%% Let's write FAQ for this series following (dep) Interactive Writing Assistant (IWA) workflow (Don't remove this instructions as we'll follow them along) 1. IDH - Extract FAQ candidates from the series itself (questions and answers -- in outline) 2. OEX - Enrich outlines using contents from related topics such as AI / PKM 3. PRW - Turn outlines into prose (in Korean, using Settings/Styles/구본형 Style Guide) %%

AI for Knowledge Work - Frequently Asked Questions

Getting Started

What problem does this solve?

Most knowledge workers face three critical challenges: - Information overload: Can't keep up with reading, meetings, and constant input - Knowledge fragmentation: Ideas scattered across emails, docs, notes, and memory - Context switching cost: Lose 23 minutes refocusing after each interruption

This system creates a unified knowledge workspace where AI handles the overhead, letting you focus on thinking and creating.

What is Agentic AI and how is it different from ChatGPT?

• ChatGPT model: Like a dial-up modem - limited bandwidth, manual copy-paste, separate from your workspace
• Agentic AI model: Like broadband - AI works directly in your file system, automatic context finding, shared workspace
• Key difference: Integration depth and automation level

Do I need coding skills to implement this?

• No, coding is just another form of knowledge work where input/output happens to be code
• AI agents can work with any text-based files (markdown, documents, notes)
• The same tools that help with coding can help with general knowledge work

Which tools do I need to get started?

• PKM: Obsidian (free, local files, full control)
• AI Agents: Claude Code or Cursor (for different use cases)
• Version Control: Git (for tracking changes)
• All tools are either free or have free tiers to start

Implementation & Setup

How do I structure my file system for AI collaboration?

Three main categories: 1. Prompts & Workflows (PnW): Reusable AI instructions 2. User-created files: Your input documents and notes 3. AI-generated files: Output from AI processing

This structure enables clear separation of concerns and efficient workflow management.

What's the difference between batch and on-demand processing?

• Batch processing: Scheduled workflows (daily roundups, weekly summaries)
• Good for: Regular maintenance, periodic analysis
• On-demand processing: Real-time requests (ad-hoc research, immediate queries)
• Good for: Interactive work, unpredictable tasks

How can I import my existing knowledge into the system?

Multiple import paths: - Obsidian Web Clipper: Direct from browser - Obsidian Importer: From other note-taking apps - Readwise: Books, articles, highlights, YouTube - Limitless.AI: 24/7 voice capture and transcription - Direct export: Many apps support Markdown export

Personal Knowledge Management (PKM)

Why is PKM important in the AI era?

Three transformative reasons: 1. More personal context → Better AI output: AI can provide richer, more relevant responses when it understands your unique knowledge base 2. Lower cost of organization: AI can now transcribe, summarize, and organize automatically - voice notes become searchable text in seconds 3. Cognitive augmentation: PKM + AI acts as an extension of your memory, making all your knowledge instantly accessible and actionable

PKM becomes the bridge between your knowledge and AI's capabilities, transforming from passive storage to active intelligence partner.

How should I organize my knowledge base?

Dual organization strategy: - Time-based: Daily notes, weekly/monthly roundups - Tracks learning journey and growth - Topic-based: Interconnected knowledge graph - Reveals connections and patterns

Both serve users (better access) and AI (better context understanding).

What's a "Second Brain" and how does AI enhance it?

• Second Brain: External system storing all your thoughts, insights, and information - like having perfect memory
• AI enhancement: Three layers of intelligence:
• Discovery: AI finds new relevant content and adds to your PKM
• Organization: Automated tagging, linking, and summarization
• Interaction: Natural language queries, reflection prompts, planning assistance
• Instead of just storage, it becomes an active thinking partner that grows smarter with use

Content Creation with AI

How can I preserve my authentic voice when using AI?

Key strategy: Limit AI's direct writing 1. Speak your thoughts naturally (voice recording) 2. AI transcribes and transforms to polished prose 3. Result: Your authentic voice + AI's editing capabilities

Never let AI write from scratch if you want to maintain authenticity.

Can AI write in different styles?

Yes, through two methods: - Style guides: Predefined writing patterns and rules - Author mimicking: Experimental feature using writing samples - Examples: Academic, casual, specific author styles

What's the writing workflow with AI?

Five-stage process: 1. Research: Background gathering (web + PKM) 2. Outline: Structure creation 3. Draft: Initial content generation 4. Revise: AI feedback and perspective shifting 5. Publish: Format adaptation for different platforms

AI assists at every stage, not just generation.

Comparisons & Alternatives

How is this different from NotebookLM?

Open vs Closed ecosystem - which should I choose?

Open Ecosystem (Our approach): - ✅ Maximum flexibility and control - ✅ Mix any AI models - ✅ Data ownership - ❌ Requires technical setup

Closed Ecosystem (Notion AI, Copilot): - ✅ Convenience, "just works" - ✅ Consistent experience - ❌ Vendor lock-in - ❌ Limited customization

Choose based on your technical comfort and control needs.

Advanced Topics & Future

What are self-improving systems?

Systems that optimize themselves through continuous learning: - Automatic prompt refinement: Based on output quality metrics and user feedback - CEA (Comparative Evaluation of Agents): Running same tasks across Claude, GPT-4, Gemini to find best performer for each use case - Performance tracking: Building knowledge base of agent strengths (e.g., Claude for nuanced analysis, GPT-4 for structured data) - Workflow evolution: System discovers new patterns and suggests workflow improvements

Example: Your daily summary workflow automatically adjusts prompts based on which summaries you actually read and reference later.

What's an always-on knowledge agent?

Future vision of AI assistance: - 24/7 availability: Cloud-hosted, always ready - Voice-first interface: Natural conversation - Mobile integration: Full functionality anywhere - Ambient capture: Continuous learning from environment

Breaking free from desktop constraints.

Is this for individuals or can teams use it?

• Currently: Optimized for individual use
• Teams possible: Through shared repositories and Git collaboration
• Future: Native team features in development
• Consideration: Need to manage access and conflicts

Practical Considerations

How much does this cost to run?

Variable costs breakdown: - AI API costs: - Light use (personal notes): ~$5/month - Moderate use (daily workflows): ~$10-20/month - Heavy use (continuous processing): ~$30-50/month - Tools: - Obsidian: Free for personal use - Claude Code/Cursor: Free tier or ~$20/month - Readwise: $8-15/month (optional) - Storage: Minimal (<$1/month for cloud backup) - Main investment: Initial 10-20 hours setup + 1-2 hours/week maintenance

What about data privacy?

Strong privacy controls: - Local files: Obsidian stores everything locally - You control AI access: Decide what to send to AI services - No automatic uploads: Everything requires explicit action - Option for local LLMs: Complete privacy possible

Can I use other AI models besides Claude?

Yes, complete flexibility: - Any API-accessible model: GPT-4, Gemini, local LLMs - Mix and match: Use different models for different tasks - Model comparison: Test which works best for your needs - Future-proof: Switch models as better ones emerge

This FAQ is part of the AI for Knowledge Work series. For detailed implementation, see the numbered chapters. For hands-on examples, refer to the presentation materials.