Artificial intelligence matters, too. Many companies have begun to focus more on AI, considering it to be the future of technology.
Consequently, the media has erupted with articles on design and articles about AI, but there has been less attention on the intersection of design and AI. It’s not for lack of effort — teams worldwide are designing artificial intelligence. Companies have created a dazzling array of robots — ranging from Boston Dynamics’ Big Dog to Tekno the robotic puppy. Chatbots have begun to replace customer service agents, and you can even hire AI-powered virtual assistants who will manage your schedule and send emails for you.
But what if you want to build AI, yet you don’t know where to start? You’re likely not alone. There are 18.5 million developers around the globe, but only 18,000 of them have expertise in artificial intelligence or machine learning. At Bonsai, we are building a platform that provides developers with tools to create AI — while abstracting away the low-level details required to build artificial intelligence from scratch.
We learned early in our design process that despite the advances made in design and technology, the user experience of developer tools is often overlooked. Accordingly, when we began creating Bonsai’s platform, we strove to design it with developers in mind: a better-designed tool allows developers to work more quickly and efficiently. In order to do so, we conducted several different types of user research with developers and data scientists interested in AI. This article will cover some of the research methods that we used, and we’ll follow up with a second article discussing our findings and how we addressed them in our designs.
Primarily, user experience research seeks to answer questions and prove (or disprove) hypotheses about users. Researchers uncover and articulate users’ needs and goals, the contexts in which they live and work, and their mental models. Once initial research is conducted, product teams can design with a deeper understanding of their user base, and they can validate their assumptions by testing designs frequently.
The Nielsen Norman Group defines these stages as “Strategize,” “Execute,” and “Assess,” respectively.
During the foundational stage, researchers learn about the problem space and potential users with the goal of choosing a design direction(s) or even new product ideas. Methods might include contextual inquiry — observing target user groups in their natural environment — surveys, and target user interviews, among other activities.
Once a direction has been determined, the product team moves into formative research. At this stage, the team will generate many design ideas and capture user feedback to evaluate their approach. For instance, the team will sketch and brainstorm together — and sometimes invite target user groups to brainstorm with them. Once they have fleshed out a design, they might then conduct usability testing to obtain feedback.
Lastly, once a product is released, the team may conduct benchmarking studies, surveys, or other similar types of research to measure the product’s success over time. Results guide subsequent product strategy and design decisions.
As we create our product, we want to ensure that our design decisions are firmly rooted in user expectations. Not everyone on our product team is a software engineer, so we can only imagine how developers would describe their ideal AI tool. Moreover, as non-engineers, we do not know what engineers value (and don’t value!) about the developer tools they are currently using. On a similar note, developers have a variety of interests, expectations, and working styles, so we want to be cognizant of different mental models as we design.
Collecting, analyzing, and applying user insights also helps us to avoid our own biases in the design process. For instance, those of us who are engineers by training have favorite developer tools. However, what our team prefers might not be what the greater developer population likes, and so research will help us check our assumptions.
At Bonsai, we begin our UX research process by collecting a list of research questions. For instance, we might want to learn more about a particular user group’s motivations, or we might want to understand if users perceive the value of a given product feature. We transform these questions into a set of hypotheses, and then we select methods to test them.
Research is especially important in the initial stages of product development. Before we began designing our product, we needed to learn more about our users. We knew that we were creating a product for engineers and data scientists with an interest in artificial intelligence, but who are these people?
We sought to answer these questions — and more — through a series of interviews with developers and data scientists. Over the course of several months, we spoke with dozens of individuals in our office, out in the field, and at meet-ups and conferences.
We used the information we gathered to inform initial designs. We learned which tools the developers we spoke with found frustrating, and we discovered what they considered a good onboarding experience. In turn, the feedback we obtained guided our preliminary wireframes and flows. For example, developers described the types of documentation they found most helpful, so we designed a help and documentation section with their suggestions in mind.
Once we created preliminary designs, we conducted user feedback sessions to evaluate them. During these sessions, we held one-on-one conversations with developers and data scientists (“participants”) during which they reviewed and interpreted our design concepts. As a result, we learned where our designs were successful and where there were opportunities for improvement. Speaking directly with participants also gave us the opportunity to discuss issues and brainstorm solutions alongside end users. The insights and suggestions our participants produced drove our next round of designs.
We also used our findings from our initial research to develop personas. A persona summarizes a type of user, depicted as a real person but is actually created by synthesizing data from user interviews or other research activities. Our persona guides contain basic demographic information, the persona’s occupation, a description of their personality and working style, goals (related to AI and otherwise), and AI expertise, among other information. We use personas throughout the product development process to guide our design decisions. For instance, we imagine how each persona might approach a given problem and what their needs and expectations might be. Personas also help us build empathy for our end users and educate the entire company — not just the product team — about our target audiences.
As we continue to design our product, we are looking for data scientists and software engineers to provide feedback! Your ideas and suggestions will help us craft a better product. If you are interested in participating in a concept feedback or usability session, please email email@example.com. Bonsai will compensate you for your time.
Stay tuned for Part 2, which will discuss our findings and how we applied them.