AI Integration and Automation

AI can dramatically accelerate documentation while maintaining quality. This guide shows how to integrate AI tools into your documentation workflow.

Why Use AI for Documentation?

Benefits:

• Speed: 10x faster photo descriptions, summaries, analysis
• Consistency: AI maintains style and format guidelines
• Completeness: AI catches details you might miss
• Analysis: Extract insights from complex information
• Scaling: Handle large volumes of documentation

When AI Helps Most:

• Photo descriptions and analysis
• Meeting note summaries
• Product research and specification gathering
• Decision analysis
• Template generation
• Content organization

AI Tools for Documentation

Claude Code (Primary Tool)

What it is: AI assistant with file access and code execution capabilities

Perfect for:

• Creating and editing documents
• Analyzing photos and generating descriptions
• Researching products and specifications
• Organizing and restructuring content
• Fixing errors and maintaining consistency

Example from Garage Project:

• Fixed “Cratherm” → “Creatherm” spelling across 11 files
• Created comprehensive Products Used page with manufacturer links
• Generated this documentation system guide
• Added inline product links throughout documents

ChatGPT / Claude (Web)

Perfect for:

• Quick queries and brainstorming
• Summarizing complex topics
• Analyzing decisions and trade-offs
• Research assistance

Garage Project Use:

• HVAC system analysis
• Insulation strategy research
• Contractor question preparation

Specialized Tools

For specific tasks:

• Whisper AI: Transcribe meeting audio
• GPT-4 Vision: Analyze construction photos
• Perplexity: Research products and specifications

Setting Up Claude Code

Installation

1. Install Claude Code from https://claude.com/code
2. Authenticate with Anthropic account
3. Configure for your vault

Configuring for Your Vault

Create CLAUDE.md in your vault root:

# CLAUDE.md
 
This file provides guidance to Claude Code when working with this repository.
 
## Project Overview
 
This is an Obsidian vault for managing a [YOUR PROJECT TYPE] project.
The vault contains structured documentation following a phase-based organization.
 
## Repository Structure
 
- **10-Planning/**: Planning phase documents (timeline, budget, decisions)
- **20-Design/**: Design specifications and technical planning
- **30-Vendors & Contacts/**: Vendor information, orders, and products
- **40-Meetings/**: Chronological meeting notes
- **50-Build/**: Construction/execution phase documentation
- **60-After Build/**: Post-completion items
- **90-Reference/**: Reference materials
- **Templates/**: Document templates
- **pictures/**: Photo documentation
 
## Document Types
 
Documents use YAML frontmatter with standardized fields:
- `type: meeting` - Meeting notes
- `type: vendor` - Vendor/contact information
- `type: order` - Purchase orders
- `type: design` - Design specifications
- `type: build` - Construction documentation
- `type: reference` - Reference materials
 
## Naming Conventions
 
- **Meetings**: `YYYY-MM-DD - Type - Subject.md`
- **Decisions**: `Decisions - [Topic].md` or `Decisions Log.md`
- **Orders**: `[Product/Service] Order.md`
- **Photos**: Match image filename (`PXL_20251022_190752216.md`)
 
## Common Operations
 
### Adding Product Documentation
1. Research product specifications
2. Add to `30-Vendors & Contacts/Products Used.md`
3. Include manufacturer links
4. Link from related documents
 
### Creating Meeting Notes
1. Use `Templates/Meeting.md`
2. Name: `YYYY-MM-DD - Meeting - Subject.md`
3. Location: `40-Meetings/`
4. Link to related documents
 
### Photo Documentation
1. Create markdown file matching photo name
2. Include description, significance, technical details
3. Link to related documents
 
## Best Practices
 
- Use wikilinks: `[[Document Name]]`
- Maintain frontmatter consistency
- Cross-reference related documents
- Update Timeline and Decisions Log
- Keep Products Used current
 
## Project-Specific Notes
 
[Add any project-specific conventions or important details]

Why this matters: Claude Code reads this file to understand your vault structure and conventions, providing context-aware assistance.

