Study finds AI debt collection callers persuade less than humans

In short: New research suggests AI can handle basic debt collection calls, but it is noticeably worse than humans at persuading people to pay and stay on track.

What's going on

Researchers studying delinquent consumer loans compared AI callers with human callers. The AI could understand speech, provide information, and answer questions. Even so, it was substantially less effective at getting borrowers to take a costly step, like making a payment.

In the study, calls with AI were about 31 seconds shorter on average and were more uniform. The AI also got fewer “binding” verbal promises, meaning clear spoken commitments that people may feel bad about breaking. The researchers suggest one reason is that people may feel less moral pressure when they are talking to a machine, compared with a person.

The gap showed up quickly. By day 5, the difference in expected value of repayments had widened, even when comparing AI to strong human performers. If a company switched borrowers to human callers after about six days of AI contact, much of the short term gap shrank.

But the earlier AI contact still mattered later. Borrowers first reached by AI repaid about 1% less of their initial late payments after one year. They were also more likely to miss future payments, suggesting weaker long term cooperation with the company.

What to watch

As AI spreads into customer calls and sales, this research points to a limit. AI can be like a helpful script reader (good at facts and routine questions), while humans are often better at reassurance and social pressure. Companies may keep AI for simple tasks, but rely on people when the goal is to change behavior, especially when trust and empathy affect the outcome.

Source: NYTimes