More Customers Chose the AI Agent Than Anyone Expected | Tom Chen, Aircall

0:00I've had the experience of stumbling onto an AI voice agent. You can hardly tell at this point that it's not a human. Typically, a human agent can only take one call at a time. But with an AI voice agent on Aircall, they can take 100 concurrent ones at a time. I'm still getting phone trees or chatbots that have canned answers. And it's just frustrating. It shouldn't be like that. How do you see the market? Is an AI voice agent going to be as performant as your smartest agent who's been around, who kind of has all the knowledge of the things that are never documented? Probably not yet, but I do think that it is better than the average rep in the call center.

0:44Can you start by introducing yourself to listeners, give a little bit of your background so far as it's relevant and how you got to Aircall and then what Aircall is, and we'll start talking about it more deeply. So I, Tom, I work at Aircall as the chief product officer. When I first joined Aircall about two years ago, it was a little over two years ago, almost getting two and a half, we were in a completely different world. I mean, this was about a year after I think ChatGPT had launched. There was some early promise, but I don't think anybody sitting in my shoes or most people's shoes would have guessed how quickly it might have moved.

1:25Of course, there were folks who are maybe more prescient or more future looking, better future predictors than some of us. But I think it probably caught most of us by surprise in terms of how fast it's moved, especially in certain general AI spaces like coding and others. And so since my time joining Aircall to now, we've really turned from what you would describe as a software SaaS business to much more of an AI business. because almost all customers around the world no longer look at a phone system or a customer communication software as just the software itself, but more around, what can you help me actually resolve?

2:12Whether it's tickets for conversations or can you actually autonomously handle conversations? And while deployment and uptick is not across the entire world uniformly at 100%, We're still in the single digit early innings. You can bet that all businesses are thinking about it and trying to find a future partner who have the capabilities that they can really lean on and trust moving forward. And trust is an interesting thing because it runs across the gamut of not just the AI services, but also your base reliability in, you know, whether it's your chat or telephony services, your ability to navigate the different telco regulations around the world.

2:56All of those things matter. And I think over the course of the next few years, we are going to see some more definitive winners, especially for certain cohorts of, let's say, ideal target customer profiles. You might have a couple of winners in enterprise and a few down market, some verticalized. verticalize. But anyhow, so that's a little bit about maybe not just about me, but what's changed in the market and what Aircall is about. I'll add one more thing a bit about me. Prior to my time at Aircall, I worked at various different Silicon Valley tech growth startups or growth stage companies, always in product here.

3:43Most of my career has been in B2B. So pretty familiar, I think, with how kind of businesses think about things. We specialize at Aircall a little bit more on the smaller business side. So there are consumer elements in how we do business. And that's fun, right? You're maybe sometimes not as bogged down by the slow moving natures of some enterprise. But I'll stop there. I'm sure there's a lot more things to talk about. I'll stop. Yeah. Yeah. Yeah. Well, we were speaking before we started recording. You were saying that one of the advantages you guys have is that you started overseas and are now in the U.S.

4:29market, so you have a multinational footprint, which is helpful for multinationals. Does that include multilingual services? Absolutely. I think there's a couple of points to it. I think with AI nowadays, you can easily get access to multilingual services, especially if you look at transcriptions or even voice agents. you can look at any of the providers and they'll be able to offer 100 plus languages and you're good to go but where Eric Hall really excels is that we have local go-to-market sales teams and support teams who are of that nationality so a good example of this is like we have teams on the ground in outside the United States we have it in London Paris Madrid Sydney Berlin Mexico city.

5:32And I think we're one of the few who serves like our, you know, a little bit more down market who are boots on the ground. And the difference between, and I think everyone who's international can probably attest to this difference between just being able to speak the language, but also kind of knowing the local culture and way of doing business. There's a huge difference between those two things, right? It's abundantly accessible to just be able to have some services that can transcribe into those languages. I think it's more rare to have, you know, a customer success team or a sales team and support agents who really know the local business culture.

