My AI note-taking workflows  🔗 **Source:** [My AI note-taking workflows](https://nicolevanderhoeven.com/blog/20260529-my-ai-notetaking-workflows/) > [!note]+ Document Notes > tags: [ai-note-taking, speech-to-text, pkm-workflows, voice-capture, obsidian] > author: Nicole van der Hoeven > source: Nicole van der Hoeven > reference: https://nicolevanderhoeven.com/blog/20260529-my-ai-notetaking-workflows/ > date created: {{date:YYYY-MM-DD}} {{time:HH:mm}} > > # 🏷️ Title: My AI note-taking workflows > <!-- deep-read:sections start --> ## 🧭 Table of Contents - [📝 Summary](#-summary) - [📖 Main Content](#-main-content) - [🗒️ Notes](#️-notes) - [📚 References](#-references) - [🗣️ Paraphrase](#️-paraphrase) - [📰 Newsletter](#-newsletter) - [✅ Action Items](#-action-items) - [🧪 Hypotheses](#-hypotheses) - [📊 Methodology](#-methodology) - [❓ Questions](#-questions) - [🔮 Future Research](#-future-research) - [💡 Implications](#-implications) - [🧠 Learning Styles Assessment](#-learning-styles-assessment) - [📊 Accessibility Assessment](#-accessibility-assessment) - [🌍 DEIB Assessment](#-deib-assessment) - [✅ Logical Validity](#-logical-validity) - [🛡️ DARVO Analysis](#-darvo-analysis) - [🎯 Conclusion](#-conclusion) ## 📝 Summary - Nicole van der Hoeven shares her evolution from AI-skeptical to AI-integrated note-taking workflows over the past year. - She presents three tiers of AI voice-to-text systems: free (phone apps), paid (convenience services), and "overkill" (custom hardware + AI agent integration). - Key distinction: AI helps with note-taking (collecting information) while humans focus on note-making (connecting and synthesizing). - Voice capture reduces friction to near-zero, enabling presence in meetings while maintaining comprehensive records. - Privacy and consent considerations are critical, especially in tier 3 setups with always-on recording capabilities. ## 📖 Main Content Nicole frames her approach around Nick Milo's distinction between note-taking (collecting, recording, highlighting) and note-making (connecting, active thinking, linking ideas). Her thesis is that AI should handle the tedious collection work, freeing humans for synthesis and connection-making. ### Tier 1 — Free Solutions Basic phone recording apps with built-in transcription: iOS Voice Memos, Samsung Voice Recorder, and desktop tools like Spokenly. The workflow requires manual steps: remember to record, transcribe, and transfer to your PKM system. This tier serves as a no-cost entry point for voice capture workflows. ### Tier 2 — Paid Convenience Services like Voicenotes, Otter.ai, Letterly, and Granola collapse friction by auto-transcribing and often providing AI summaries or Q&A capabilities. Nicole demonstrates using Granola with Obsidian integration via QuickAdd macros to automatically link transcripts to relevant vault notes. The tradeoff is vendor lock-in and potential data loss from service shutdowns. ### Tier 3 — Custom Integration Nicole's personal setup combines a Plaud AI hardware recorder with her custom AI agent "Iris" powered by OpenClaw. The system runs continuously or on-demand, syncs to Obsidian via a forked plugin, and generates daily summaries, task extraction, and contextual reminders. "Hey Iris" voice memos allow natural language task delegation that processes automatically. The key insight is parallel usage across all three tiers for different contexts: Spokenly for dictation, Granola for meetings, Voicenotes for journaling, and Plaud for travel conversations (with explicit consent). ## 🗒️ Notes - **Friction elimination**: Voice becomes the lowest-friction capture method - no device finding, app opening, or typing required - **Presence optimization**: Knowing transcription is handled automatically allows full engagement in conversations and meetings - **Note-taking vs note-making distinction**: AI handles the mechanical collection work, humans focus on synthesis and connection-making - **Privacy considerations**: Always-on recording requires careful ethical consideration and explicit consent from others - **System redundancy**: Using multiple tools in parallel optimizes for different contexts rather than seeking one universal solution - **Integration complexity**: Custom workflows require technical setup (BRAT plugins, QuickAdd macros, AI agent configuration) but offer personalized automation ## 📚 References - [The Mechanics of Knowledge Management podcast episode](https://nicolevanderhoeven.com) - [OpenClaw