Clay Chatbot

Clay Chatbot

Brief

In the spring of 2016, chatbots were the new frontier. Facebook, Kik, and several others all launched chatbot platforms with the lofty goal of replacing apps. Naturally, we at Clay Games saw this as an opportunity to further expand the reach of our own game platform, by providing users all-new ways to discover and play games. I was tasked with designing the entire Clay chatbot experience, from how it would communicate with users, to determining the best way to optimize games for chat-based interfaces.

My role

As with my previous projects as the sole designer at Clay, I owned the entire design process, from initial ideation through to shipping the live product. In this role, I relied on close collaboration with front-end and backend developers to help problem solve and provide feedback every step of the way.

Process and solution

Chatbots were foreign territory for me, my team, and most everyone else. There were few, if any, guidelines or best-practices early on. That meant I'd be improvising to start, relying on rapid iteration and constant user feedback to get it right. But first, I needed to figure out what exactly our bot's uses should be. After extensive brainstorming with my team, we came up with these use cases:

  • Chat with the bot to get game recommendations based on current mood/interests/situation. Example: Brian has long commutes every day. He asks the bot for word puzzle games to play while on the train
  • Use the bot to play games with friends in a group chat. Example: Susan and her friends are feeling bored, so they ask the bot to start a card game with them

Now with a general idea of what the bot could do, it was time to plan out how it would do it.

I began by creating several flow diagrams for how users would interact with the bot, how the bot would guide them through various games, and how it would behave in group scenarios. These diagrams would go through numerous iterations as we worked to streamline the bot's interactions as much as possible.

Along with the flows, I also began writing rudimentary scripts for the bot. As a chatbot, its primary interface is all text-based, so it was crucial for the bot to communicate in a clear, easy to understand, and friendly way.

As the bot developed further, it became clear that it needed more personality to make it friendlier, more approachable, and less robotic. However, we didn't want to imitate a human, for fear of dipping into the uncanny valley. Our solution? Clay, the Red Panda.

Based on an original sketch by Andrew Augustin, I redrew Clay digitally, building him out of modular components. We would be able to programmatically mix and match individual components to allow Clay to convey all manner of emotions. He could be excited, silly, embarrassed, shocked, angry, sad, and more. This accomplished my goal for Clay to react in a semi-realistic way to any scenario.

His ears, eyes, and mouth were all interchangeable

I'd created a bot with personality—next was to design games for him to play with users. I ended up designing a combination of word and trivia games that could be played directly over chat, plus a few that would launch fullscreen within the messaging app. As a general rule-of-thumb, the games needed to be playable in very short bursts, and asynchronously when in a group.

A few of my games: Doodle Draw, Connect 4, Hoops, and Texas Hold'em

Finally, to round Clay out, we curated a selection of games across a variety of genres from both the App Store and Google Play that he could recommend to players either when prompted or periodically (with their permission, of course).

Example game recommendations Clay could give

Outcome highlights

  • Launch parter & first approved game developer for Facebook Messenger's and Kik's chatbot platforms
  • 2+ million unique users
  • 13+ million messages sent