Running large-scale online courses exposes a structural tension: as enrollment grows, routine work balloons. Designers and instructors spend more time on administrative tasks, while the quality and intentionality of learning design erodes. With AI and automation, I transformed course design and delivery workflows to reduce manual effort without removing human judgment, freeing people to focus on what matters most: teaching and learning. This approach led to: - **30% faster course deployment**, by automating high-friction setup tasks - **20% improvement in design efficiency**, freeing time for higher-order design work - **Reduced duplication and rework**, across media, assessments, and course assets - **More instructor time spent on teaching**, rather than administration ### The Problem Online programs scale through standardization, but that efficiency carries hidden costs. Tasks like group creation, media management, and content scaffolding expand linearly with enrollment. At scale, even simple actions consume weeks of collective effort. **How do you reduce operational drag in online learning without lowering quality or removing human oversight?** ### Constraints - Courses enrolled hundreds to thousands of learners - Core platforms (Canvas, Slack, media repositories) could not be replaced - Designers and instructors needed to retain editorial control - Any solution had to work across programs, not as a one-off optimization These constraints ruled out platform overhauls or fully automated course generation. The focus shifted to automating _repetition_, not _decision-making_. ### Design Decisions #### 1. Automating high-friction setup tasks Course launches were slowed by manual group creation across multiple tools. In large courses, setting up Canvas groups and corresponding Slack channels could take days. I initially reduced friction using structured Google Sheets to generate consistent group and roster data. In parallel, I worked with Canvas and Slack vendors to automate group and channel creation directly. This eliminated manual setup. Work that once took hours or days now completed in minutes, without instructor involvement. #### 2. Making media assets searchable and reusable Over time, the design team accumulated a large image library that was effectively unsearchable. Assets lacked alt text, tags, or descriptive metadata, forcing designers to recreate visuals they already had. I built a system that processed each image through AI to generate descriptive alt text, searchable tags, and ready-to-use embed code. The result was a searchable, accessible media library that supported reuse and faster design decisions. Designers stopped rebuilding assets simply because they couldn’t find them. #### 3. Reducing repetition in course design artifacts Course development repeatedly required the same foundational elements: learning objectives, rubrics, quizzes, discussion prompts, and accessibility metadata. Designers rebuilt these components course after course, often aligning them manually to external standards. I developed a suite of AI-assisted tools to generate first-pass versions of these elements. Designers reviewed and refined outputs, but no longer started from a blank page. By removing repetitive setup and generation tasks, designers and instructors regained time for judgment-driven work: sequencing, feedback, interaction, and instructional intent. AI functioned as infrastructure rather than pedagogy. The learning experience remained human-designed, but no longer constrained by manual overhead.