On a balmy April afternoon in Portland, Oregon, Ben got a call from Eve, an AI agent from ProCollect, asking him to pay a $266 debt. He says the bot knew his name and the amount tied to a former landlord, and it stayed on the line even after he tried to make the conversation sound like a joke.

“Would you like to resolve it today by card or bank transfer?” Eve asked, according to Ben, who tried to steer the call toward repayment details and then into role-play, telling the bot he was “just a little guy.”

What made the call stand out was not only that the collector was software, but that the software kept going. Ben said he expected the moment he started asking about structure and other technical details to send him to a person. Instead, the bot played along for a few minutes, then transferred him to a call center employee who looked him up and found the balance was zero.

That result is where the story lands hard. Ben says he had settled with a collection agency five months earlier, but Eve did not seem to know that when she called. The gap between what the system was trying to collect and what a human later found in the file captures the central risk in automated debt collection: a machine can be persistent, polished and wrong at the same time.

The practice is spreading because the business of chasing overdue bills is changing fast. Debt delinquency in the United States is swelling as inflation and stagnant salaries squeeze pocketbooks, and companies are increasingly using AI agents to make calls, send emails, texts and letters at scale and without sleep. An analysis by the collections agency Kaplan Group estimates AI debt collectors could become an industry worth nearly $16 billion within the next decade.

That growth sits uneasily beside a job many people already dislike. CareerExplorer ranks debt collection in the bottom 1 percent of professions for job satisfaction, while the Consumer Finance Protection Bureau received 11,000 debt-collection complaints within six months when it first began taking them. Michael Bovee, who follows the field closely, said, “We have, right now, the highest amount of collections in the courts that I’ve ever seen.”

For Ben, the call ended with a zero balance and an awkward handoff to a person who could see what the bot had missed. What remains unanswered is how often automated systems are still pressing people for money that has already been settled, and whether the companies deploying them are keeping records clean enough to stop it.