For a job interview, I was given the exercise to create a script for a based on a current application. After connecting the dots that UI had everything to do with AI, voice, ML, and chatbots it a pure delight to try out.

In this article, I document my design process behind this short exercise with the goal of outlining a potential process for to use when practicing or designing for conversational UIs.

Researching Conversational UI

I started by reading this article that describes conversational UI in-depth calling it “interacting with the computer on human terms.”

After this cursory search of conversation UI, it is clear that the technology is creating the infrastructure that increases a computer’s capacity to solve problems whiling conversing with humans as naturally as possible, and learning from it.

For example, you can barely tell the difference between this Google voice assistant and the front desk assistant at this salon. In the “age of assistance” we are demanding more experiences that do not disrupt the lived reality of our lives. The impersonal and restrictive syntax of past and present bots and voice assistants served a purpose as their focus was automating simple tasks like connecting you to your “party’s extension” or gathering basic information to support human representatives of the problem or purpose of your inquiry.

Now, chatbots, voice assistants, and similar technologies are training to reflect the same natural language patterns we use as humans. The goal is to make the technology indistinguishable from humans by being social and user-led, allowing the computer to give feedback to customer queries and inputs.

Choosing a Company

This didn’t take long at all. I picked up my phone and started flipping through it’s screens. Considering the apps that built on search functions, I landed on Groupon. Surprisingly, I found no remnant of the chatbot or voice assistant technology in the app or desktop experience. I liked the idea of starting from scratch so I settled on Groupon as my company.

Setting the Tone

Once I chose Groupon, the next logical step was to find documentation on Groupon’s brand personality and voice. Luckily, being the consumer facing product it is, the website detailed five characteristics that described the “feel” of the brand and it’s commitment to the innovation the product.

I could have stopped here, but I wanted to add more depth to my understanding of the brand’s tone. I decided to make a “this NOT that” statement to define the intention of each characteristic. In addition to the statement I wrote a description clarifying the distance between the characteristics in the “this NOT that” statements.

Fun NOT Childish — All touch points carry a touch of playfulness that resonate with mobilizing the customers ability to connect with great deals that enrich their lives. An energetic and dynamic experience that inspires not drains from “overdoing it.”

Clever NOT Condescending — Take the guesswork out of getting good deals. Think about the things customers may overlook and use subtle cues to guide customer to their goal. Let their ‘aha’ moments intuitive and reinforce Groupon’s cleverness.

Using these statements as guiding principles, I created a container for the conversational UI.

Use Case

Next, I began weaving the story of my Groupon chatbot experience. My use case came first, “I want to find and purchase a deal for entertainment so that I can take my family’s kids out for fun.”

If you wanted a deeper story that’s it — it’s the first thought I came up with so I challenged myself by seeing it through. Having a clear use case became the foundation for the conversation that met my user’s need.

User Persona

Now, it was time to think of who was speaking to the chatbot anyway. Who was the I? With a use case in hand, I created a fictional user persona that gave me the remaining context I needed to start the conversation UI.

Meet Aaliyah.

  • A 35-year-old woman with family in town for holiday at her house. She is new to Atlanta, GA, so she is only willing to go 10–20 miles away from her home. She doesn’t have a strong sense of what the kids in the family like, but promised them a night out in Atlanta so she needs help finding an age-appropriate activity.

Based on my sister, I feel this persona was realistic and representative of a market of Groupon’s current users.

User Flow

How does the user get from their having a need to fulfilling a goal? By utilizing user flows, I was able to think through the conversation as if I was creating flows for a UI design. When we talk about user experience even a lo-fi flow can help define the scope of a particular feature and ensures key steps of the envisioned process are not missed.

Being this was an exercise, I went straight to my goal with minimal consideration to alternatives of my scenario. What if my bot couldn’t answer her question? What if my user didn’t have the answers the bot needed to make a decision?

