Online discussions often fail in two ways. Asynchronous discussions feel impersonal. Synchronous discussions don’t scale. In a large, fully online graduate program enrolling thousands of students, I designed a discussion workflow that preserved live interaction without increasing instructional overhead. This approach led to: - **Stronger peer connection**, as reported by students compared to prior cohorts - **No increase in instructor grading time**, despite longer and more substantive discussions - **Longer and deeper live conversations**, no longer constrained by grading feasibility These outcomes made synchronous discussions viable at scale in fully online courses and led to **broader adoption** among instructors teaching similar courses. ### The Problem Asynchronous discussion boards are easy to manage, but students rarely care about them. Text-based participation often feels performative, and with AI, surface-level responses can be generated in seconds. Live discussions improve engagement, but they introduce a different constraint. To grade participation, instructors have to attend sessions or review recordings. At scale, this quickly becomes impractical. **The core tension was clear:** _How do you preserve the value of live interaction in online learning without making it costly to sustain?_ ### Constraints - Instructors could not realistically attend or watch recordings of all discussions. - The solution should not introduce a new non-content related learning curve for students. - Student submissions had to be reviewable quickly and consistently. - Tools had to be university-approved and compliant with student privacy requirements. ### Design Decisions I designed a discussion workflow that allowed students to participate live while enabling instructors to evaluate participation asynchronously. Instead of requiring instructor presence, full video submissions, or manual student summaries, discussions were captured directly using AI-powered transcription and summarization tools. Students submitted either: - a transcript of the session, or - an AI-generated summary highlighting key points and individual contributions. This provided instructors with a concise, searchable artifact. Instructors reviewed these submissions against a discussion rubric to assess contribution quality and engagement. Before Zoom AI Companion was available, I redesigned discussion guidelines to help students enable and use Otter.ai during virtual meetings. Students submitted transcripts and summaries generated by the tool. When Zoom AI Companion became available, I initially created setup guides for students. Later, I partnered with the university’s Zoom team to automatically enable these settings for student accounts in programs using this workflow. Compared to Otter.ai, Zoom AI Companion offered structured summaries, speaker attribution, and the ability to jump directly to relevant moments in the video when needed. This shift removed earlier constraints. Previously, discussions were capped at 10–15 minutes to keep grading manageable. With summaries in place, students could engage for longer periods without increasing grading workload. Over time, more students began continuing conversations outside required sessions, and instructors retained visibility into student participation without monitoring every conversation in detail.