AI-Assisted Workflows

Photo Documentation with AI

Workflow:

1. Take photo and transfer to pictures/ folder

2. Use Claude Code to analyze:

@pictures/PXL_20251022_190752216.jpg

Please analyze this photo and create a detailed documentation file.
Use the pattern from pictures/PXL_20251022_174100432.md as a template.

Context: This is from garage construction, foundation phase.
The photo shows [brief description of what you see].

3. Claude Code generates:

---
title: [AI-Generated Title]
type: photo-documentation
tags: [relevant, tags]
date: 2025-10-22
time: [extracted from filename if possible]
phase: foundation
---
 
# [Descriptive Title]
 
![Photo](PXL_20251022_190752216.jpg)
 
## Photo Details
- **Date**: October 22, 2025
- **Phase**: Foundation
- **View**: [AI describes viewpoint]
 
## Description
[AI-generated description of visible elements]
 
## Visible Elements
- [Detailed breakdown]
 
## Significance
[AI explains importance/context]
 
## Related Documents
- [[50-Build/Initial Build]]

4. Review and enhance with:
   • Specific technical details
   • Personal observations
   • Decision context
   • Cross-references

Real Example from Garage Project:

Request:

Please analyze pictures/PXL_20251022_190752216.jpg and document it.
This shows our pressure test gauge for the radiant heating system.

Claude Code produced detailed documentation including:

• Gauge reading (60-70 PSI)
• Significance of pressure testing
• Why this checkpoint is critical
• Related to quality control
• Links to contract, HVAC strategy, build progress

Time saved: 20-30 minutes per photo

Product Research and Documentation

Workflow:

1. Identify product from contract/invoice:

Product: Creatherm radiant floor panels
Spec: 2" thickness, 110 sheets

2. Ask Claude Code to research:

Please research Creatherm radiant floor panels and add comprehensive
documentation to 30-Vendors & Contacts/Products Used.md

Include:
- Manufacturer information
- Technical specifications (R-value, dimensions, materials)
- Product links to manufacturer pages
- Installation requirements
- Relevant use cases

Then add inline links in Hershberger's Contract.md where "Creatherm"
is mentioned.

3. Claude Code:

   • Web searches for product information
   • Finds manufacturer spec sheets
   • Extracts key specifications
   • Locates product pages
   • Creates formatted entry
   • Adds inline links in existing docs

4. You review for accuracy and completeness

Real Example:

Task: Document Genie 6170 jackshaft garage door opener

Result:
- Complete specifications (850 lb capacity, 24V DC motor)
- Features (Aladdin Connect, battery backup)
- Warranty information
- Product links to manufacturer
- Added inline links in contract document

Time saved: 1-2 hours of research

Meeting Note Enhancement

Workflow:

1. Take rough notes during meeting (minimal)
2. After meeting, ask Claude Code:

Please enhance these meeting notes following our template:

Raw notes:
- Met with Marcus
- Foundation forms look good
- PEX going in tomorrow
- Discussed slab sensor idea
- Decided to install conduit
- Pour Friday if pressure test passes

Context:
- Meeting date: 2025-10-22
- Attendees: Me, Marcus (Hershberger's), concrete crew
- Topic: Foundation progress and PEX installation

3. Claude Code produces:

   • Properly formatted meeting document
   • Expanded discussions with context
   • Structured action items
   • Clear decisions section
   • Suggested related documents

4. You add specific details and verify accuracy

Benefit: Focus on conversation, not note-taking

Spell-Check and Consistency

Use Case: Fix terminology across entire vault

Example from Garage Project:

I noticed "Creatherm" is misspelled as "Cratherm" in many documents.
Can we correct this across all files?

