Defining a future vision of Watson Assistant that reduces the time and effort to build a chatbot

My role

Lead Designer, collaborating with product management, engineering, and design research

Timeline

6 months

Outcome highlights

The final concept presentation and validation research findings convinced executive leadership to prioritize its development and release by reworking the upcoming product roadmap and accommodating our resourcing needs
When presented with the concept, one prominent existing customer responded by incorporating the solution into their roadmap and signed up to be a beta user
I enabled a newly formed squad – composed of engineering and product management members who hadn't been on a squad with design before – to quickly arrive at a place where they were working with and contributing to design in a collaborative, trusting style

Skill highlights

Design leadership - at the squad level, got the team on board up front with the importance of user outcomes and user research; and at the executive level, influenced the stakeholders to shift their mindset about the potential of this project
Collaboration - Cultivated a collaborative environment for our squad to feel empowered to contribute user experience ideas and be a part of the design process, feeling a sense of ownership in the conceptual storyboard we created
Innovation - my design ideas created a solution that offers capabilities that no other chatbot platform on the market does, while utilizing new tech from IBM’s innovation labs

The challenge

Our team was responsible for designing and selling to internal stakeholders a vision to deeply and intuitively integrate Watson’s AI search tool (Watson Discovery) into their existing chatbot product (Watson Assistant) in a way that would differentiate Assistant in the conversational AI market.

The task was to plan out an innovative integration concept, which would eventually enable a non-technical customer care professional to easily build a chatbot solution that could pull responses directly from their company’s existing content. This new search-enabled chatbot could then respond to customers’ problems with accurate, conversational answers.

Adding to this challenge, the design process and product vision needed to be developed with a newly formed cross-functional squad composed of several teammates who had never worked with design before.

Design process

Perform generative research

First, we needed to develop a deeper understanding of our users in order to find opportunities for solutions and innovation. I partnered with design research to establish a generative research study. We worked with them to define research objectives, a screener and interview discussion guide.

The research findings we got from talking to 8 customer care professionals unearthed key insights and opportunities. These findings informed all of the subsequent design work we did.

Set goals

With our user data in tow, I set up a series of collaborative design sessions with my squad so we could ideate as a group, utilizing all of our unique perspectives. We kicked off the series by brainstorming hopes and fears, the problems needed to be solved, and overall goals for the project.

We turned the results of this exercise into a strategy document that we presented to leadership stakeholders in order to gain alignment before moved forward.

Create scenario map

Next, we were ready to draft a rough vision of what we saw as the ideal experience for our user.

In this session, we worked together to build out a high level to-be scenario map that laid out the journey of our persona’s intents and the actions taken to complete her goal. Within the map, we then identified 9 "problem spaces" that we wanted to dig into deeper.

Ideate in problem spaces

I led a 1 hour session for each problem space. The squad diverged and sketched ideas of ways to solve this particular problem for the user. We then converged and discussed the ideas a group.

After each session, we summarized the ideas we wanted to advance into a flow. By the end of the sessions, we had a skeleton flow that would serve as the base of our concept car.

Produce concept car

We took the ideas from the visioning sessions and boiled it down to something more realistic, and that told a story. It came to life in the form of a concept car (a design technique borrowed from the auto industry) that tells the story of how we can solve a customer’s problem, without getting hung up on technical constraints. With each iteration of the concept car, we shared with stakeholders and evaluated the concepts with users.

Evaluate concept

Now that we had our concept car, we needed to see if the concept resonated with existing and prospective users. We paired with design research to define the goals and discussion guide for the evaluation research. We talked to 8 participants and learned that overall our idea was novel and valuable. We were also able to identify some areas that needed further refinement.

Results and next steps

Based on the results of the evaluation research, our design vision was successful – our concept would garner interest in the market and reduce setup time for customer care professionals building a chat bot.

We also achieved alignment with the Assistant and Discovery leadership and were green-lit to move forward with production.

The concept car artifact empowered the engineering member to produce a notional architecture map so that they could start producing baseline technology. The concept car and architecture map were used to determine the staffing needed to bring the vision to life.