Conversational chatbot

McDonald's

Building a chat-based cashier that delivers service with a smile.

DURATION

4 Weeks

TEAM

Pete Wise
Sara Ma

My Contributions

User Research
Conversational Design
Coding & Integrating AI
User Testing

TOOLS

Voiceflow

THE CHALLENGE

Creating Space for Conversation in a Fast-food Restaurant

In fast-food, speed is king, and the opportunity to build connections at the counter gives way to efficiency. In my conversational UX class, my team built a chatbot that creates space for customers to connect with the McDonald's brand while ordering.

TAKEAWAYS

Learning as I Built

While building my first chat agent, I had to learn new terminology, design process, and software as I built.

TAKEAWAY #1

Test Utterances Early and Often

Building a chatbot that feels human requires a rich dataset and a deep understanding of how customers want to interact with it. Testing early and often generates both the knowledge and the data.

DEFINING INTENTS

Misplaced Focus on Placing Orders

While conducting initial research, we extracted intents from the McDonald's mobile app and read social media threads on people's go to McDonald's orders. Our research approach inadvertently limited the scope of our understanding to users that already know what they want to order.

show the user flow going through the McDonald's app to order food with major tasks extracted to be used as intents in building the chatbot

WRITING UTTERANCES

Too Specific

When translating user intents to utterances, we assumed people were mainly familiar with the McDonald’s menu. Early version of utterances focused on requesting specific menu items.

❌ I want a McFlurry

❌ I'll have a Big Mac

âś… I need to wake up

TESTING

Create a Collaborative Conversation

During our first round of user testing, we learned that people wanted to converse with the agent to help make decisions. We rewrote our user utterances to be more open to vague phrases that the lead the conversation to suggestions. Then, we continually tested these utterances to build our phrase bank (right).

Example Chat with Suggestions

I'm craving something sweet
Something sweet does sound good right now! Which one of these look good to you?
I could use a cookie!
Ooo chocolate chip cookies are one of my favorites!
How many do you want right now? You can order 1 ($1.25), 3 ($2.75) or a dozen ($8)

Utterance Phrase Bank

a bank of utterances in voiceflow

TAKEAWAY #2

Stay True to Brand Values

The chat agent should embody the beloved McDonald's brand. From brand guidelines and analyzing the tone used in the McDonald's app, I crafted the agent's personality to be unpretentious, lighthearted, and approachable.

Unpretentious

Stay casual but use clear diction. Mcdonald's is loved across cultures, so inclusion is a priority.

Nice! Our fries are iconic.

Lighthearted

Playful language to affirms choices and create positive, memorable interactions.

Ooo chocolate chip is my favorite cookie!

Approachable

Always be ready to help and never judge. Offer suggestions to help guide the customer.

How can I help you today? Check out deals. See the menu.

TAKEAWAY #3

Don’t Let Technical Skill Limit Design

At the start of the project, Pete and I got caught up in figuring out how to customize each order item and replicating the menu, we ended up building our chat to only accept one item at a time instead of how people naturally want to order food.

BUILDING VUI

Adding One Item at a Time

We mirrored the functionality of the McDonald’s app, where users add one item to the cart at a time. This works well for users who decide as they browse, but not for those who already know their order.

showing the voice user interface flow of ordering a single item at a time built on voiceflow

VOICEFLOW PROTOTYPE

From Menu to Check Out

Our final design retained the flow of adding one item at a time.

FUTURE CONDSIDERATIONS

Ordering Multiple Items Matches Mental Models

Repeat McDonald's customers prefer to order multiple items at once, so if we had more time, we would implement that functionality in the future. I learned for conversational UX projects to use the mobile app to extract intents but don't try to mimic the functionality of the mobile app.

Example ordering multple items

I want a chicken nuggets, a McDouble with no cheese and coke
That sounds good! I'm going to add nuggets, a McDouble with no cheese and a coke to the bag
How many nuggets do you want?
You can get 4 ($3.59), 6($5.19 ), 10($6.49), 20 ($9.99), or 40 ($17.99)
Let's do 6
Great! I'll edit the nuggets to be 6
For the coke, is medium for $1.49 okay?
Yeah, that's good
Okay! Your total is $10.77 with a 6-piece chicken nuggets, McDouble with no cheeese, and medium coke. Do you want to add anything else?