Claude Code:

1. Searched all files for misspelling (found 11 files)
2. Created task list to track progress
3. Fixed each occurrence systematically
4. Verified all changes
5. Committed with descriptive message

Time saved: 30+ minutes of manual search-and-replace

Decision Analysis

When facing complex decisions:

I need to decide between three HVAC options for my garage.
Here's the context:
[paste details]

Can you help analyze the trade-offs and create a decision document
following the pattern in 10-Planning/Decisions - Slab Sensor Conduit.md?

Claude Code provides:

• Structured analysis of each option
• Pros and cons comparison
• Cost-benefit analysis
• Recommended approach with rationale
• Decision document template filled out

You provide: Final judgment and project-specific context

Content Organization

Use Case: Reorganizing or cleaning up vault

I have several documents about HVAC scattered across folders.
Can you help organize them better?

Files:
- HVAC notes in 10-Planning/
- Mini-split info in emails
- Radiant heat in multiple places

Claude Code:

• Analyzes current organization
• Suggests improved structure
• Moves files with automatic link updates
• Creates index/overview document
• Updates cross-references

Best Practices for AI Assistance

Do’s

✅ Provide context: The more context, the better the output

Good: "Analyze this photo of our radiant floor pressure test during
foundation phase. The gauge should read 60-70 PSI."

Bad: "Describe this image"

✅ Reference examples: Point to existing documents as templates

"Create a vendor document for ABC Roofing following the pattern
in 30-Vendors & Contacts/Services/HVAC Contractors.md"

✅ Review all AI output: AI can make mistakes, always verify

• Technical specifications
• Dates and numbers
• Links and cross-references
• Formatting

✅ Iterate: Ask for revisions

"Good start, but please add more detail about the installation requirements
and include links to manufacturer installation guides"

✅ Use for research: AI excels at finding and summarizing information

"Research oxygen barrier PEX tubing specifications and explain why
it's important for radiant floor heating"

Don’ts

❌ Don’t blindly accept: Always review and verify

❌ Don’t skip human judgment: AI provides analysis, you make decisions

❌ Don’t expose sensitive info: Be careful with:

• Personal addresses (use placeholders)
• Phone numbers
• Financial details
• Passwords or access codes

❌ Don’t expect perfection: AI is a tool, not a replacement

❌ Don’t forget to maintain: Update CLAUDE.md as your vault evolves

Advanced AI Workflows

Batch Operations

Process multiple photos at once:

I have 10 photos of the foundation work in pictures/ folder from
Oct 20-22. Can you create documentation files for each, following
our established pattern?

Photos focus on:
- Excavation and forms
- Creatherm installation
- PEX loop placement
- Pressure testing

Claude Code: Processes all photos, creates consistent documentation

Template Generation

Create new templates:

Can you create a template for tracking contractor performance?

Should include:
- Contact information
- Services provided
- Quote history
- Performance ratings (responsiveness, quality, pricing)
- Would use again? (yes/no)
- Notes

Result: Well-structured, reusable template

Content Summaries

Summarize complex topics:

Please create a summary of all HVAC-related decisions we've made.
Search through meeting notes and decision documents.

Format as timeline with dates, decisions, and rationale.

Use for: Project reviews, stakeholder updates, lessons learned

Link Maintenance

Find and fix broken links:

Can you check for any broken internal links in the vault?
Look for:
- Links to non-existent documents
- Links to moved files
- Incorrect link formatting

Claude Code: Scans vault, reports issues, suggests fixes

Claude Code Example Sessions

Example 1: Adding Product Documentation

User:

Can we create a comprehensive document that lists all products used in
the garage construction? Include manufacturer links and specifications.

Save in 30-Vendors & Contacts/Products Used.md

Claude Code:

1. Searched through contract, invoices, design docs
2. Identified all specified products
3. Researched each product online
4. Found manufacturer specifications and links
5. Created organized document by category
6. Added cross-references to related documents

Outcome: Comprehensive Products Used page with 14 products documented

Example 2: Fixing Spelling Errors

User:

I noticed "Creatherm" radiant floor panels is misspelled as "Cratherm"
in many documents. Can we correct this?

