Apple ConversationKit

I founded and designed Apple's in-house no-code tool for building AI for sales and support.

00

Background

In 2015, Apple announced internally that they would begin driving customers to interact with customer sales and support (S&S) via iMessage. My team was responsible for building tools that allowed Apple to handle extremely high S&S volumes while delivering world class quality. This announcement meant that volume was about to sky-rocket. My team and I realized we needed automations that could effectively handle incoming messages and, otherwise, route customers to the correct S&S team. The experience needed to surprise and delight and uphold Apple's quality standards.

Solution

Enter ConversationKit: I led the product and design, from 0-1 and from 1-2, of a tool that allowed non-technical service team members to build out conversational interfaces for Apple's S&S globally. Our platform made it easy to train a base LLM with custom intents, collect contextual data from user responses, interact with API's within a conversation and build out rich media messages and interactions across multiple chat channels (starting with iMessage), while integrating seamlessly with the Apple ecosystem. The platform saved Apple millions in customer support costs in the first year and serves hundreds of millions of Apple customers globally eight years later.

What started as a small prototype between my colleague, Matt Austen, and me, became a well funded startup within Apple.

We made tens of iterations and worked across multiple organizations like Apple Messages, Apple Maps, and Apple Marketing to build a tool that seamlessly integrated with the existing ecosystems.

In two long weeks, Matt and I built a functioning prototype. We demo’d our solution alongside IBM Watson's team. Our model accuracy and precision was on par, while the flexibility and usability of our tool met Apple’s needs far more effectively. Apple chose to move forward with our solution.

By the end of my two years on the project, we were a team of 13: one additional full-time designer, a design intern, four full-time engineers, and six part-time engineers dedicated to building a platform that has served as Apple's gateway to S&S for the last eight years.

year

2017

year

2017

year

2017

year

2017

timeframe

2.5 years

timeframe

2.5 years

timeframe

2.5 years

timeframe

2.5 years

tools

Sketch, Framer, Vue.js, Stanford ML

tools

Sketch, Framer, Vue.js, Stanford ML

tools

Sketch, Framer, Vue.js, Stanford ML

tools

Sketch, Framer, Vue.js, Stanford ML

category

UX + AI

category

UX + AI

category

UX + AI

category

UX + AI

Impact

Won VP-level stakeholder adoption Saves Apple hudreds of millions annually Still in use and growing 8 years later

Impact

Won VP-level stakeholder adoption Saves Apple hudreds of millions annually Still in use and growing 8 years later

Impact

Won VP-level stakeholder adoption Saves Apple hudreds of millions annually Still in use and growing 8 years later

Impact

Won VP-level stakeholder adoption Saves Apple hudreds of millions annually Still in use and growing 8 years later

01

I designed many novel UI patterns that were adopted and implemented by the iMessage team.
I designed many novel UI patterns that were adopted and implemented by the iMessage team.
I designed many novel UI patterns that were adopted and implemented by the iMessage team.
I designed many novel UI patterns that were adopted and implemented by the iMessage team.

02

The final UI is confidential, but this is an early wireframe showing the potential for different types of nodes and how logic trees can be nested and collect context.
The final UI is confidential, but this is an early wireframe showing the potential for different types of nodes and how logic trees can be nested and collect context.
The final UI is confidential, but this is an early wireframe showing the potential for different types of nodes and how logic trees can be nested and collect context.
The final UI is confidential, but this is an early wireframe showing the potential for different types of nodes and how logic trees can be nested and collect context.

03

I presented my innovations with Niko Grupen, our Machine Learning expert, at the  Apple Machine Learning Summit in 2017.
I presented my innovations with Niko Grupen, our Machine Learning expert, at the  Apple Machine Learning Summit in 2017.
I presented my innovations with Niko Grupen, our Machine Learning expert, at the  Apple Machine Learning Summit in 2017.
I presented my innovations with Niko Grupen, our Machine Learning expert, at the  Apple Machine Learning Summit in 2017.

.open for collaboration

Shoot me an email about your next adventure

.open for collaboration

Shoot me an email about your next adventure

.open for collaboration

Shoot me an email about your next adventure

.open for collaboration

Shoot me an email about your next adventure