CASE STUDY | Smart (Voice) Shopping

Jetson AI

ON VOICE AND TEXT

 

ABOUT

Jetson AI is a “voice-first” AI platform that allows restaurants to offer conversational food ordering through a virtual assistant, both on and off-premise, and be easily discovered by new customers.

PROBLEM

Small business owners and brands need to expand their market reach, increase sales, and reduce the service bottleneck by prioritizing convenient, safe, and pandemic-friendly shopping.

SOLUTION

Jetson AI agent offers multi-step, cognitive conversation while helping customers find personalized location-based restaurants and decreases wait times during breakfast, lunch, and dinner rushes through consistent, pleasant experiences augmented into personal devices or channels.

ROLE & RESPONSIBILITIES

My job was to design conversational and voice interfaces for a mobile app guiding customers through the ordering process and replicating an in-store experience.

01 | Discovery (Research)

To kickstart the assignment and understand why users needed a voice assistant to order from a quick service restaurant, who uses them, the volume of interaction, and their pains and gains, I led a series of customer and stakeholder interviews.

I discovered that professionals working in co-working spaces use voice assistants for the following reasons:

  1. To save screen time and cognitive load searching and selecting a store.

  2. To expedite an order placement at the nearby restaurants

  3. To cut delivery fees.

02 | Competitive Analysis

What’s out there? Siri, Amazon Alexa, and Google Assistant just to name a few, but… the voice applications of today suffer from an inherent design problem.

Digital voice assistants and IoT devices offer IVR experience with one-sentence responses, that generate a “robotic” feel without natural conversational continuity.

Why conversational AI?

For millenniums people have been buying and selling through conversation.

Consumers are increasingly more concerned about screen time spent using devices.

Fewer apps to download through with conversational commerce.

03 | Value Proposition

Voice-first conversational ordering solution across today’s most popular devices, channels, and ecosystems is the most innovative approach for businesses and their customer’s experience, seamlessly improving employee engagement, enhancing communication efficiencies, and elevating user experiences.

04 | Conversation Design

The main objective of the conversation was to create a quick order flow through continuous human-assistant interaction mimicking the natural dialogue.

Features and capabilities included filtering, location preferences, and integration with Stripe for payment flow.

The platform matching the technical requirements, customizations, integrations, and pricing was IBM Watson Assistant, allowing the development of intents, entities, dialogs, error messages, and other conversation design elements.

I started my design process by creating high-level diagrams and dialogues in Luicidchart to map out the industry-specific conversational strategy and communicate with the product owner, visualize flows, make edits, and encourage cross-functional collaboration.

Order Flow for QSR Restaurants

Before diving into IBM Watson, I reviewed 7 intent pathways for clarity, concision, and relevance to the customer’s goals. I mapped the utterances to visualize the experience and created a system to communicate dialogue revisions to the team.

Payment Flow for QSR Restaurants

Once the dialogue flows and dialogs were reviewed and approved by the stakeholders and development team I created scripts incorporating marketing strategy, for the product to meet user and business goals.

Every dialog is designed with the help of agent responses that maintain and manage the direction of the conversation flow. Responses vary from simple, interrogative, conditional, and complex.

Agent Response Design

05 | Prototype (Mobile App)

While designing across multiple channels and platforms, the MVP was to create a native app experience where end users can order by exchanging messages with the Jetson AI agent with the following features:

Instant access to intelligent browsing and a lower cognitive load on users

Can order directly from the app and a wide range of IoT devices

Displays lists of menu items (entities) making the brand content availability clear

Concise dialogue leads to faster and safer payment transactions

After several rounds of user testing and collecting feedback for visual, conversational, and voice experiences I’ve applied user data to complete the ordering flow design that was clear, concise, and engaging.

Jetson gives users the opportunity to begin their journey regardless if they are on the go, at home, or at the office, so that when they want to order from a nearby restaurant, they can do so instantly.

ERROR HANDLING (CONVERSATION REPAIR)

An important element of designing better experiences and unlocking the potential of conversational AI is error handling. I used the following techniques for when there is a no-input and when there is a no-match:

Disambiguation
Jetson helps users navigate through the ordering experience while letting them know what they can or can’t order.

Suggestions & Discoverability Jetson may not be able to make the right API call without knowing the type of content the user is looking for.

07 | Summary (Key Takeaways)

Designing for Voice. Without visual cues such as nods, facial expressions, and eye contact, words hold even more weight. I was given a thorough list of training phrases, but the NLU wouldn’t always get it right when testing. This is a great accessibility feature that can be explored further but was out of scope for the sprint.
Conversational Design. Continuous dialogue gives the user a feeling of control over the correct solution and alleviates some pain points such as:

  1. Making conversation more efficient and concise minimizing screen time.

  2. Lowered cognitive load on a user expediting the food ordering process.

Challenges. Given the nature of the conversational technology, the phrases were not always captured and understood by Jetson assistant and I had to apply different methods to smooth the interaction:

  1. Eliminating words that did not move the conversation forward.

  2. Giving prompts to guide users to the phrases the assistant could understand.

08 | Next Steps

  1. Create a multimodal conversational experience for smart devices.

  2. Revisit the dialogue. The rhythm and balance of the back-and-forth exchange are as crucial as the writing of the scripts

Being “conversational” is an interaction, not a monologue. The feel should be inviting, encourage cooperation, and be user-centered. A conversational assistant shall always provide an improved, more efficient experience.

UX/UI| Merchant Dashboard

Another task of the project was to design a business-to-business dashboard that allows instant product upload via Delivery.com and Shopify Integrations and a freemium B2C customer-facing app.

Listing in minutes: The easy-to-use dashboard allows sellers to upload products or import your entire product catalog in a snap.

Getting paid seamlessly: a sale is followed by instant Stripe payment directly to a business bank account.

Ship, pick-up, or deliver: allowing customers to pick up purchases in-store or have them delivered. Shopify and Delivery.com integrations enabled.

Next steps: more integrations with Brandibble and Chowly, gamification, and animation of voice commands to maximize user engagement.