Business problem
Turning interview transcripts into a complete profile requires repeated extraction, prompting, review, and formatting.
Previous manual process
- Read a full interview transcript and pull out the relevant profile details.
- Prompt an AI tool separately for each profile section.
- Copy, review, and format the output into a final document.
- Track whether source data and generated content have been approved.
What I built
Three connected n8n workflows. The first extracts structured data from Fathom transcript PDFs into Google Sheets and requests human review in Slack. The second generates separate profile sections from editable prompt templates only after the data is approved. The third copies a Google Docs master template, replaces its placeholders with approved content, and returns the final document link.
Workflow steps
- Detect transcript PDF
- Extract structured profile data
- Pause for data approval
- Generate sections from editable prompts
- Pause for content approval
- Create the Google Docs deliverable
Apps connected
My responsibility
Workflow architecture, structured extraction, prompt-template setup, approval gates, status tracking, document generation, notifications, and project documentation.
Safeguards
- Human approval
- Structured AI output
- Status gates
- Editable prompts
Removes repeated transcript review, prompting, copying, and document formatting while keeping both the source data and final content under human review.