6:14And that I think stands out to a lot of our customer base. In the world of AI, when there's so much just AI spam, you know, at the end of the day, I think what becomes a little more scarce in the world and humans are always gravitating towards scarcity to a large extent is just a team that can really help them all right and so we we we pride ourselves we pride ourselves on that especially for those companies that are more multinational being able to really talk the talk like talk their talk knowing how businesses really work and then of course our products are also then built with this in mind uh just across the board like we we really care about the different cultural aspects from all the different local markets.

6:58And we think we're very thoughtful about how we do this. So that's maybe what makes us a little bit more unique. And that does come from perhaps a more European heritage, where on day one, you're forced to somewhat tackle all the markets. Whereas in the United States, you kind of get the, you know, one large market, great market, but you're not forced to do that on day one. You know, Aircall was forced to do that on day one. And since then, we've been scaling in that manner. And United States just happens to be one really large market. And some of our best tech talent is obviously out of the United States.

7:33But we have large presence here as well, from San Francisco, New York, Seattle, et cetera. Yeah. And it's funny, my family and we talk about it just now that they're living in those states, how that people don't appreciate that cultural sensitivity and it's one son in particular is is in a star japanese uh startup and he he talks about their colleagues that you know he's the favored guy because uh he's he's comfortable and he can read the room and he understands those nuances when we spent a lot of time in Asia. Yeah, it's an underappreciated skill or aptitude, I should say. Not that people that have never left the States, some may have it, but not many have it.

8:32And you guys are really focused on voice as opposed to text. I mean, when General DeVeyer first came out, there was this huge shift chatbots that would then hand off to human agents at a certain point. I continue to be frustrated. I've mentioned this many times on this podcast. the pace of adoption, I'm still getting phone trees or chatbots that have canned answers. And, you know, it's just frustrating. It shouldn't be like that. How do you see the market? And as you said, you're at the very beginning of this massive market. How do you see that adoption continuing or hopefully accelerating? Yeah, I think it's very much accelerating.

9:36So that's the good news. What I would say is that right now you get this tension between perhaps what folks actually think customers might want versus what feels safer for their jobs, let's say. and like all adoption or adoption curves of kind of a frontier technology I think you're going to be faced with that what makes this space what makes me feel more bullish about this space in particular is that you a lot of companies now understand that you don't have to do everything in one step that is to say look let's say you have a phone tree an IVR that all your operations are like based off of that and you don't want to muck with that, right?

10:24Like you got people who are employed, who have specific shifts, and especially in large contact centers. In some of the largest contact centers, you can get up to like 10 ,000 plus even 100 ,000 reps. Not all at the same time, but with all the different shifts that are happening across the year, right? There's numbers that are really large that you wouldn't even fathom. And to even get there, you must have had a really well-oiled machine. So companies don't really want to mess with these things. It's a headache. It's a potential. There's downsides of the service. But what I think companies are discovering is that they can first start to get a taste of how well generally AI is behaving or how well it can be deployed in things like after hour use cases, overflow use cases.

11:16And these are all a little bit less scary than trying to deploy, you know, from eight to five. And then what I see is the moment customers have taken that first step for a use case in which is a little bit more upside only, meaning like they weren't even picking up these calls anyways. So even if they just answer a few questions and maybe the customer left through a structured few questions on a voice agent, what they're really looking for, it's more upside than there's almost very minimal downside. Once when they get experience of that, they start seeing so much more opportunity to then basically get higher and higher coverage.

11:59And so part of what we try to do at Aircall is try to educate. I think you have to really educate your customers. And if there's any customers listening, it's like you want to find a partner who can help you along on that journey. Because it's companies, customer service or even sales, it's so personal for most of these companies. It's kind of the lifeblood that it's not the same as just procuring any random piece of software. And so what you often typically see now in the enterprise market is most of these voice agent or AI agent companies are going to be coming with a full, like a forward deploy engineer.

12:41They're going to build the solutions with you on the side. And I do think in the early stages, that is what we all have to go through to kind of get through the right levels of automation that ultimately is reasonably comparable to what a human agent could possibly achieve. And I think most companies that start small, they all see the value quickly. They get a taste of it. And then they start really expanding their usage. And that's certainly what we've seen across our customer base. And the other thing I would add is like really making sure that, you know, we're a partner to our customers, not just like a random piece of software that is being procured.

