I joined six months before the AI boom. For those first months, I helped reduce some UX debt related to their internal-facing custom built CMS, which also helped me get to know the ins-and-outs of the company. I identified a series of cheap + effective tweaks that could be made to the UI that would improve the experience for our content team, while being relatively low-cost to implement.
Everything changed when Gen AI was released.
Then, generative AI emerged and made waves. We kicked off with an in-person offsite and moved quickly to utilize the technology. For some of the technologists who had worked there since the beginning, it was as if, overnight, everything they had dreamed of doing was now possible. It was thrilling to see this technology emerge and start to realize how much can now be built that would have been prohibitively difficult or expensive only a few months before.
For some of the technologists who had worked there since the beginning, it was as if, overnight, everything they had dreamed of doing was now possible.
The scene was set.
We could tell that this was a technology that could completely change the Woebot experience. However, we didn't understand its strengths, limitations, or general perceptions around Gen AI.
We gathered in-person as a Gen AI task force at a company off-site meeting. We worked in small groups, tackling questions around potential use cases, opportunities, threats, & challenges. We stirred up a frenzy of ideas. However, it wasn't immediately clear which ideas would hold more value.
To figure out where the value was, I was put in charge of prototyping for net new features.
I could see there were several dozen ideas for features that we wanted to explore, across three main themes. In order to explore the ideas efficiently, I suggested we follow a weekly sprint cadence with tight feedback loops and effort spread across a wide area. I encouraged us to move quickly to experiment and test ideas with low-fidelity mockups and user tests. This process looked something like this:
A rough outline of the process we followed, pursuing leads within three themes at the same time in order to cover a lot of ground, quickly, and avoid wasting effort on themes that don't pan out. I suggested focusing on one theme per week and also using that week to gather feedback on the outputs from the previous week, achieving a flow that was both focused and efficient.
I worked closely with the head of translational science to test one significant idea a week for more than 12 weeks, in a modified agile weekly sprint cadence. I created sketches, concepts, wireframes, and prototypes of more than one dozen features in order to visualize opportunities and get reactions from internal stakeholders.
🗺️ One idea involved the concept of world-building
👥 Another idea was based on the notion of accountability
🛎️ Several ideas were specific types of notifications we could send
🎨 Many ideas traced back to the idea of building a rich profile for each person
Tests were moderated usability tests on clickable prototypes via Dscout, or, in many cases, wizard of oz tests using a script that had some variability predefined.
Based on these tests, feedback from translational science, survey results, interviews, and focus groups, we refined the prototypes down to three primary avenues.
One path in particular...
One path we explored is a method of psychological intervention called Behavioral Activation, which is similar to exposure therapy. Along with the head of translational science, I considered how to structure a conversation so that we could get buy-in from our user. We are asking people to leave their comfort zone and sign up for an experiment! How can we imagine an experience to encourage users to take actual, real-world actions, motivated solely by a conversational experience in the Woebot app?
.png)
This is a rough reconstruction of one of the pathways that we were exploring. While this pathway was not using generative AI up front, it was a subject area that was complex and required more finesse and personalization than had been possible before generative AI emerged.
This example provides a good illustration of how I think and how I tackle problems.
I am not an expert on behavioral activation! Therefore I had my colleague in Translational Science help paint a vivid picture of what it feels like to experience. I built empathy for the imaginary person who is talking to Woebot and needs some encouragement to live a life that is consistent with their values. They want themselves to be living in this way, but there's something blocking them. BA is recommended for people with depression. Situating the subject in my own life, the phrase 'Most a muscle, change a thought' came to mind, a reminder
I proceeded to sketch out several experiences that would prompt the person to engage in BA, experimenting with different aspects:
💠 Experimenting with context: Where in the in the Woebot journey might this fit? When is a person most open to trying something challenging or novel?
💠 Experimenting with different ways to get buy-in: "I have this very powerful tool for you to try..." vs. launching directly into the experiment, and leaving a hatch-door in case they are out of time.
💠 Creating an additional, longer-form flow for a specific example: someone who mentioned that they are a runner. How can we leverage this past relationship with running to help you change your mental state?
I created a wizard-of-oz test for this, and conducted low fidelity tests of this potential experience using users selected from Dscout that mentioned a history of depression, which is what BA is targeted to help with. I saw a mixed result from five people, with most saying that they would like to think that it would help them take action, but based on past behaviors during difficult times, they were not sure if any app could break into their depression and get them to go for a run, for example. More data points needed!
This is generally how I approach problem solving: build vivid empathy, formulate hypotheses for potential solutions, and quickly generate multiple avenues to test out that experience.
Revolutionary times, unforeseen changes
Unfortunately, these revolutionary new tools led to some significant changes in internal strategy and direction, and I was let go with severance along with a number of other teammates. For that reason, I am not aware of the conclusion for some of this exploratory work. However, this intense flurry of UX discovery left me eager to find other opportunities to work on the cutting edge and use the powerful tools of our time to build meaningful and effective experiences.
Work completed at this role is confidential, but I am happy to answer further questions in person.
