Transforming course content into validated, reusable, multi-format assets.
💡 The idea: Structure course content once, so designers receive clear, reusable building blocks tailored to their medium, whether a marketing document, a lead magnet, or a video script.
📐 Prototype: Website course pages serve as the source of truth. Content is extracted and structured into predefined document schemas, then transformed into three document variants, including a lead magnet PDF, with tone and format review suggestions generated per version.
Without access to internal workflows, tooling, or friction points, I began by mapping potential touchpoints for designers within the content lifecycle.
The initial goal was simple: understand where designers interact with course material and where friction might occur.
This led to broader questions:
While mapping these designer touchpoints, it became clear that design does not exist in isolation. The preparation of course content upstream strongly influences how efficient and controlled the design phase can be.
What began as a designer-focused investigation expanded into understanding the full course creation lifecycle. Supporting designers effectively required mapping how content moves from subject matter experts and compliance review through marketing preparation and into design execution.
This shift clarified that friction in design often originates upstream, in how content is structured and prepared before it reaches visual production.
Click any step to see Vision, Prototype, and Example.
The Content Engine prototype (GitHub: content-engine) implements this with a Python extractor, Next.js API routes, OpenAI for generation and validation, and Playwright for PDF export.
Structured pull of facts and sections from source—no generative guesswork. In the prototype: a Python (FastAPI) service fetches the course page and returns structured JSON; no LLM in this step.
Narrative generation from extracted data, constrained by templates. In the prototype: POST /api/write calls OpenAI with extracted data and template type (e.g. lead-magnet), returns markdown and layout spec.
Compliance checks and consistency rules before export. In the prototype: POST /api/guardian uses OpenAI to review the draft (claims, tone, suggestions); human review still required before final use.
Design and marketing review before finalisation. The UI shows extracted data, generated draft, and guardian report so reviewers can approve or adjust before export.
Structured output for design tools and version control. In the prototype: a fixed HTML template is filled with generated markdown and rendered to PDF via Playwright; the same model could drive other formats later.
Impact is measured by improvements in preparation time, structural consistency, validation accuracy, and designer workflow efficiency.
Reduced Preparation Friction
Structure course content once so teams spend less time reorganising before design begins.
Measured by: time-to-first-structured draft and reduction in manual restructuring.
Increased Structural Consistency
Ensure every output follows the same predictable content model across labels and formats.
Measured by: template adherence rate and reduction in missing required components.
Reduced Repetition Across Formats
Eliminate repeated reinterpretation and prompt rewriting for each new medium.
Measured by: prompt reuse rate and reduction in duplicated content preparation steps.
Improved Validation & Trust
Strengthen confidence in generated drafts through structured validation before human review.
Measured by: guardian pass rate and percentage of drafts marked "ready for refinement."
Designer Workflow Efficiency
Enable designers to focus on refinement rather than reconstructing content structure.
Measured by: time spent on formatting vs visual design and number of revision cycles.
The current prototype demonstrates a single output format, but its purpose is architectural. It shows how course content can be extracted, structured, validated, and prepared as a controlled foundation for generation.
The broader vision is to standardise course content into reusable schemas that can feed different content creation flows, such as whitepapers, video scripts, voiceovers, sales materials, or brochures.
The goal is not to automate one format, but to structure the content layer that all formats depend on.
Designers and marketers remain in control. The system prepares structured drafts and consistent context, while human expertise governs tone, compliance, and final execution.