Agentic AI has officially graduated from basic decision trees to unscripted autonomous agents, turning customer service into a massive $15 billion market. Startups like Decagon AI are pulling in a $4.5B valuation because of high enterprise demand. Hertz is already routing 50% of its phone interactions through AI voice, dropping its per-call costs by more than half. And the car rental giant is hardly alone. Companies are fast adopting AI customer service agents. However, about 80% of consumers still crave a human on the other line, and 40% of us would literally pay a premium just to bypass the bot. Fortune’s AI Editor Jeremy Kahn takes us behind the scenes to break down the delicate balance between instant resolutions and human connection.

Almost 80% of Americans Prefer Human Agents, but Why Is AI Growing in Customer Service?

Tech

fortuneaifeatures

AI is making customer service even more unbearable. Three second delay and it goes ha ha ha, and I couldn't talk to a representative. How often does someone come in and they're just like, 000 agent agent agent. In 2026, AI and autonomous agents have taken over customer service. A recent report says 88% of contact centers are now leveraging AI, with a quarter of them running on fully integrated automation. Most of the inquiries that happened today in five years, there's no question that AI. will be able to handle all of them, but there's a catch. Another study reveals that nearly 80% of Americans still prefer talking to a living, breathing human being. And the hidden costs of this automation, it's staggering. Research shows that simply disclosing a chatbot's identity before a chat even begins can plummet purchase conversion rates by 80%. There's always that little human touch that that we always need. Yet we're a generation obsessed with instant gratification. How do we square that? As the AI editor at Fortune, I focus on how AI is shaping business and the wider economy. With AI poised to transform customer experience, I wanted to talk with business leaders on the front lines of this shift to find out how they're adapting and what the future of business really looks like. Despite our collective craving for human connection, the AI customer service market has ballooned to over $15 billion in 2026. So agentic AI or AI systems that actually take actions in the world, whereas previously you had automated agents that had to follow a set decision tree. With the current systems, they actually are able to have much less structured flows of how they operate. One of the front runners leading the charge is Decagon AI. Founded in 2024, they've already hauled in $481 million in funding, commanding a $4.5 billion post-money valuation. That is a major indicator of just how eager enterprises are to adopt this tech. Decagon's mission building faster, smarter AI without damaging the customer experience, and it's working. They're maintaining a 93% customer satisfaction score while driving revenue from fully automated interactions with. Some individual contracts now clearing $1 million. What do you think that looks like going forward in, in 5 years? Most of the inquiries that happened today, I would say in 5 years there's no question that an AI agent will be able to handle all of them. In a lot of these mass, like huge volume use cases, the human touch doesn't really matter, right? It's, it's not like you need the human touch to book a new flight or to, you know, reorder a lost credit card. All you want in that situation is just like get it done. It's, it's a little bit more mundane. These are not canned responses. This is something way more personalized. It's very different from what you've tried before. You're kind of changing the customer's behavior. But get this, over 40% of consumers say they'd actually pay a premium just to guarantee they can speak to a real human representative. Around the time of the industrial revolution, people wanted to buy a Ford Model T because the presumption was the quality was higher because it was built in a factory. Now today, curated, handmade experiences. Seem to have more of a premium association with them. Look across the industry and you'll see giants leveraging AI in different ways. Take Chipotle, a behemoth with over 100,000 employees, while some, like Decagon, are adapting to customers' changing behavior, others are doubling down on the interpersonal interaction. What kind of customer experience do you really want people to have going forward? One of the things that I think sets us apart and makes us special is the human interaction. Hi, how are you? Those curated moments of human connection and service. Are always going to be a fundamental part of what the Chipotle experience is. Back in 2024, they automated their hiring process with an administrative AI system called Avocado. And what was the idea behind that AI system? The average restaurant manager was spending more than 10 hours a week on recruiting activity, 20% of their week just on that one task. It's a fraction of that now. And how has AI changed sort of what your daily routine looks like? It allows me to, you know, be in front with the guest, serve them, going over recipe cards, you know, tasting food, a lot of training, people development for