iFind, an AI-assisted video editor, was created to aid intervention methods in the FIND Program. It was meant to speed up the process of converting parent-child interaction sessions into coaching videos that highlight program-specific, developmentally supportive techniques, but adoption of the software was less than 10% due to user confusion and a plethora of usability concerns.
Users repeatedly expressed being uncomfortable moving forward with the task because they wanted to evaluate the source of the provided AI clip suggestions.
Grant users autonomy in reviewing the serve and return instances identified by AI, with playback of each instance and manual override for clip selection if desired.
The first 6 users all expressed confusion toward the stated goal. They were unsure how to identify the specific developmentally supportive technique because of nomenclature inconsistencies in the introductory presentation.
None of the 20+ participants afterward had conceptual understanding issues.
The combination of identified usability violations and uncovered mental model mismatches led to iFind being adopted and regularly used by 100% of the FIND research team after 5 weeks of testing.
This work was completed in late 2023 and early 2024, early in my UXR career. Today, I would prioritize setting KPIs to measure impact quantitatively, along with more concrete measures of intuition in usability studies, such as first-click analysis, task completion rate, and time to complete the task. This prioritization would have made it easier to measure the effect of each change in user experience across iterations, rather than solely observing the overall increase in adoption, even though complete adoption was the initial goal.
Want to learn more? Feel free to reach out!