An executive coach built an AI content pipeline that keeps her voice in charge

MindValley: 2026 AI Mastery Program Submitted Artifact Title: Agentic-Human Content Creation Workflow   Capstone Final Project Category The Optimization Alchemist Award Description Effortless yet impactful workflow enhancements with simple AI tools like Custom GPTs- turning low-hanging AI fruits into opportunities for growth. Date Submitted Aug 8, 2026 Submitted By Dr. LJ (Linda Johnson) ljohnsondba@gmail.com PROJECT DESCRIPTION This project serves as a working demonstration of Stage One in the front half of a content development pipeline used by LJ Executive Coaching’s (LJEC) for producing YouTube and social media deliverables. The three-stage workflow (WF) results in a longform YouTube video first and then repurposes the transcript for social media deliverables across Instagram, LinkedIn, and a website blog. Claude.ai serves as Chief of Staff (COS) and Chief of Operations (COO) and functions via SKILLs initiated by Dr. LJ. While the WF design could have been automated end-to-end, a strategic decision was made to keep the human in the loop ·      During Stage One, the COS conversationally interviews Dr. LJ to surface the weekly content domain focus and bring strategic clarity to expectations and directions for developing the content further down the pipeline. Dr. LJ and the COS solidify the video topic, theme, scope, and source material primarily with the help of Perplexity, Google Gemini, YouTube, and the LJEC Self-Improving Agentic Wiki.  ·      During Stage Two, the COS is responsible for drafting a publish-ready script for Dr. LJ to review, edit, and film. ·      During Stage Three, the COO orchestrates social media production with Marblism Agents Sony and Penny. Dr. LJ completes the process with final edits and review with Legal Linda prior to publishing.   3 Attachments ·      Script-BlindSpot-Whale-Trust-Communication-DrLJ.docx ·      3-Frame Story Arc-Whale Mane-DrLJ.png ·      WIKI NOTE-Grok AI Whale Language-DrLJ.png   ACHIEVEMENTS ·      Ran a live topic-selection interview start to finish, landing on a locked topic and Hook in one conversation ·      Captured that interview pattern into a reusable skill, "echolocator-interview," rather than writing one cold from theory ·      Drafted a full Hook-Proof-Reflection-CTA script (approx. 4 min runtime) against Brand Command Deck voice and structure rules, revised once against LJ's live edits ·      Used Obsidian's web clipper to ingest the source YouTube research video directly into mv-wiki with structured properties (source, author, published date, tags), demonstrating low-effort, high-value ingestion ·      Generated a 3-frame concept storyboard externally (ChatGPT) as a visual proof-of-concept, without committing to full video production ·      Surfaced and enforced a classified-material safeguard mid-conversation: proprietary CreatingWE®/C-IQ training material was correctly held back by default until explicit permission was given   TOOLS •       MindValley AI Mastery 2026 lessons •       Claude.ai, Code •       ChatGPT •       Obsidian, Obsidian Clipper •       Kepano's obsidian-skills (obsidian-markdown, obsidian-bases, JSON Canvas, Obsidian-CLI, Defuddle) •       YouTube •       Grok AI •       NotebookLM (NBLM) •       Google Gemini Mac •       Mini M4 (local build environment) •       MacBook M2 •       Dual monitors   METHODOLOGY-STAGE ONE 1.    COS (Claude) interviews Dr. LJ conversationally to surface the week's topic; open prompt, no menu of options offered 2.    COS separates the "felt" hook detail from the "useful" strategic detail through follow-up questioning 3.    COS checks for supporting research material; classified-material default (no-touch, ask first) applied live 4.    Wiki material selection and formal ingestion deliberately deferred until the project package is complete, to keep the wiki's "official" filing clean 5.    COS drafts script against confirmed brand voice/structure standards 6.    Dr. LJ reviews and edits directly in the working document 7.    Interview pattern extracted into a packaged, reusable skill 8.    Concept visual generated externally; package assembled: script, skill, screenshots, images stopping short of filming   LEARNINGS ·      The emotionally "felt" detail and the strategically "useful" detail are often different; asking twice, differently, is what surfaces the real hook ·      A skill built from a live, real conversation captures nuance (like when to ask permission) that a cold-authored skill would likely miss ·      Classified-material boundaries need to be explicit; repeatable check needed to be built into this skill, assumed knowledge ”in the moment” mattered here, was named out loud, and fixed ·      Low-effort ingestion tools (Obsidian's web clipper) is time-saving extension that works well with the custom note templates in this wiki ·      Proof of connection use case - Checking the wiki for a specific term ("low road") in the C-IQ™ context, and finding it absent, confirmed a voice decision made earlier: that language stays attributed to Creativity Centered Coaching®, not C-IQ™, since reviewed C-IQ™ source material was reviewed, searched, and found it does not include the language. The ability to fact-check within the wiki has also proved to be a significant time saver.   RELEVANT URLS ·      https://drljcoaching.com ·      Source video: [Grok AI Finally Decoded Whale Language-The First Message Shocked Scientists]- https://www.youtube.com/watch?v=C1WdSpe39js