메뉴
BL
Wired AI 5일 전

AI가 대체하는 세계 최악의 직업

IMP
7/10
핵심 요약

미국의 만성적인 물가 상승과 체불 채권 증가 속에서 추심 업계가 비용을 절감하고 효율을 높이기 위해 AI 콜봇을 도입하고 있습니다. AI 추심원은 24시간 불평 없이 업무를 수행하며 기존 콜센터의 낮은 직업 만족도와 높은 소비자 불만을 해소할 대안으로 주목받고 있습니다. 이는 고강도 단순 반복 노동이자 기피 업무였던 채권 추심 산업의 구조를 완전히 바꿀 수 있는 중요한 비즈니스 전환점입니다.

번역된 본문

댓글 로더 스토리 저장 스토리 저장 댓글 로더 스토리 저장 스토리 저장

그녀는 자신을 이브(Eve)라고 소개했지만, 벤(Ben)은 전화선 너머의 목소리가 봇이라는 것을 바로 알아차렸다. 이브는 그의 이름을 알고 있었다. 또한 그가 이전 집주인에게 빚지고 있던 금액(266달러)도 알고 있었다. 하지만 그가 5개월 전에 추심사와 합의를 봤다는 사실은 모르는 듯했다. 이브는 자신이 추심 회사 프로콜렉트(ProCollect)의 AI 에이전트이며 빚을 추심하기 위해 전화했다고 말했다. "오늘 카드나 계좌 이체로 해결하시겠습니까?" 그녀가 물었다.

벤은 오리건주 포틀랜드의 따뜻한 4월 오후, 전화를 받기 위해 밖으로 나왔다. (그는 재정적 문제에 대해 자유롭게 이야기하기 위해 와이어드(WIRED)가 가명을 사용해 달라고 요청했다.) 햇빛 아래 서서, 벤은 이브를 어떻게 해야 실제 상담원으로 전화를 돌려줄지 궁금했다. "상환 구조나 더 기술적인 내용을 물어보면 그냥 사람에게 전화를 넘기겠거니 했죠." 그가 말했다. 하지만 이브가 계속 통화를 유지했기 때문에 벤도 그렇게 했다. 그는 '왜 안 될까?'라는 생각에 봇을 좀 놀려보기로 했다. 벤은 자신이 '그냥 꼬마'이고 자신의 빚이 그를 짓밟으려는 거인 여성 같다고 설정하는 롤플레잉을 봇에게 요청했다고 말했다. 그는 이브가 얼마나 이상한 반응을 보일지 보고 싶었다. 그는 봇이 몇 분 동안 더듬거리며 그의 말에 맞장구를 쳤지만, 이내 갑자기 콜센터 직원에게 전화를 돌려버렸다고 말했다.

실제 상담원은 벤과 AI 사이의 기이한 대화를 들었는지 밝히지 않았다. 하지만 그들은 혼란을 빠르게 해결해 주었다. "시스템에서 저를 조회하더니, 잔액이 0달러인 것을 확인해 주었죠." 그가 회상했다.

벤의 경험은 점점 더 흔해지고 있다. 인플레이션과 정체된 임금이 사람들의 지갑을 옥죄면서, 미국의 채무 불이행 비율이 급증하고 있다. 채무 조정 전문가 마이클 보비(Michael Bovee)는 "지금 법원에 계류 중인 추심 건수는 제가 본 것 중 가장 많은 수치입니다."라고 말했다. 전례 없는 많은 사람이 빚을 갚기 위해 고군분투하는 가운데, 채권 추심 회사들은 노력을 배가하기 위해 기술에 눈을 돌리고 있다. 사람들이 돈을 요구하는 전화, 이메일, 문자, 편지를 받는 대부분의 작업은 이제 AI 에이전트에 의해 수행된다. 이들의 어조는 예의 바르고 심지어 아첨하는 듯할 수도 있지만, 결코 화를 내지 않는다. 또한 절대 잠을 자지 않는다. 이들의 강점은 끈기와 규모에서 나온다. 추심 대행사 카플란 그룹(Kaplan Group)의 분석에 따르면, AI 채권 추심은 향후 10년 내에 약 160억 달러 규모의 산업이 될 것으로 추정된다.

