and AI models. And each model or tool has different strengths and weaknesses.

[00:21:18] I think it's just important that we have one toolbox and can help ourselves to whichever one that we need for a particular task. So for anybody listening that wants to learn more about how the front platform is enabling AI and people to work together in one place and ensuring all that customer work moves forward without losing context, where should they go? Yeah, we would love to chat with anybody that's got, as I said, support needs. That's thinking about, you know, a different vision of the future. We think we're working on some really interesting things.

[00:21:48] You can find us at www.front.com. We publish some first-party research, which I think is really interesting as well. We recently rosed out some research around the coordination tax, which is people spend a lot of time coordinating and what's the impact on that business. And especially in an agentic world, you can find that at www.research.front.com. And then I'm on LinkedIn. I'm pretty active at posting and replying to messages. So if anybody ever wants to reach out, you can find me there. I'm happy to say hello.

[00:22:15] Well, I genuinely really believe that the organizations that get the greatest value from AI will be the ones that can build a flexible foundation, can support innovation, adapt to change and improve customer experiences over time. So I absolutely love what you're doing here. I'll include links to everything you mentioned, including the research. But just thank you for bringing all this to life and sharing your story today. Appreciate your time. Yeah, I appreciate it. Thanks for having me.

[00:22:41] I think Dan's argument is a useful reminder that choosing an AI agent is only part of the customer service decision. The harder work actually involves carrying the right context through every single handoff. deciding when a person should review an action and measuring whether the customer's problem was genuinely solved.

[00:23:04] Because an open platform gives teams options, but these options still need governance, testing, audit trails and clear ownership of the result. So a big thank you to Dan for explaining how Front is building around native, third party and self-built agents without forcing every company into the same operating model. Incredibly refreshing to hear. So remember you can find Front over at Front.com.

[00:23:33] Research on the coordination tax at Research.Front.com. But over to you as agents multiply across your business. Who is making sure they share context and remain accountable for the customer experience? As always, techtalksnetwork.com. Browse to episodes and there will be a blog post associated with this interview today with all the links we talked about and much more. So I invite you to come check that out.

[00:24:03] And while you're on techtalksnetwork.com, browse to events. I'm traveling a lot of miles between now and the end of the year. I'll be going to Austria, Vegas three times, Silicon Valley, Orlando, San Francisco. If you're near any of these areas and you're attending one of these events, let me know. Look to meet you in person. But that's it for today. So thank you for listening as always. Speak with you soon. Bye for now.