13:21And I think this one is a little bit more well known now in the industry broadly. but S &B is probably not as fast on some of the AI adoption in some of these cases but I do anticipate with the things that we're doing that there would be massive acceleration in the next couple of years because what you might say as ROI is quite frankly just too high to ignore. The other thing is companies can scale customer conversations without scaling headcount. I mean, so your funnel can widen without having to take on more people. Yes. Can you talk about embedding AI directly into daily workflows and and knowing when to involve humans and and your customers?

14:19Are they contact centers or do you partner with contact centers because the AI cannot handle everything? And excuse me, most of our customers are not contact centers. They might use, they might have an offshore contact center or onshore contact center that they, you know, then procure to who might procure air call. But most of our customers just due to the size, it's more of a direct relationship with air call as opposed to, hey, I'm going to a contact center. And then, you know, the air call just happens to be inside that contact center. Excuse me. And so maybe, you know, that helps maybe set the stage a bit to answer the first question, which is, you know, how do we know, if I were to repeat it back, like, how do we really help customers figure out when to insert humans in these AI workflows?

15:10And I think there's maybe two main ways. The first is like, as it pertains to the business that we do and what customers do on our platform, we have one suite of products that we call assistance technology that is simply in the background assisting humans anyways. So in those cases, it's pretty much all human-led, except now AI is much more in the background. We have a lot of live assistance technology that can really hear, you know, tone, pick up on words that are key topics, reps who can follow a playbook. It really helps speed up the training and onboarding of new reps. And inevitably, there's a lot of business out in the world where even if the AI can perform like exactly the same as human, their customers just feel better about talking to a human.

16:00Right. And in those because it's a relationship business. And and now I don't know if that's definitively true down the line, but certainly the businesses think so. And I think there's reason to believe that their customers might think so. So in those cases, I get it. Like some customers might be pretty hesitant to have voice agents talk to humans because it's not part of their brand. They're trying to convince a customer that it's going to be a longstanding relationship. And the last thing you want to do on day one of that relationship is to throw an AI agent at you. That does not come off like a long-lasting, valuable relationship business, or at least how one of those would operate.

16:45And so the way to really have AI help there is to be more in the background, to be able to train reps faster, to identify key things that they can object to or help out. Because humans, we could all use some help. We can't store everything in our brain. We certainly can't do everything just the right way at the right time. And so that's where the technology really, really does help. Now, there's different types of businesses that like to deploy AI, whether it's chat agents or voice agents. They're higher in transaction volume. They have a different way of doing business. and generally what we do with these types of companies when they deploy these agents is that there are very easy ways for us to instruct the AI agent and build a workflow with the AI agent in which escalation points are triggered and then the call would route to humans and to your team and that tends to be part of the design process we go through there are certainly tasks in which you can take like refunds, for example.

17:53Most companies don't want, unless it's like crazy volumes in enterprise world, but most companies in our world don't want like a refund workflow to necessarily have no humans in the approval, right? There could be fraudulent things happening, all that kind of stuff. So a customer would decide, okay, here's a bunch of topics that I want to escalate to a human. They can actually design the AI agent workflow with us to say when there's detection on topic A, let's route to this team when there's detection on B. And they can even tweak the knobs a bit on how much an agent should attempt to answer the question or solve the problem before escalating.

18:34And it's really interesting when you work with these customers, how some customers are a lot more sensitive to the companies who are like, hey, the moment there's frustration, just escalate because that's just how I want to run my business. And then there are others who are more in the deflection kind of mindset. It's like, hey, try to deflect this, you know, as long as we can. And these are just prompts that go into the AI agent. Try to deflect this as long as we can. And then only after, you know, three or four occurrences of frustration, then you. And, you know, I do think businesses have a right to choose how they want to weigh this tradeoff between deflection and customer service, maybe satisfaction.

