Developing effective agentic AI requires a new research playbook. When systems plan, decide, and act on our behalf, UX moves beyond usability testing into the realm of trust, consent, and accountability. Victor Yocco outlines the research methods needed to design agentic AI systems responsibly. Agentic AI stands ready to transform customer experience and operational efficiency, necessitating a new strategic approach from leadership.
A recent study suggests that 65 percent of our daily behaviours are done on "autopilot," meaning that we do them without thinking. These automatic behaviours occur because they are the result of a habitual process. Habitual behaviours are formed through repetition. They can be helpful, like washing our hands, or unhelpful, like biting our nails. Since so many of our day-to-day actions are habitual, understanding how habits form and how we can change them is essential for improving health and productivity.
It's rooted in old philosophy but runs today on experiments, observation, and data. Back in the late 1800s, Wilhelm Wundt had the bold idea to treat consciousness like something you could study in a lab, the same way you might study a chemical reaction or a falling apple. But people aren't atoms. We don't follow fixed laws. We're emotional and shaped by the world around us. Context, memory, trauma, culture, and love all matter.
A hypothesis is simply an assumption about user behavior, needs, or preferences that you want to validate (or invalidate). Think of them as "bets" you're making in your design. I typically use the following format for a hypothesis: I believe that [user group] will [do something/use something] because [reason or context]. I believe that first-time users will skip account setup if it feels too long, because they want immediate value.