Prepare to Design the Future
March 19–20, 2017: Training
March 20–22, 2017: Tutorials & Conference
San Francisco, CA

Bots may solve some of our problems; here's how they'll put us on the hook for others

Desiree Zamora Garcia (IBM Watson)
2:05pm–2:45pm Wednesday, March 22, 2017
Beyond the screen
Location: California East
Level: Intermediate
Average rating: **...
(2.50, 2 ratings)

Who is this presentation for?

  • User experience designers, product designers, product managers, user researchers, and backend engineers

Prerequisite knowledge

  • Familiarity with automated support tools, customer service software, and enterprise software
  • Experience working with teams that include multiple engineers (including backend)
  • Comfort designing with APIs

What you'll learn

  • Explore lo-fi prototyping of a bot
  • Understand why the most important thing about building a bot is designing with content first
  • See why the harder the job, the longer the training a bot needs


We know bots are everywhere. Chances are, you’ve interacted with pleasant, simple bot—or perhaps you’ve interacted with a less impressive one (sometimes cute, sometimes nerve-racking).

Have you ever wondered what it would be like to interact with a supercharged bot—one that could detect your emotions so it could respond back with reassurance, one that built a personality profile on you, down to the hobbies and interests you had, so it could make small talk?

Desiree Garcia shares how she messed up in designing a bot using IBM Watson. This isn’t meant to be a sensational story—it won’t repackage what happened with Microsoft’s Tay, for example. Instead, it is a story about the importance of design process, designing with content first, and the impact of constraints.

It started off as a simple bot to help Nike+ customers troubleshoot their products. A hasty, last-minute decision to impress a stakeholder led the team to switch out the Nike+ domain for Twitter’s main support line. From a pure business standpoint, the bot was a customer service tool, and it was still working with the same type of content. But what the team didn’t think about was how much the context had changed and how crucial it is to train a bot before even demoing it. The type of user that asks for help for a fitness device is in a much different situation than one who is asking for help because they’re being harassed or stalked online. The type of content is still 140 characters long, but even the most advanced bot today will find it difficult to read between the lines of a tweet.

Ultimately, the project was released as a sample app—a resource for developers to learn how to build their own bot with Watson. While the bot won’t actually be responding to customer service concerns on Twitter, it is still a point of view. Without context in this case, precedent is set for what’s acceptable in bot communication. Moreover, it may even validate the type of rhetoric that sends victims of online harassment to seek help in the first place.

When designing for “the future,” sticking to our principles as designers will be even more important. Bots will continue to solve many of our problems. But they will put product teams even more on the hook for others.

Photo of Desiree Zamora Garcia

Desiree Zamora Garcia

IBM Watson

Desiree Garcia is a designer for IBM Watson. She works on Watson Conversation, a product that enables people who are not developers to build, train, and scale conversational interfaces. Before, she worked on the Watson Developer Cloud, helping developers to leverage specific domains of cognitive technology in their software. She is interested in the notion of diverse teams setting precedent with “emerging AI,” a term she uses to describe a range of current technology that will shape what mature AI looks like someday. Her official background is in psychology research and practice, but she has been tinkering with computers and the web for much longer. Ending up on the Watson design team was unexpected, but makes pretty good sense.