19:15And we allow them to configure that. We obviously, depending on industry, would have some best practices. But otherwise, it is up to the business to decide how they want to tweak this workflow and decide when to move it over for human intervention. Hope that helps. I'm not wanting to talk to an AI. I've had the experience of stumbling onto an AI voice agent. And it's kind of a relief because, first of all, you can hardly tell at this point that it's not a human, but the conversation tends to be much cleaner and much, I don't mean cleaner, like people are swearing. I mean, just simpler, just more direct.

20:03It's funny that you say that. We have a lot of customers who over time have shifted towards this point of view. I do feel the more folks use it, the more they can see the benefits. Now, is an AI voice agent going to be as performant as your smartest agent who's been around, who kind of has all the knowledge of the things that are never documented? Probably not yet. But I do think that it is better than the average rep in the call center. And there's a few reasons for that. But one is, at least from the business point of view, I'll put it that way, but even from the consumer point of view, from the business point of view, they're actually more adherent to what you want them to do.

20:47They don't, you know, so they don't really go off script or do things that they don't, you know, they're not instructed to do. Earlier models were worse at this, but newer models are much better at it, right? So that's one thing. And then on the consumer side, I think one very easy to see quality is their patience level is infinite. They never get upset. Their tone of voice never changes. They're always patient with you. And I think sometimes businesses don't quite grok those two things immediately because they are comparing against your absolute best human agents oftentimes. times. And I think for the more experienced customer service leader, even sales leaders, they start realizing really they should be comparing against maybe the median and then they might have a different perspective.

21:40So I'll give you a couple of other interesting examples. We had a customer who, I believe this was out of Australia actually, but I've heard it from a couple of US customers as well, where they actually, instead of just forcing every customer to a a voice agent right off the bat, they first asked the customer if they would like to talk to a human or get faster service through an automated AI agent. And surprisingly, they found that people who selected the latter was much, much higher than what they thought initially. And I think it's because they gave the customer a choice of, you know, and a relatively accurate trade-off of choices, which is faster service, but understand as an AI agent versus, hey, do you really want to talk to a human?

22:29And I think a customer then weighs in their head, okay, do I have a simple thing that should get resolved? Or do I have some super complex one that I just want a human to handle? And by doing that, you really increase the operational efficiency with kind of how you run your customer service center. But more importantly, the satisfaction, like the CSATs of these operations are actually higher because you're giving more kind of optionality and control over to the end customer and they appreciate that. They can call back if something's not working well, right? So I think if it's well-designed, more and more consumers are going to move to the channel that they deem probably have a higher probability of solving these things.

23:16So that's one interesting kind of example that is starting to be more frequent. The percentage is just getting higher and higher where people in many ways prefer to be talking to an AI voice agent because they think it's going to resolve faster. And there's quite a few of those types of examples. But I could probably go on more on some other ones. But I'll leave it at that for the time being. Yeah, I mean, the other thing about voice agents is, well, maybe it happens with live agents too. I don't know, but you collect a lot of data on the back end that you can analyze and use to improve the service.

23:59Does that happen with a live agent? Are those calls recorded and then analyzed as well? it's it's it's funny because i um in generally if you are a modern you know if you're using a modern phone or contact center system the answer is going to be yes there are countries and regulations in which recordings are not allowed they're not they're banned unless you have prior consent so that's probably why when you dial into something the very first thing that this ivr tree or this tree might say is like, hey, your phone, you know, this call is being recorded for training, whatever purposes. That's a very specific law in the United States, in different states.

24:42Not all states even have the same, just like you would expect the U.S. to behave. But most countries have something like that, right? So when you're calling into a company's phone line to get service, I think the reasonable expectation is that, yeah, you're being recorded. It's honestly a little bit odd if they don't have that recording that kind of says that. If that happens for live agents, as well as AI voice agents, if anytime you're being recorded, it needs to be somewhat exposed to the customer. And that's just law across. So now, what I would say is that more than 40 % of the world are not on a modern contact center or phone system.