AI 지지자들은 자동화가 고도화됨에 따라 인류가 세계에서 가장 끔찍한 직업을 없앨 수 있는 흔치 않은 기회를 맞이했다고 종종 강조한다. 콜센터에서 일하는 것 자체도 이미 끔찍하다. 특히 사람들에게 돈을 갚으라고 조르는 콜센터에서 일하는 것은 비참함을 배가한다. 커리어 매칭 플랫폼 커리어익스플로러(CareerExplorer)는 채권 추심원을 직업 만족도 하위 1%로 선정했다. 추심원들이 자신의 일을 싫어하는 만큼 사람들도 추심원을 혐오한다. 소비자 금융 보호국(CFPB)이 채권 추심에 대한 불만 접수를 처음 시작했을 때, 6개월 만에 11,000건의 불만이 접수되었으며, 이는 금융 서비스 중 가장 많은 분노를 샀던 모기지 산업 바로 뒤를 이었다. 만약 일자리 상실에 대한 큰 반발 없이 사라질 수 있는 산업이 있다면, 아마도 이 분야일 것이다.

이브 같은 봇에게 지구상에서 가장 미움받는 직업을 대체하기 위해 무엇이 필요할까? 이브의 능력을 더 잘 파악하기 위해 직접 전화를 걸어보기로 했다. 하지만 벤이 알려준 번호로 전화했을 때, 프로콜렉트의 실제 직원이 전화를 받았다. 나는 기자라고 신분을 밝혔다. 상대방은 나의 질문에 답해 줄 사람이 없으며 내일 다시 전화해 달라고 했다. 다음 날 다시 전화했더니, 다른 직원이 회사는 AI를 사용하지 않는다고 말하면서도 인사부서와 이야기해 보라고 했다. 인사부는 이메일로 질문을 보내달라고 했고, 나는 그렇게 했다. 나의 질문 중 하나는 이: 이브는 어디서 왔는가? 였다.

추심 회사의 답변을 기다리는 동안, 벤은 친절하게도 직접 이브에게 다시 전화를 걸어보겠다고 자원했다. 그는 전화를 연결했고, 봇을 자극해 누가 이를 만들었고 어떻게 훈련되었는지 알아내려 했다.

