Message Us V2

Amazon, 2019

Reinventing a stagnating platform.

Message Us (MU) is a customer service chatbot platform aimed at decreasing the amount of customer contacts handled by humans. This platform was first launched in Q4 of 2016, after a 6-month research and design effort. The first version of this product was designed and optimized for mobile, as no existing chat functionality was present at the time. Since then, the team has solely focused on bringing more workflows and automation into the platform.

Now, MU faces a new set of challenges. Research done on both the desktop and mobile experiences highlight gaps in regards to customers’ perception of the platforms, effort required, and its limited interactions. With a restricted UI library and interaction patterns, the chat platform is at risk of stagnating, and failing to engage customers. Iterating on the depth and breadth of workflows will achieve less than expected results, unless we improve upon the growth-limiting factors of engagement and delight.

Here's how I began crafting an MU V2.

• Presented December 2019

Design Brief

In order for my managers and stakeholders to allow me the time and space to embark on this effort, I created a design brief that explained my reasoning for this project and defined the scope. This project would focus on how we could perform distinct iterations on the designs in order to make the product more engagement and delightful.

The design brief for MUV2

Discovery

Some upfront work was needed to understand the opportunity space. I interviewed all the MU stakeholders to understand what experience-related problems are causing them issues in their work. I also looked at the research reports from past MU Baseline studies to gather customer painpoints. From these, I then compiled notes and affinitized them into themes.

I then looked at other chatbots from competitors to understand what aspects worked well for them. Again I compiled these findings and affinitized them into themes.

One of the boards where I posted the annotated competitive analysis

SWOT Analysis

I also looked into industry research that explained the value of chatbots, good practices, and design pitfalls. With this, and the findings from previous exercises, I completed a SWOT analysis of chatbots with my team. We came up with this list:

Defining the area of opportunity

With months spent gaining the foundational understanding of the product and chatbot space, I now was able to define exactly what areas we would be exploring in MU V2. With each area, I also defined the driving questions and open questions that would have to be answered through this effort.

The focuses would be on Interactions, Visuals, and Branding.

The full list of exploration areas with driving questions attached

Design Principles

In order to focus our thinking and make decisions on what ideas to pursue in the future, I developed design principles:

• 1. Match the level of humanization to the bot’s capabilities.
  As the bot becomes more capable, it should increasingly emulate the personality and characteristics of a Customer Service Agent.
• 2. Educate in stages.
  Set realistic expectations early, and avoid unintended deception. Describe user benefits, not technology. Describe the core value initially, but introduce new features as they are used.
• 3. Tailor the experience to the customer.
  Understand the emotional state and journey of the customer before entering Message Us, and personalize the experience accordingly.
• 4. Allow the customer to express their voice in the conversation.
  The customer can interact with MU in the way they want to.
• 5. Make the experience memorable through delightful issue resolution.
  We encourage the customer to return to MU by providing delightful and memorable experiences. The customer is never stuck in a dead-end. If we aren’t able to solve the issue, we will direct the customer to a human who can.

Ideation

Now's the fun part. With the broad areas defined, and specific opportunities in each area, I could go create and document my design ideas.

One of the exploration boards that I had filled with ideas

In order to get some additional feedback, we reviewed the design ideas with our Director. We currently have two separate products in our space—one conversational (MU), and one GUI-based (Foresight)—that solve customer issues. Upon seeing my explorations, our Director questioned why we don’t just combine the two products. This was because many of the ideas had visualized how conversational and GUI interfaces could interact seamlessly. While this question wasn’t in scope for our MU V2 project, it was something that we would have to tackle in the following year.

Design Concepts

After we were done exploring we started grouping the ideas and organizing them into 2 concepts:

This concept extended and iterated on the traditional chatbot experience with explorations in voice & tone, explicit bot identity, UI cards to present information, and a more modern visual skin.

Some screens from the Bot-Forward concept

In this concept, I re-imagined what a chatbot experience could be—turning it into a dynamic workspace, rather than an explicit conversation. I explored hiding the explicit bot identity while maintaining a conversational tone & voice, and using more UI inputs to facilitate this conversation.

Some screens from the Controller-Forward concept

Testing

We learned that there were aspects of each concept that customers enjoyed. While making iterative changes to the chatbot bettered the traditional experience, the flexibility of the Controller-Forward concept excited customers more with the ease of access to various types of information.

First-time Onboarding

Informational Cards

Alternative Controls

Examples of the ideas that tested well with customers

Review & Launches

After reviewing these findings with leadership, we decided to proceed in 2 parallel directions. In the short term, we would make the iterative changes suggested by the Bot-Forward approach to the current Message Us platform. These were built in Q1 2020.

• Enhanced item picker. Added search functionality and infinite scroll to the chatbot's item picker. Resulted in 1.03M annualized Customer Service contact reduction ($2.678M annualized savings).
• End chat functionality. Added the ability for a customer to end a chat conversation. Resulted in 715K annualized Customer Service contact reduction ($1.859M annualized savings).
• Typing indicator animation. Added an animation to slow down the speed of the bot messages and to draw customer focus back to the new messages. Resulted in 213K annualized Customer Service contact reduction ($554K annualized savings).
• Visual reskin. Redesigned the visual design of the UI components to make them more modern and engaging. Resulted in 156K annualized Customer Service contact reduction ($406K annualized savings).
• Customer satisfaction survey. Integrated a question from the bot that allowed us to establish a Bot Satisfaction metric.

A video of the Protopie prototype I put together to show our Front-end engineers how the new animations will behave

In the long term, we would invest heavier into exploring and building out the controller-forward concept. I iterated on this concept and performed a more in depth usability study to validate this direction. After that, I created a vision deck and pitched this as our 3-year vision. After receiving alignment and doing an extensive road-show across our Customer Service (CS) organization, this became the current direction CS is driving towards.