agent runtime](https://nicolevanderhoeven.com) - [Voicenotes app](https://nicolevanderhoeven.com) - [Author's Patreon with QuickAdd scripts](https://nicolevanderhoeven.com) ## 🗣️ Paraphrase Nicole van der Hoeven describes her transformation from AI skepticism to integration over the past year, particularly for converting speech to text in her personal knowledge management system. She presents three implementation tiers: free phone apps with manual workflows, paid services that reduce friction through automatic transcription and AI features, and a custom "overkill" setup using dedicated hardware and AI agents. Her core framework distinguishes between note-taking (mechanical collection) and note-making (synthesis and connection). AI excels at the former, allowing humans to focus on the higher-value latter. Voice input eliminates traditional capture friction - no device handling or typing required. Her personal system uses a Plaud recorder with custom Obsidian integration and an AI agent called Iris that processes daily transcripts, generates summaries, and handles voice-delegated tasks. She emphasizes using multiple tools in parallel for different contexts rather than seeking a universal solution, while noting the critical importance of privacy and consent when using always-on recording capabilities. ## 📰 Newsletter **Voice-First Note-Taking: From Skeptic to Convert in 12 Months** Nicole van der Hoeven's journey from AI skeptic to voice-capture enthusiast offers WordPress users practical insights for content creation workflows. Her three-tier approach - from free phone apps to custom AI integration - shows how voice input can eliminate the friction between ideas and documentation. The key insight for content creators: AI handles the mechanical work (transcription, initial organization) while you focus on synthesis and connection-making. Whether you're recording client calls, brainstorming sessions, or content ideas during walks, these workflows could transform how you capture and process information for your WordPress sites. Check out her Obsidian integration techniques and consider how voice-to-text might streamline your content planning process. The privacy considerations she raises are particularly relevant for agencies working with client conversations. ## ✅ Action Items #### DevRel - Evaluate voice-to-text tools for developer advocacy content creation: demos, talks, and documentation drafts - Test Granola or similar tools for recording and transcribing community calls, office hours, and workshops - Develop guidelines for consent and privacy when recording developer conversations or feedback sessions #### Community - Explore voice-based contribution workflows for accessibility: verbal issue reports, spoken documentation reviews - Create community guidelines for recording meetups, WordCamps, and contributor sessions with proper consent protocols - Pilot voice-to-text for meeting notes in contributor team calls #### WordPress - **Core**: Test voice workflows for bug report dictation and code review commentary - **Documentation**: Experiment with voice-first content creation for DevHub and HelpHub articles - **Training**: Explore voice-based lesson plan creation and course content development - **Accessibility**: Evaluate voice input as an alternative interaction method for wp-admin ## 🧪 Hypotheses - Voice capture reduces the cognitive overhead of note-taking, allowing for more presence and engagement in conversations - AI-assisted transcription quality has reached a threshold where manual correction overhead is minimal for most use cases - Multi-tier tool usage (rather than single-solution approaches) optimizes for different contexts and friction levels - The gap between thought and documentation can approach zero with proper voice-to-AI workflows - Always-on recording capabilities create productivity gains that justify the privacy and consent complexity - Integration between voice tools and PKM systems (like Obsidian) creates compounding value over standalone transcription ## 📊 Methodology - **Tier comparison study**: Test free vs. paid vs. custom solutions across different recording contexts (meetings, walks, desk work) measuring accuracy, friction, and integration ease - **Friction measurement**: Time-track the complete workflow from thought to searchable note for voice vs. traditional text input methods - **Privacy impact analysis**: Survey participants