Making sure you take time for these considerations is key when you develop scripts with real world application because the happy path is rarely the reality. A flow chart can help you plot the happy path and alternatives which together tell a more robust story.

UX Writing — the crux of Conversational UI

Unlike the other sections which follow my actions and thought processes, I’m going to use examples from my script, I share a few key insights to wrap up.

  • Actions Speak Louder than Words

Leverage the tone and personality characteristics in the actions of the UI. We get the most robust characters from good indirect characterization. That’s also true for people, you know — actions speak louder than words.

When thinking about how to bring in the traits from your research consider the “how” your bot does something. How does it ask a question? How does it respond when it’s stumped? How does it respond to an annoyed user? How does it communicate an error?

MailChimp is a good example with it’s quirky copy being reflective of it’s brand personality. The reality is I don’t remember the words at all. I remember the feel from the actions taken that create the experience — like the monkey hi-fiving you after a campaign.

Think about the feeling you want your customers to leave with. This is the most important.

  • Align the Language of Platform Experiences with the Script

Aaliyah: Hello?

Dealbot: Hey Aaliyah, what kinds of deals are you looking for today?

Aaliyah: I need to find something to entertain a group of five kids for 2–3 hours tonight.

A chatbot does not stand alone, it should speak the language of the website and app experience. It’s key for a Groupon chatbot to ask, “what deals are you looking for,” just like Facebook asks “what’s on your mind, AmberNechole?” Become aware of how the entire ecosystem of language that your script exists in and build with it in perspective.

  • Don’t Expect Human Kindness

Dealbot: Would you like me to find similar activities based on relevant past orders?

Aaliyah: Yes.

Dealbot: Sounds good. I’ve sent you a couple options, am I going in the right direction?

Aaliyah: Kinda, the kids I’m with are a mix of teens and pre-teens. All the stuff for young kids should be excluded.

In my first drafts, my user speaks to my bot like a regular person. Upon reflecting on the script, I realized that unless someone is talking to a bot for pure fun, they want to get a job done. I intentionally made her answers short, like ‘yes’ and ‘nope’ to juxtapose the bot’s characteristics.

In creating scripting for conversational UI, remembering the “customer is always right” is a good rule of thumb to design by. Your bot should reflect the best of your brand with an angry customer or a gentle one.

  • Display Your Bot Using Earlier Information

In brainstorming, especially before the data rips you to shreds, it’s good practice to show your bot using earlier information to make a decision. It reflects continuity in your design and understanding of the dynamic nature of chatbots and voice assistants. They are constantly learning how to respond to new questions and using past information to make inferences like you and I.

Dealbot: Paintball Park has a deal for $49 for 12 open-play passes with safety gear and gun rental for 12 people ($420 value). It does have some stipulations, Compressed air and paint not included and a minimum 500 paintball purchase per person required. Additionally, it is not recommended for children under 13 without parental consent. There is no food at the venue. I can help you find a deal if you’d like.

In the above script, my bot recalled an earlier “input” of the user saying this:

Aaliyah: Kinda, the kids I’m with are a mix of teens and pre-teens. All the stuff for young kids should be excluded.

Did you catch what my bot did? Aaliyah said she had a mix of pre-teens and teens. The bot doesn’t know the exact ages, but due to the ‘string’ pre-teen it triggers the response “it is not recommended for children under 13 without parental consent.” Although subtle, I thought it was important to have to reflect it’s ability to make informed decisions about existing deals based on a customer’s information.

  • When You Get Stuck, Switch Directions to Refresh Your Brain.

This one is for you, not the UI. Just like writing a story or article, if you get stuck start on the other end. I think scripting is especially cool to do this with because meeting yourself in the middle can show blatant inconsistencies or the perfect integration of problem and solution. UX writers get writer’s block too, so it’s important to change perspectives and use design-thinking strategies to facilitate your scripting.

Source link


Please enter your comment!
Please enter your name here