Customer help desks are a top target of money-saving schemes involving generative AI, as I reported earlier in an attempted takedown of the chatbot. Now I’m convinced that automating customer service will bring the apocalypse somewhat nearer, not because machines will become sentient, but because the machines turn perfectly normal human beings into morons.
It started when a family member included in my T-Mobile plan lost a handset. It happens.
Each month I fork over $18 to T-Mobile’s partner Assurant for handset protection, so I clicked on Assurant’s website to file a claim. But the website went into an infinite loop: To file a claim, the site demanded a one-time password sent to the handset in question, which is the handset that was lost, and therefore couldn’t receive a one-time password or anything else.
Use another handset from the same account to get the oneoo-time passwor, Assurant’s convenient pop-chat function told me. I should have known better. ChatBots make things up, as a couple of New York lawyers discovered when their AI-generated legal brief presented cases that didn’t exist. The same thing can happen with help desks.
Stupidly, I did as instructed. But Assurant’s system assumed that the lost handset needing tobe replaced was the one that had received the one-time password, namely my own handset. After I forked over a $250 deductible, Assurant promptly sent out a replacement for my handset, a venerable Samsung device that had worked uncomplainingly for years and remained comfortably nestled in my shirt pocket.
Once I was in receipt of the wrong replacement handset, I called Assurant customer service and explained the error (it was Assurant’s error, not mine). I returned the handset with the next UPS pickup. Assurant apologized for the error, and took a second $250 deductible to replace the handset that actually had been lost, promising to refund my first $250 deductible when the first replacement handset arrived.
In due course, the correct replacement handset arrived, and all seemed well, except for the deductibles, of course.
But I was unaware that the unquiet spirit of Franz Kafka had quietly possessed the soul of T-Mobile and Assurant customer service.
My handset stopped connecting to the network in the middle of a weekday afternoon, ten days after my initial encounter with the Catch-22 on Assurant’s website. I was traveling and between meetings. I had a backup phone (I’m the sort of Airport Dad who always has a backup) and I spent two hours on the line with T-Mobile customer service until someone figured out that Assurant, not T-Mobile, had blocked the SIM card on my handset, because it had been reported lost.
It never was reported lost – its only offense was to receive a one-time password – and Assurant customer service had known of its error for ten days. But the misinformation had worked its way through the system over those days.
At length I talked to someone at Assurant, who said that it would take one business day, or maybe one to three business days, to unblock my SIM card. I begged, cajoled and threatened the Assurant tech as well as her supervisor, to no avail. The phone still isn’t working.
Fortunately, most of my friends have stopped using cell phone connections in favor of encrypted messaging apps, and these apps run on WiFi. I can turn on the data hotspot on my backup phone and use the messaging apps on my handset with the still-blocked SIM Card.
A supervisor finally took my call at T-Mobile. I asked her if I could record the call (which T-Mobile says it does whenever you call them). She hung up. Her AI system probably flashed, “Abort! Abort!”
If you’ve been trying to call me on my regular line, and I haven’t answered, now you know why.
T-Mobile, Verizon and AT&T cost a lot more than the discounters. What you pay for is superior customer service, which means a lot in an emergency. But as AI systems metastasize through customer service departments, the advantage disappears. If you’re going to be mistreated by a machine, might as well pay less for the privilege.