Chipotle, you know, we always need that human touch, you know, we always need, uh, people tasting the food because that's what we sell at the end of the day. We definitely always gonna need that human aspect. I don't know what kind of day they might be having before they come here, you know, we give them a burrito, a bowl, you know, we serve them with gratitude, with a smile, you know. Thank you. That could make anybody's day. But to see the customer facing side of this revolution, I synced up with Hertz. As an international car rental giant, they handle massive call volumes. By deploying Decagon's tech since early 2025, they're now resolving 75% of customer inquiries autonomously. We want to take every single one of those calls that comes our way, and we want that to be as seamless for our customers as possible. To see this AI agent interaction, I tried giving Hertz's AI, Haley, a call. Hi, this is Haley. I see you have a rental scheduled for pickup. How can I help you today? Oh hey Hayley, I'd like to change the drop off location. Let me see, where would you like to drop off the car instead? Yeah, if I could drop it off at LaGuardia Airport, is that possible? Great, you're all set with the new drop off at LaGuardia. Yeah, that was pretty good. I mean, I think that's, you know, that's was smooth, and it definitely didn't have to wait, which is nice not to have to queue at all. That's the new expectation. They're going to expect faster. Response times, better experiences. When we look at the future of contact centers, we're gonna see more and more AI agents taking on this workload. Every business will have an AI agent that's kind of like the front door to me that actually isn't to say that the human agents will go away because there's gonna be a lot more things that can be handled. Like why are people doing more of the mundane conversations because they have to, um, like those are the bread and butter. But if the AI can take that away, it frees them up to do more conversations that are not being had today. In fact, Hertz told us that AI voice is now live across 50% of their total phone interactions. In 2026, the cost of a customer interaction dropped by more than half to under $1.50. That shift means companies like Hertz are now seeing roughly $3.5 back for every $1 they put into AI customer service. Yet, depending on the issue, Hertz still reroutes calls to human agents. Hi, this is Hayley. How can I help you today? Hi, Haley, I've got a Hertz rental car right now and I have a flat tire. I'll connect you to a live agent now. Thanks for your patience. Thank you. They've got a flat tire or maybe they're on the side of the road and they're at risk. They absolutely want to hear a human voice and have that level of connection where someone understands that they may be at risk and is doing everything they can to support them in that scenario. Earlier you mentioned that billing is an area that still reverts to the human agent. Why is that? Why, why is the billing so important to be handled by a human? Yes, we think that the billing conversation is one about trust and the customers looking to understand why maybe a charge is not what they believe. It should be. So that's one where we very much default to someone being able to explain those pieces to them, and we found customers much prefer that. But even with those undeniable margins, the trust gap persists. Studies show that only 13% of consumers completely trust AI. Generally speaking, people will say I would be a little bit nervous about that, or in some cases a lot nervous or being very careful to consider PII or other information that might be in those systems and make sure that the models are both protected and secure from that perspective. that were responsibly handling people's information. How do you make sure there aren't hallucinations, because there have been these cases, I mean this famous Air Canada case where the judge said Air Canada was sort of liable for that incorrect information that its AI system had provided as part of a customer support experience. The only way to prevent that in a nondeterministic system is to have all these checks around. We call them supervisor models, and you kind of think of it as like the main agent is executing and you have these like, you know, small models around it that are running in parallel. There's all these models around it to make sure it stays on track and stays within the guard rails. Data proves we want more human interaction and will actively penalize businesses that deprive us of it. But as agentic AI gets increasingly good, will that sentiment hold? Are we staring at a future where consumers are forced to choose between instant resolutions or the human touch? Is there anything you think AI agents? We never really be able to automate when it comes to the customer contact, honestly, no, I think at a certain point it's almost indistinguishable. You'll get into a situation where if you didn't tell, inform the user that you might not know, consumers will now learn the other side, which they're, they're now used to interacting with great AI agents that they'll be disappointed if they're waiting on hold. We have very high conviction that that's where, where the world is moving towards.