When the pandemic hit, most businesses closed their doors. From the 405 to the 3rd street Promenade in Los Angeles, not a soul was in the road. The only brick and mortar in town were grocery stores. People invaded those stores and snatched every last roll of toilet paper. Other businesses awaited quietly while they bled through their remaining cash. Even the online world was not spared. Customers were not sure how things would play out, so they saved their money. Or reserved it for toilet paper delivery.
I was lucky. I worked in customer service automation. Not only was it the right business to be in, it was one of the most thriving ones. When uncertain companies cut down the number of customer service agents, they relied on automation to fill the gap.
We saw an overwhelming number of people turn to customer service to ask questions.
"Please cancel my subscription!"
"When can I expect the next shipment of toilet paper?"
"Why is my package not here yet?"
It wasn't just retail stores that were affected. UPS, USPS, FedEx, they all suffered major delays. With little to no staff, customer service departments were completely overwhelmed. Their response time grew from 24/48 hours to 2 to 4 weeks.
To many companies, customer service automation came as relief. It was the only reason they were able to stay in business. In the industry, companies track their Customer Satisfaction or CSAT score. It's a simple metric that tracks the total number of positive customer responses divided by the total number of responses multiplied by 100. With automation they started seeing this number go up.
No one had answers. No one knew how long the toilet paper shortage would last. No one knew when FedEx would finally be able to deliver packages. One thing we knew was that if you gave the customer an answer immediately (in seconds), even if it is bad news, they'd feel better.
Honesty and speed were the key differentiator. No one wants to wait weeks to get a response. Positive or negative. When customers write a message, it is because they are frustrated now. Making them wait can only worsen the situation.
One client who was reluctant to use automation finally caved in. They signed a contract and we started automating their responses. It didn't take long for our AI to send one of the worst emails we could imagine.
I got a slack message from our client. All it said was "WTF!!!" And attached was a screenshot of the message we sent to their customer. It was a frustrated customer who was complaining about slow delivery. We responded with:
Hey Bitch, Thanks for reaching out.
I didn't even read the rest of the email. I promptly apologized to the client and promised to investigate right away.
When your company's tagline is "exceptional customer service", there is no other way to put it. We screwed up. I turned off our service and started investigating the issue. I pinged a few of my colleagues and we spent the next minute collectively going through the 12 stages of grief.
There is a study I often see quoted on LinkedIn. Calling customers by name validates them. It defangs a faceless corporation. It humanizes the customer. I believe it makes people feel subtly better, but the study might as well be a myth. I've never read or seen that study.
We added a feature to greet customers by name early on. But a name is a variable, it is complicated, it is whatever the person says their name is. "Henry Kills the Enemy" is a valid name. In fact, by reading it you can rightly assume that Henry is the first name. If you are in the US, you can even guess this person's heritage. But try to explain that to your program.
Our client blamed our AI? "Isn't the AI smart enough to know not to call someone a Bitch?" I learned a long time ago that explaining the inner workings of software engineering is always the wrong answer. Especially in a tense situation. Our machine learning application took no part in calling the customer names. Our codebase does not contain the word bitch, so I assumed this was human error somewhere.
Helpdesks are a type of application companies use to provide customer service. When you write an email to a company, the message doesn't just pop into someone's email. Instead, it goes into the helpdesk where your message is put in a queue for the next available agent to handle. These applications often save the customer's name, email address, and some other metadata to help the agent the next time around.
I logged into the client's helpdesk to look for that extra metadata about the customer. I was hoping that the customer name was "Bitch". Who knows, maybe it was just a rowdy customer or like I said earlier, names can be complicated. But no, the customer had a common name. But since I built the integration, I had admin access to the helpdesk. I decided to read the logs.
I saw that someone had changed the name. It went from Bitch to a common name. What did that mean? It meant that when the customer filled out the customer service form, they entered Bitch as the first name and Babe as the last. Because the helpdesk recognized the email, it added a recommend update button next to their name. Sometimes later, an agent clicked on the button to update to the original name. Thus replacing Bitch Babe with the real name. But not before we had already sent our message to the customer.
Now that I had answers and my heart was no longer palpitating, I was ready to send the client a message. To cover all my bases, I went to read the message and our response one more time. I noticed that the customer had sent a response.
"Oh my God, you guys answered so fast I didn't even notice. Thank you so much for the info."
The customer was happy. And they even left us a positive review. It wasn't the answer they wanted to hear, they were happy that we had responded in a timely manner.
It doesn't work all the time. I've seen customers be pissed because we responded too fast. i.e. "I don't want a robot, I want to talk to a manager!" But those are rare. The majority of time, speed is one of the most important factors in customer service.