25:27I don't know the last time you saw a perhaps a desk phone or an old landline. There's still a lot of the world who's operating on perhaps a more modern version of that. But at the end of the day, it's not a digital service. So without it being digital, it's not something in which you can get transcripts. And I do think there's an argument to be made that right now, why there's so much waves in maybe customers looking at just modernizing, not even talking about AI, just modernizing from an old phone system or contact center to a modern one that's digital is because of the AI wave, that people are seeing so much potential with what AI can bring them that, hey, maybe in previous eras, just purely the sake of digitization was not enough to move the needle.

26:19Now, suddenly it's enough to spur a lot of movement. So I do think in the market that there is that element that's kind of also happening. So we'll see. But that's kind of the observation that I've had. I don't have any hard data to prove it, though. Yeah. What do you learn from the back-end data from Aircall's system? I mean, what kinds of insights do you gather? So we as a platform, we do not actually train on our customers' data or do things of that nature because ultimately it's our customers who have some way of ownership towards that data. So what we do is we get consent from our customers if they ever opt into a service in which Aircall needs to have, you know, probably do any form of analysis on their data or otherwise for Aircall purposes.

27:19But if it's simply just, hey, us parsing the data to show it to our customers, that is part of the general AI service that we have. And I'm not sure folks get the nuance. Like if we wanted to use it for our purposes, that's generally not something that, you know, a company like us do. Right. And I think most contact centers and things like that probably have a similar stance towards this, unless we have some kind of opt in program and et cetera. And that's a sacred thing. I wouldn't recommend customers potentially sign up for service that don't have that outlined very clearly. But if you're asking like, hey, what do our customers do with this data?

28:02This is where all the magic happens. And I think this is a, you know, post 2015, I would say, this is where you have tremendous amounts of innovation happening because this data, if Digitize transcribes and customers have ownership over it, they can feed it into a number of different services first, right? They obviously can feed it into the CRM and then use that in conjunction with other data, more structured data in CRM to get a lot more insights. So that's kind of a probably the number one use case all prior to a lot of the newer voice technologies. But nowadays, what our systems are all doing, and especially at Aircall, is we're providing a lot of live services.

28:47So we're transcribing this information live and showing it to the agents of our customers and then adding a lot more other value add live services on top. Some of them I mentioned earlier around detection of sentiment, detection of keywords in which the companies have configured other answers to detection of these things. And that just reminds these agents of what they should be saying, making their lives a little bit easier as well. And so that's, you know, I would say if you add all of these things up, it's almost like a lot of our customers now just have this really smart, I don't want to say omniprescent, but very aware agent that is an assistant, you know, it's more of an assistance agent that kind of accompanies you prior to your call that you can ask questions to, you know, in the middle of the call when they're surfacing things.

29:43And then post-call, as they just automatically take this transcription, they parse it, they make sure the right structured data is going to get fed into your various CRMs, right? And so this assistant is not only helping you kind of just save time and all the diligence work, but it's also making the key moments matter in your conversations. A key moment in a support call or a sales call. So you add all that up. You know, sometimes we like to think that this assistance technology is as valuable as a whole kind of sales coach, which you're probably paying upwards of, you know, 150 to 200 ,000 USD for.

30:25And that coach can't even survey all the calls. It can't be with you live on all the calls. At best, they're just taking a small sample. So a lot of our customers are really buying into that as well. Just they're seeing the significant enhancement in kind of how fast reps onboard, how much better and quality that they're keeping assured, let's say, with every single call that's happening through their human agents. Yeah. How do people implement this? Is this, I mean, is it primarily through, you know, a phone number, a publicly listed phone number? Is it through a chat bot on their website or?

31:10When you ask how to implement, like, you mean the end to end of Aircall or just the AI parts of it? Yeah, so if I'm an Aircall customer and I have, I don't know, a network, a global network of sales, you know, a sales force that's selling a complicated set of products that I manufacture, how where are people touching the air call system is it when they they call a company or when they look at its website or or you know yeah so so i think the first is like air call is uh is business facing uh for from the end consumer customer side they don't know what you know phone system you might be using so irrelevant to them of course you might run into an ai agent and it's an air call agent, but for their purposes, it's like any other voice agent.