When Self-Service Backfires: Why Anxious Customers Need Human Reassurance

What’s the cost of cutting people out of the process? And can companies deliver the efficiency of automation without sacrificing the connection customers crave under stress?

To answer these questions, my co-author Ryan Buell of Harvard Business School and I conducted one field experiment and four lab studies set in financial service contexts. As we report in our paper, which was recently published in the journal Management Science, we found that anxious customers using self-service technologies become less confident in their own decisions—even when those decisions are objectively aligned with their goals. This internal doubt leads to reduced trust in the company offering the service—although the company neither had control over the sources of their anxiety nor influenced their decisions. But our research also points to a powerful and surprisingly simple fix: just making it easy for customers to connect with a person can offset these negative effects.

Our field experiment took place at a U.S. credit union that was launching an online loan application system. Their goal was to increase the percentage of approved applicants who followed through and closed their loans. We randomly assigned over 300 applicants into three groups:

1. No contact until their loan decision was made
2. Text messages with updates on the review process
3. The same messages, plus a loan officer’s phone number and an invitation to reach out with questions

Providing status updates alone didn’t improve outcomes. In fact, the second group was slightly less likely to close their loans than those who received no updates at all. A follow-up study offered an explanation: reminders that they were being evaluated—even though expected—amplified feelings of anxiety. In contrast, the third group saw completion rates jump from 64% to 80%. Transparency only helped when it came with the opportunity to connect with a human.

We then explored the psychological underpinnings of this effect in the lab. We created a simulated online retirement planning platform where participants invested a hypothetical $100,000 portfolio across stocks, bonds, and cash. In our first experiment, we made it so that some participants experienced the normal ups and downs of U.S. market history while others were exposed to a higher chance of returns that reflected the worst years of the U.S. stock market.

Even when participants’ made decisions that produced investment gains while the stock market fell, those in down-market conditions reported higher anxiety and lower satisfaction with their decisions. The result: reduced trust in the service provider—even though the firm had no control over outcomes.

We replicated this effect in another study where all participants received identical returns, but some watched a stressful movie clip beforehand. Despite facing equal stock market performance, anxious participants again felt worse about their decisions and trusted the platform less. This showed that it wasn’t just poor outcomes of the service that hurt customer confidence—feeling anxious for any reason was enough to damage the experience.

In a follow-up study, we tested whether a human connection could help restore satisfaction and trust. Participants were split into market conditions as before, and split again into three groups: no support, the option to chat with another investor, or the option to chat with an expert. Few used the chat—but simply having the option significantly improved how anxious investors felt about their decisions and the company.

However, when we offered those chat options to participants who were made anxious by the movie clip, it did not have the same effect. The human connection only helped restore decision confidence and trust in the firm when the anxiety was inherently linked to the service itself.

One of the most surprising findings was how rarely participants used the available human support. Just knowing that help was available was enough to boost satisfaction and trust. This suggests companies don’t need to flood customer journeys with live agents—but they do need to make human help visible and accessible in those moments when self-service customers are likely to feel high anxiety.

What’s the bottom line? As companies scale self-service systems, they should pay attention to customers’ emotional states. In high-anxiety settings, the convenience of self-service technology isn’t enough to garner customer satisfaction. People need reassurance as they make high-stakes decisions, even if they never actually ask for it.

Just letting customers know that a real person is available—a phone number, an invitation to connect, a visible chat feature—can go a long way toward restoring confidence and building long-term trust.