원문 보기
원문 보기 (영어)
Comment Loader Save Story Save this story Comment Loader Save Story Save this story She introduced herself as Eve, but Ben knew right away that the voice on the other end of the line was a bot. Eve knew his name. She also knew how much money he’d owed a former landlord ($266). She didn’t seem to know that he’d settled with a collection agency five months prior. Eve said she was an AI agent from ProCollect and was calling to collect a debt. “Would you like to resolve it today by card or bank transfer?” she asked. Ben had stepped outside on a balmy April afternoon in Portland, Oregon, to take the phone call. (He asked that WIRED use a pseudonym so he could speak freely about a financial issue.) As he stood in the sun, he wondered what he’d have to say to make Eve hand off a call to a human. “I figured it was just going to kick me over to a person when I asked about repayment structure or anything more technical,” he says. But Eve stayed on the line, so Ben did, too. He decided—why not?—to mess with the bot a little. Ben says he asked the bot to engage in some role-play, in which he was “just a little guy” and his debt was like a giantess prone to trampling him. He wanted to see how weird Eve would get. The bot haltingly played along for a few minutes, he says, but then abruptly punted him to a call center employee. The human agent didn’t disclose whether they’d heard Ben’s bizarre conversation with the AI. They did, however, quickly clear up the confusion: “They looked me up in the system,” he recalls. “Found that the balance was zero.” Ben’s experience is increasingly common. As inflation and stagnant salaries squeeze pocketbooks, debt delinquency in the United States is swelling. “We have, right now, the highest amount of collections in the courts that I've ever seen,” says debt settlement expert Michael Bovee. As an unprecedented number of people struggle to repay debtors, the companies chasing down debts are turning to technology to amp up their efforts. Many of the calls, emails, texts, and letters people receive asking for money are now carried out by AI agents. Their tone may be deferential, even sycophantic, but they never fly off the handle. They also never sleep. Their edge comes from persistence and scale. An analysis by the collections agency Kaplan Group estimates that AI debt collectors will be an industry worth nearly $16 billion within the next decade. AI boosters often stress that, as automation becomes more sophisticated, humanity has a rare chance to get rid of the worst gigs in the world. Working at a call center already sucks. Working at a call center specifically to hound people for money compounds the misery. Career matching platform CareerExplorer ranks debt collection in the bottom 1 percent of professions for job satisfaction. As much as debt collectors hate their jobs, people also hate debt collectors. When the Consumer Finance Protection Bureau first began accepting complaints about debt collection, it received 11,000 within six months, putting it just behind the mortgage industry as the financial service that provoked the most ire. If there’s any employment sector that could go poof without too much fuss over job loss, it might just be this one. For a bot like Eve, what does it take to beat the least-liked people on Earth? To get a better handle on Eve’s capabilities, I decided to call her myself. But when I tried the number Ben gave me, a human ProCollect employee answered. I identified myself as a journalist. They told me that no one was around to answer my questions and suggested I call back the next day. When I did, another human told me that the company does not use AI but also that I should talk to human resources. HR told me to email my questions, which I did. One of my queries: Where did Eve come from? While I awaited a reply from the collection agency, Ben graciously volunteered to try calling Eve back himself. He got through and prodded the bot to reveal who created it and how it was trained. Eve demurred on the first question and spat out a generic answer to the second: “I use a mix of conversation patterns and account-related training to help.” Ben said he still had questions. Eve asked if he wanted to speak to a human. Ben said he did. Eve said the office was closed. Because debt collection is such a no-duh industry to automate, any number of companies could have been involved in creating Eve. Pedro Fernández, cofounder of the startup Altur, which operates a “human-less call center,” says collections agencies were among the industry’s “best early adopters.” Altur’s agents complete more than 2.5 million calls a month about debt issues, Fernández says, and the company’s clients include major banks in Mexico. Altur is one of half a dozen debt collection and settlement startups to have incubated at Y Combinator in the past six years alone. Some already operate on an even larger scale, like Domu, which was founded in 2023 and “automates debt-collection calls, texts, and emails” for clients in health care and financial services. Domu describes itself as a company “doing what nobody else is willing to do.” According to Isaac Choate, Domu’s head of product and growth, the company’s AI agents hit 70 million monthly connected calls in March. These companies go to great lengths to make their agents as deft as possible. Choate says Domu’s agents adjust how they speak based on who they’re calling, even varying their Spanish accent based on whether they’re calling someone in Mexico versus Colombia. “Situational tone” is crucial, he says. “An agent that's handling a hardship conversation is going to sound a lot different from one handling a routine reminder, which is more of a friendly, neighborly sort of conversation.” Companies have different thresholds for when they shift calls over to humans. Mentioning bankruptcy, or raising anything categorized as a “vulnerability signal” (illness, a death in the family), will trigger Altur’s agents to transfer the call to a human, while other companies will allow their AI agents to continue talking. A chipper young salesman from another AI debt collection startup, Moveo, showed me how agents review information on each person they call. Moveo’s agents adjust how they speak based on whether someone brings up a medical condition, job loss, or lawyers; they even create “psychographic profiles” for the humans they call by analyzing transcripts of prior conversations. Debt collectors are notorious for badgering their targets, but they do have rules to follow; laws like the Fair Debt Collection Practices Act in the United States are meant to shield people from manipulative tactics, like threatening to tell someone’s spouse about the money they owe or calling repeatedly in the middle of the night. All of the AI startups I spoke with stressed that their systems have been designed to comply with existing laws. Some industry insiders wonder if that’s enough. Credit counselor Martin Lynch worries that a buggy agent could inadvertently disclose a debt to the wrong person—which would break the law—and calls the industry “a minefield of legal trouble.” Despite the risks, evangelists for AI collectors see them as a kinder, gentler alternative to an old-fashioned, crabby, low-paid human chasing down payments. They stress that bots can get a surprisingly positive reception. In some cases, they say, the issue isn’t that people don’t want to speak to the AI agents—it’s that they don’t want to stop. “We've had cases of people asking the agent for a personal number to keep talking and a few where the person pretty clearly developed some interest in the voice on the other end,” says Altur’s Fernández. “Not something we expected to see.” A lot of AI debt collection agents sound like women. As in the wider AI industry, consumers are coaxed to forget that the agent they’re talking to is not a sentient being with thoughts, feelings, and a gender identity. Domu calls one of its agents Taylor and