in recorded conversations about comfort levels and consent preferences across different tools and contexts - **Synthesis quality assessment**: Compare note-making outcomes when AI handles transcription vs. manual note-taking during the same conversations - **Integration effectiveness**: Measure how often voice-captured notes get linked, referenced, or acted upon compared to traditionally captured notes ## ❓ Questions - How do voice-to-text accuracy rates vary across different accents, languages, and technical vocabularies in WordPress contexts? - What consent protocols work best for recording community calls while maintaining psychological safety? - Which integration patterns between voice tools and WordPress editorial workflows show the highest adoption rates? - How does voice-first content creation affect the quality and structure of final published articles? - What privacy-preserving alternatives exist for the benefits of always-on recording? ## 🔮 Future Research - Local-first voice processing capabilities and their privacy advantages over cloud-based solutions - Multi-language voice workflows for global WordPress communities - Integration patterns between voice capture and WordPress editorial workflows - Accessibility implications of voice-first content creation tools - Long-term knowledge retention differences between voice-captured and manually typed notes ## 💡 Implications - **Practice**: Voice-to-text workflows could significantly reduce barriers to documentation in open source projects - **Policy**: Communities need clear guidelines for recording and consent in distributed, global contexts - **Security**: Always-on recording devices create new attack vectors and data retention considerations - **Education**: Voice-based learning and content creation methods may improve accessibility for different learning styles - **Funding**: Investment in local-first AI tools could reduce dependency on centralized transcription services ## 🧠 Learning Styles Assessment Reading level estimate: Flesch-Kincaid Grade 10.2 - accessible to general tech audience #### 👀 Visual learners - Present: Three-tier comparison structure, tool screenshots, workflow diagrams - Improve: Add visual workflow flowcharts, before/after interface comparisons, decision tree for tool selection #### 🔊 Aural learners - Present: Voice-focused content naturally appeals to auditory processing, includes video walkthrough - Improve: Add audio version of the article, podcast-style explanation of each tier #### 🗣️ Verbal learners - Present: Clear definitions (note-taking vs. note-making), analogies about friction reduction - Improve: Include more technical terminology explanations, comparison frameworks #### 🧩 Kinesthetic learners - Present: Step-by-step setup instructions, hands-on tool demonstrations - Improve: Add practice exercises, downloadable setup checklists, trial periods for each tier #### 🧠 Logical learners - Present: Structured tier progression, clear tradeoffs analysis, decision frameworks - Improve: Add cost-benefit matrices, privacy risk assessments, tool comparison tables #### 🧑‍🤝‍🧑 Social learners - Present: Community context, consent considerations, shared workflows - Improve: Discussion prompts about team adoption, collaborative note-taking scenarios #### 🧘‍♀️ Solitary learners - Present: Personal workflow optimization, individual tool selection criteria - Improve: Self-assessment questionnaires, reflection prompts on current note-taking pain points #### 🌿 Naturalistic learners - Present: Organic workflow evolution, adaptation to different contexts - Improve: Environmental considerations of always-on devices, sustainable tool choices ## 📊 Accessibility Assessment #### Screen reader semantics - Present: Clear heading hierarchy, descriptive text for workflow steps - Improve: Add ARIA landmarks for tier sections, alt text for any embedded media #### Keyboard navigation - Present: Text-based content navigable without mouse - Improve: Ensure embedded videos have keyboard controls, skip links for tier sections #### Contrast and readability - Present: Good text contrast on white background - Improve: Check code block contrast, ensure link visibility #### Text resizing and reflow - Present: Responsive design supports text scaling - Improve: Test at 200% zoom, ensure no content overlap #### Media and captions - Present: Video content referenced - Improve: Ensure video has captions, provide transcript downloads #### Forms and interactions - Present: Simple navigation, no complex interactions - Improve: Any embedded tools should have accessible alternatives #### Standards reference - Meets WCAG 2.2 AA for text content; video accessibility needs verification ## 🌍 DEIB Assessment #### Cultural breadth - Present: Focus on English-language tools, Western productivity frameworks - Improve: Include multilingual transcription options, global privacy law considerations #### Inclusive language - Present: Gender-neutral technical descriptions, avoids exclusionary jargon - Improve: Acknowledge economic barriers to paid tools, include alternative approaches #### Neurodiversity support - Present: Multiple input methods accommodate different processing styles - Improve: Discuss cognitive load benefits, executive function support aspects #### Representation - Present: Single author perspective limits viewpoint diversity - Improve: Include community examples, diverse use cases beyond knowledge work #### Psychological safety - Present: Strong emphasis on consent and privacy considerations - Improve: Address anxiety around recording, power dynamics in workplace contexts ## ✅ Logical Validity #### Strong reasoning - **Causal reasoning**: "Voice capture reduces friction → enables presence in meetings" supported by specific workflow examples - **Analogical reasoning**: Note-taking vs. note-making distinction provides clear framework for AI role allocation - **Inductive reasoning**: Three-tier progression from simple to complex shows scaling adoption path #### Fallacies - **Appeal to novelty**: Assumes newer AI tools are inherently better without comparing to refined traditional methods - **False dilemma**: Presents voice vs. text as binary choice rather than complementary methods - **Hasty generalization**: Personal workflow success may not transfer to different work contexts or personalities #### Tone of voice - Conversational and practical with examples: "The slightly absurd thing is that I still use all three levels" - Balanced acknowledgment of limitations: "Cons: the privacy of people around me" - Technical without being intimidating: clear explanations of complex integrations ## 🛡️ DARVO Analysis No DARVO patterns detected. The author presents a straightforward workflow explanation with appropriate acknowledgment of limitations and privacy concerns. The emphasis on consent and ethical considerations demonstrates responsible approach to potentially invasive technologies. ## 🎯 Conclusion Nicole van der Hoeven's three-tier voice-to-text framework offers a practical roadmap for integrating AI into note-taking workflows. Her key insight - using AI for mechanical collection while humans focus on synthesis - provides a sustainable approach to AI adoption that enhances rather than replaces human cognitive work. For WordPress professionals, the workflows suggest immediate applications in content creation, client communication documentation, and community contribution processes. The emphasis on privacy and consent considerations is particularly relevant for distributed teams and public communities. The multi-tool approach (using different solutions for different contexts) offers a more nuanced alternative to seeking single universal solutions, allowing optimization for specific use cases while maintaining workflow flexibility. ## 🔄 Refactor Appendix — My AI note-taking workflows ### 📚 Literature Atoms (H3) #### Voice input eliminates traditional capture friction barriers > [!quote] > For me, voice is the lowest-friction capture method that exists. No stopping, no finding my phone, no opening an app, no typing. Just talk — and the system handles everything downstream. **Context & use**: - Core argument for voice-first workflows - Contrasts with traditional text input methods requiring device interaction **Tags**: #ai-automation #workflow-optimization #voice-capture **Highlight**: Primary workflow philosophy <!-- atomicity: 1 idea ✓ --> #### AI should handle collection while humans focus on synthesis > [!quote] > Note-taking is collecting — recording information, paraphrasing what an author said, highlighting. Note-making is connecting — active thinking, engaging with the information, linking it to other ideas, working out how it fits into your PKM system. **Context & use**: - Nick Milo's distinction adopted as AI workflow framework - Defines clear role separation between AI and human cognitive work **Tags**: #ai-automation #knowledge-management #human-ai-collaboration **Highlight**: Foundational framework concept <!