Improving customer experience through personalized voice and chat bots

Ask yourself (or your boss, or your team) why are we leveraging conversational technology? Here are some good reasons our clients have presented us for exploring chat and voice bots.

• Providing 24-7 support for their customers (without staffing 24-7)
• Reducing support call volume for topics that have common solutions
• Simplifying valuable actions that can be automated for their customer with a simple prompt
• Leveraging machine learning models to improve personalized, yet consistent, customer experience (humans are... well, human, and bring their mood to work with them sometimes)

Using the Jobs to Be Done framework, we work with our clients to understand why their customers would be “hiring” voice or chat to complete this task rather than another channel. Our goal is typically to help a customer complete a task or consume information. Common reasons we’ve seen customers utilize voice or chat include the following:

• I want to get support, when it is convenient for me (not necessarily during work hours).
• I want a more accessible way to connect with the brand for customers with vision, mobility, or hearing impairments.
• I want to be able to multi-task and use voice to be hands-free.
• I want to document my interactions with a brand for reference.

One of our automobile manufacturer clients worked with us on a strictly in-car experience. This meant very little or no dependency on a visual UI and brought unique challenges, like other people talking in the car.

These are the roles we consider the bot providing support for. It’s important to identify which experiences it’s going to personify, for the user to meet their needs.

• Receptionist: Chatbots can greet your online users and offer to direct them to the information they need.
• Account Clerk: Chatbots can take a customer’s details and check their account to retrieve information. For these kinds of routine, repetitive account chores, chatbots are an effective (and often invisible) service tool.
• Sales Person: A chatbot could be programmed to start conversations based on what a user is browsing, such as a feature set or price list or suggest deals and answer any product questions.
• Shop Assistant: Commonly asked FAQ’s can be answered by a chatbot effortlessly, and allows the bot to fill in any knowledge gaps.

If you are like most brands, you’ve already been interacting with your customers… What are they asking? One of the first things we do with our client is buddy up with their customer insights and support teams to look at the data. What type of support is requested the most? Alleviating what type of support would provide the most return on investment? What support would be most viable via a chat or voice interface?

When creating a voice experience around something as complex as a restaurant menu (and all its many modifications) it is important to have a knowledgeable designer or architect onboard.

Your everyday person, the eye-rollers I mentioned earlier, typically find bots to be lackluster, dumb, and more frustrating than they are worth. Therefore, it is essential to work with a UX researcher and convo UX designer to explore and evaluate the customer experience. Our UX team members analyze transcripts, create and test prototyped experiences, and carefully design flows and language that align with your brand to ensure we are setting the customer up for success and, dare we say, delight?

Well, there will always be the person who wants to get to a human as soon as possible. We try to accommodate this whenever we can, while noting advantages to solving the problem through the bot (usually wait time). There will also be things your bot doesn’t know. We make sure to provide fallbacks, or graceful ways to let the customer know we can’t help with that… yet. These become significant indicators in the analytics, which we will cover later in this series.

We work with our clients on bot personality spectrums based on existing brand guidelines and the willingness to purposely steer this channel away from what exists today. Leveraging roleplaying activities, we test the different ways a bot could affirm what the customer is saying, respond when it doesn’t understand, or even say thank you. There are benefits to being bot-like and human-like, so we collaborate to understand how we want to leverage those benefits and how far we want to push the “human” side.

It was important for one of our QSR clients that their brand’s witty personality came through without alienating their customers. Through customer research we were able to calibrate a tonality that was endearing and successful. Depending on where the customer is engaging, they may even be open to more conversational dialogue or small talk. Think about all the strange things you ask your Google Home or Alexa. What might someone ask your brand uniquely?

For a conversational designer, there are some emerging tools and activities for the trade and we’ve homed in on a few that keep the team aligned during our iterative process.

Transcript/Log Analysis: Customer Service Teams are your best friend at the start of a convo project. Leveraging existing transcripts and logs helps you identify high-volume support questions and determine how customers ask them, which can inform training phrases.

Conversational Flow Design and Copywriting: Even simple conversations will have complexities within their intents in order to ensure your customer does not get caught into a loop and that errors are handled gracefully. We use an iterative approach to test and learn how customers interact with a specific brand through conversational interfaces as automated technology is becoming the new norm and behaviors and expectations are changing. You can check out some other great tips for Voice User Interface Design from my fellow Kin + Carta strategist Becky Pierson.

In closing, I am expecting a lot less eye-rolling when it comes to conversational UX bot as we roll into the next year. The use cases around customer support and labor efficiency are hard to ignore. People are now more self-reliant and remote than ever. Bad bots can, are, and will still be created but we are seeing product definition and design mature in the conversation space so bad bots will become few and far between. As with any technology, as great experiences emerge, that bar will go up. We are looking forward to the challenge. Stay tuned for part three of our series, where we'll dive in deeper into the tech behind convo experiences.