Words at work

– 9 min read

How does Amazon approach customer support now?

Amy Cuevas Schroeder

Amy Cuevas Schroeder

“Money is an emotional thing, and many people live really close to the margins.”

That’s what Carol Valdez keeps top of mind when she and her team write the copy that millions of Amazon customers see in emails, chats, and across the retail site every day. Valdez believes money is likely an especially sensitive topic right now, during the COVID-19 pandemic.

“Many people don’t have a lot of wiggle room in terms of their finances. The checks and balances of a user’s experience are important. You can’t assume that, ‘Oh, this transaction is just about a couple of bucks.’ You can’t be cavalier in terms of finance,” she said during a Zoom interview from her home in Seattle in May 2020.

As the principal UX writer and voice design lead for Amazon Customer Service, Valdez spends a fair amount of time thinking about people’s sensitivity about money. She started learning about people’s emotions around money when she worked as the principal content designer for TurboTax, in 2012, and before that, she started her career as a language teacher. 

Valdez works on an Amazon UX team of about 25 to 30 people who sit within Amazon worldwide retail customer service. “We’re a small but mighty craft,” she said. Valdez manages a UX writer and voice designer, and her team supports both customer service–facing products and Amazon associate–facing products. Her primary responsibilities include working on tools that are intended for retail use cases, such as post-purchase and on amazon.com. Valdez’s team works on customer-facing tools, including chat, text-based chat box, voice box on the phone, voice forward Interactive Voice Response (IVR), and the Alexa help domain getting customer service on Alexa-enabled devices.

Carol Valdez of Amazon
Valdez Valdez is the principal UX writer and voice design lead for Amazon Customer Service.

Do you concentrate on messaging for particular product categories or particular kinds of customers?

I work across product categories. Amazon serves so many people. Solving problems at scale means hitting the sweet spot for most people, which lands you in a neutral place. I’ve always realized how much pressure I’m under because the software has to work for a huge audience. This feeling started when I worked for Turbotax, and it’s the same here at Amazon.

When people contact customer support, there’s usually some sort of emotion involved. If a delivery was a bit late, people aren’t usually going to contact customer support. If there’s another factor, we keep that in mind. What are the customer’s emotional needs, and how do we address them in a scalable way? We’re trying to automate what is usually a human interaction. We have to find a commonality. 

We ask ourselves, ‘What can we do to predict what customers might need help with and make them feel like we’re taking care of them? Can we leverage something that we already know about the customer to help ‘read their mind’ to a certain degree?’.Click To Tweet

Money tends to be an emotional topic for people. When customers are waiting to get their money back for a return, you need to reassure them of when they’ll receive the refund and send a  reassuring follow-up communication.

Also, we ask ourselves, “What can we do to predict what customers might need help with and make them feel like we’re taking care of them? Can we leverage something that we already know about the customer to help ‘read their mind’ to a certain degree?”

How are things going for you and Amazon’s customer service team since the onset of the pandemic?

We’re all remote now, which has its ups and downs. A lot of us are surprised that there’s a silver lining to the pandemic. The challenges are real, too. Things are crazy for a lot of the people who work at the heart of customer service.

When the pandemic first hit the U.S., Amazon was hit really hard with traffic. It was all hands on deck, and we were always available because everyone was working from home. It was a first-responder situation — we’ll get in there and help the business get through this. Working through the pandemic has been a nice team-building yet exhausting situation. 

Now that we’ve made it through the initial storm, we want to find ways to continue to connect with the broader team, which is the biggest challenge. You’ll always have your core people you interact with, but there are others who you might have just run into getting coffee — how do you replace that remotely? How do you find ways to spend time with coworkers beyond your core group? 

What are your team’s top priorities now — are they different than pre-COVID-19?

There aren’t a lot of quick solutions for solving problems in a pandemic, but messaging can happen quickly, which is work for writers. That was the biggest strain on me and my team — to be available to write things on the fly.

Now that the initial traffic surge and demand for new messaging has evened out, we’re adjusting to a new normal. Our priorities are largely back to what they were pre-pandemic — we’re working on the same software just like before. Some people’s road maps shifted a bit, but now things are largely normalizing.

As for working remotely, people like the lack of commute time and the convenient work-life balance and seeing your family a bit more than before.

Have you had to do any pivoting or make significant changes in your approach, communications style, or messaging with customers?

We pushed some of the work on the docket; for example, Amazon Fresh, our grocery service, took priority at the beginning of the pandemic. We also prioritized writing the most common questions and answers. Overall, we’re now back to our original work.

What new ways of working have worked well for you and your team since the pandemic hit? Are there any particular approaches or kinds of messaging that you’ve found to be helpful when communicating with customers?

I’ve always been a big plain language advocate. While I don’t try to police grammar, I do try to police plain language to make sure we’re speaking to people in the simplest terms, of all reading levels. That’s another reminder as we do this kind of messaging.

There are more people and players — more people I need to give this message to. I evangelize the plain language message moreso now but not any differently. I’ll say, “Is there a simpler way of saying this? How might this be understood by multiple people?” The same approach as always but now it’s visible to more people at Amazon.

How do you approach writing English copy that will be translated into other languages?

Our team supports English version UI text — but it gets localized into many other languages for different markets. The main, constant challenge is about writing in a way that’s fluid. People will say, “This message might not work in Japan — what should I do, Carol?” I’ll say, “I don’t know — I don’t live in Japan.” There’s room to improve that process. More and more, we’re annotating notes about what we’re trying to achieve within a particular customer conversation. For example, “Here we’re trying to be reassuring” or “We wrote this to help with clarity.” So at least it could have a playbook to explain why we chose what we did. Maybe the Japanese doesn’t need as much reassurance. We’ve been adding annotations to our work. This is a huge opportunity [in the UX writing world] — things get lost in translation, like they say.

How do you use AI to support your work?

When we’re helping customers solve problems, we hone in on what the data tells us about what they typically need help with. This helps us provide the right experience for them. 

In our software for associates, we want AI to make it easier for them to help customers. Our customer service folks are great at talking with people — that’s their top skill. My team makes the software work on customer support agents’ behalf to help predict solutions. The machine suggests things that people might need help with. Our associates can listen to the customer and help make suggestions using AI, instead of having to hunt for the right resources.

Within conversational UI, the goal is to teach machines to speak. Like I said, UX writing is a small but mighty craft, so we’re never going to have enough writers to write every word of dialogue for bots. We collaborate with the machine learning team to help the machines speak for themselves. It’s tricky and challenging, but we’re making good progress.

Do you ever worry that AI might replace the need for UX writers?

No, but my goal is to make my job obsolete. I believe there will always be some unmet need that requires specialized human help. AI thrives on patterns, but there will always be spaces where there aren’t patterns they can draw on. That means there will always be space for people to be part of the loop.

For example, who wants to be policing grammar if you could use machines to do that? I’d rather work on problems that are much harder to solve. There’s always going to be enough work to go around to sink your teeth into.