-- atomicity: 1 idea ✓ --> #### Multi-tier tool usage optimizes for context over universality > [!quote] > The slightly absurd thing is that I still use all three levels — in parallel, for different things: Spokenly for dictating at my desk, sometimes straight into Obsidian, sometimes into a chat. Granola for meeting notes, and increasingly for transcribing videos I'm watching. **Context & use**: - Challenges single-solution approaches to tool selection - Demonstrates context-dependent optimization strategy **Tags**: #tool-selection #workflow-optimization #context-switching **Highlight**: Practical implementation insight <!-- atomicity: 1 idea ✓ --> ### 🧠 Permanent Candidates (H4) #### Voice-first workflows reduce cognitive overhead for knowledge capture **Claim**: When voice input eliminates mechanical friction, cognitive resources can focus entirely on synthesis and connection-making rather than documentation mechanics. **Evidence pointer**: [[#Voice input eliminates traditional capture friction barriers]] **Scope & conditions**: Effectiveness depends on transcription accuracy, integration quality, and user comfort with voice input **Links**: - [[Friction reduction]] — because extends beyond traditional productivity contexts - [[Human-AI collaboration]] — because applies-in knowledge work specifically - [[Context switching costs]] — because contrasts by eliminating input method transitions <!-- atomicity: 1 idea ✓ --> ### ✏️ Fleeting Queue - [F1] Investigate local-first voice processing for privacy-preserving workflows — https://nicolevanderhoeven.com/blog/20260529-my-ai-notetaking-workflows/ #fleeting #privacy - [F2] Test Granola + Obsidian integration for WordPress community call documentation — workflow evaluation #fleeting #community ### → Things 3 (max 2 actions) - [ ] Evaluate voice-to-text accuracy for technical WordPress vocabulary — testing framework #things - [ ] Draft consent protocols for recording contributor team calls — community guidelines #things <!-- deep-read:sections end --> ## Highlights - For me, voice is the lowest-friction capture method that exists. No stopping, no finding my phone, no opening an app, no typing. Just talk — and the system handles everything downstream. - Note: 📚 Voice input eliminates traditional capture friction barriers · - Core argument for voice-first workflows - Contrasts with traditional text input methods requiring device interaction · #atom/literature/h1 - Note-taking is collecting — recording information, paraphrasing what an author said, highlighting. Note-making is connecting — active thinking, engaging with the information, linking it to other ideas, working out how it fits into your PKM system. - Note: 📚 AI should handle collection while humans focus on synthesis · - Nick Milo's distinction adopted as AI workflow framework - Defines clear role separation between AI and human cognitive work · #atom/literature/h2 - The slightly absurd thing is that I still use all three levels — in parallel, for different things: Spokenly for dictating at my desk, sometimes straight into Obsidian, sometimes into a chat. Granola for meeting notes, and increasingly for transcribing videos I'm watching. - Note: 📚 Multi-tier tool usage optimizes for context over universality · - Challenges single-solution approaches to tool selection - Demonstrates context-dependent optimization strategy · #atom/literature/h3 - Voice-first workflows reduce cognitive overhead for knowledge capture - Note: [[Friction reduction]] — because extends beyond traditional productivity contexts · - [[Friction reduction]] — because extends beyond traditional productivity contexts · #atom/zettel/z1 - Investigate local-first voice processing for privacy-preserving workflows — https://nicolevanderhoeven.com/blog/20260529-my-ai-notetaking-workflows/ #fleeting #privacy - Note: ✏️ Fleeting capture #atom/fleeting/f1 - Test Granola + Obsidian integration for WordPress community call documentation — workflow evaluation #fleeting #community - Note: ✏️ Fleeting capture #atom/fleeting/f2 - Presents voice vs. text as binary choice rather than complementary methods - Note: 🔍 Fallacy: False Dilemma #fallacy/false-dilemma - Personal workflow success may not transfer to different work contexts or personalities - Note: 🔍 Fallacy: Hasty Generalization #fallacy/hasty-generalization