Claude Code:

1. Searched vault for “Cratherm” (found 11 files)
2. Created task list to track corrections
3. Fixed each file systematically
4. Updated in:
   • Main documents
   • Photo descriptions
   • Timeline
   • Contract
5. Verified all changes
6. Committed with descriptive message

Outcome: Consistent terminology across entire vault

Example 3: Photo Documentation

User:

@pictures/PXL_20251022_190752216.jpg

Analyze this photo and create detailed documentation.
This shows the pressure gauge during our radiant heat system test.

Claude Code:

1. Analyzed image (pressure gauge reading 60-70 PSI)
2. Generated comprehensive description:
   • What’s visible (gauge, fittings, background)
   • Significance (quality checkpoint before concrete)
   • Technical context (pressure test requirements)
   • Construction timeline position
3. Added related document links
4. Formatted following established pattern

Outcome: Professional photo documentation in minutes

Measuring AI Impact

Track time savings:

• Photo documentation: 20-30 min → 2-3 min
• Product research: 60-120 min → 10-15 min
• Meeting note cleanup: 30 min → 5 min
• Spell-checking vault: 45 min → 5 min

For Garage Project:

• 50+ photos documented = 15-20 hours saved
• 14 products researched = 10-15 hours saved
• Numerous edits and improvements = 5-10 hours saved
• Total: ~30-45 hours saved

Security and Privacy

What to Share with AI

✅ Safe to share:

• Technical specifications
• Product information
• Construction details
• Meeting summaries (generic)
• Design decisions

⚠️ Be careful with:

• Exact addresses
• Financial details (budget totals okay, bank accounts no)
• Contractor proposals with pricing
• Personal phone numbers
• Email addresses

❌ Never share:

• Passwords
• Access codes
• Credit card numbers
• SSNs or tax IDs
• Legal documents (without redaction)

Using CLAUDE.md Safely

Redact sensitive information:

**Address**: [PROJECT ADDRESS]
**Budget**: $XX,XXX
**Owner**: [YOUR NAME]

Troubleshooting AI Assistance

AI Misunderstands Context

Problem: Generated content doesn’t fit project

Solution: Provide more specific context

This is for a residential garage construction project, not commercial.
The systems are sized for a 960 sq ft building, not larger structures.

AI Output Formatting Issues

Problem: Markdown doesn’t render correctly

Solution: Ask for corrections

The formatting is broken. Please fix the markdown syntax, especially
the frontmatter YAML and image embeds.

AI Includes Wrong Information

Problem: Specifications don’t match actual products

Solution: Provide correct details

The Creatherm panels are 2" thick, not 1". Please correct throughout
the document. Here are the correct specs: [paste from invoice]

AI Overwhelms with Detail

Problem: Generated content too verbose

Solution: Request concise version

This is good but too long. Please condense to key points only,
max 200 words for the description section.

Future AI Capabilities

Coming soon:

• Voice-to-text meeting transcription
• Automatic photo organization
• Predictive task suggestions
• Budget tracking and alerts
• Timeline auto-updates from progress

Next Steps

06 - Advanced Features and Best Practices - Master advanced Obsidian techniques

Quick Reference

Best AI Use Cases:

1. Photo analysis and description
2. Product research and specification gathering
3. Meeting note enhancement
4. Consistency checking (spelling, formatting)
5. Decision analysis and documentation
6. Template creation
7. Content organization

AI Workflow Pattern:

1. Provide clear request with context
2. Reference existing examples/templates
3. Review AI output critically
4. Enhance with human expertise
5. Verify technical accuracy
6. Iterate if needed

---

AI saved ~30-45 hours on the Garage Project documentation. Your results will vary based on project complexity.