The Exact Claude Setup That Writes My Documents Automatically

How to Automate Document Creation Using AI

Introduction to AI in Document Creation

• The speaker shares their experience of using AI, specifically Claude, to generate polished proposals from call transcripts in about 30 seconds.

• This method is applicable not only for proposals but also for various documents like contracts, SOPs, and follow-up emails.

The Three-Step System

• The process consists of three primary steps: identifying what "good" looks like, reverse engineering existing documents, and embedding the aesthetics into a project for recurring use.

• The first step involves analyzing existing high-quality documents (contracts, SOPs, proposals) to establish benchmarks for future AI-generated content.

Extracting Aesthetics and Structure

• Key elements such as structure, styling, fonts, and spacing are extracted from existing documents to ensure that AI-generated outputs maintain brand consistency.

• Claude is highlighted as the current tool used due to its unique features; however, other platforms like ChatGPT and Gemini are expected to adopt similar capabilities soon.

Crafting Effective Prompts

• The speaker emphasizes the importance of crafting effective prompts when working with AI. They suggest using terms like "methodically" and "reverse engineer" for better results.

• A basic prompt example is provided that instructs the AI on what specific elements to extract while allowing flexibility for additional insights.

Demonstration of Reverse Engineering Process

• An example document (a fictional proposal template) is shared with Claude. It includes specific styles and formats necessary for accurate reverse engineering.

• Multiple file formats (PDF and DOCX) are recommended for optimal results during the reverse engineering process. After processing time (10–15 minutes), Claude provides a detailed breakdown of the document's specifications.

How to Create a Skill Using Claude

Creating a Skill

• The process of creating a skill is straightforward, involving simple prompts. The prompt instructs the AI to use the skills creator to generate a skill that matches an original document pixel for pixel.

Enabling Skills in Claude

• Before executing the prompt, ensure that skills are enabled in Claude by navigating to settings and checking under capabilities.

• Within capabilities, confirm that the "skills" option is turned on and also enable the "skill creator," which allows for the creation of new skills.

Utilizing Extracted Information

• After extracting information methodically, paste the specific prompt into the chat thread. This will utilize the skills creator to develop a dedicated skill that produces proposals matching original documents accurately.

Copying Skills for Future Use

• Once created, you can copy your new skill into your personal skills section for easy access. Sharing with others involves downloading and sending it so they can upload it into their own platform.

Embedding Learning into Cloud Projects

• The final step involves integrating everything into a cloud project where transcripts can be dropped in for automated processing. This includes adding reference files that provide context or inspiration for drafting proposals or reports.

Drafting System Prompts

• When drafting system prompts, it's crucial to specify what you want from the AI clearly. Include reference materials if necessary to guide its output effectively.

Writing Base Prompts

• Start with a base prompt before enhancing it using prompt improvers offered by platforms like Anthropic and OpenAI. These tools help refine prompts based on best practices in prompt engineering.

WW Framework for Effective Prompting

• Utilize the WW framework (What, Why, How) when crafting prompts:

• What: Clearly state what task you want completed.

• Why: Explain why this task is important; this helps AI infer additional context.

• How: Specify constraints such as format or structure required in responses.

Example of Base Prompt Structure

• An example base prompt begins by assigning a persona (e.g., expert proposal writer), outlining tasks (drafting proposals from meeting transcripts), and indicating available examples within its knowledge base.

Creating a Pixel-Perfect Proposal Using Claude

Overview of the Proposal Process

• The speaker emphasizes the importance of using a specific skill, referred to as "X," for creating an aesthetically pixel-perfect report. This skill name should be customized for individual projects.

• The speaker introduces their offering and mentions that further details will be provided in a markdown file within the knowledge base, summarizing two key offerings.

• A clear structure is outlined for the proposal, specifying six essential components that must be included in the final document.

Generating an Effective Prompt

• The speaker discusses how to improve the initial prompt using Claude's capabilities, highlighting that it should result in higher quality responses from AI.

• Instructions are given on accessing Claude’s platform (platform.cloud.com), where users can input their base prompt and generate an improved version.

Utilizing Prompt Improver Features

• After generating a new prompt, it's noted that users should check if they are utilizing a reasoning model (Opus 4.5 with extended thinking).

• The improved prompt includes explicit instructions about what to look for in the knowledge base, such as previous proposals and company offerings.

Structure and Format Enhancements

• The format of the prompt has been enhanced with headers to clarify each segment's purpose, aiding AI understanding during processing.

• A "scratchpad" feature is introduced, allowing Claude's models to internally critique their responses before delivering them back to users.

Finalizing and Testing the Proposal Output

• An additional section on format requirements is added in the improved prompt along with specific cloud skills mentioned earlier.

• The speaker demonstrates how to implement this project by simply entering a command ("draft a proposal from the below transcript") followed by pasting in relevant content.

• A successful output is shown where generated proposals match exactly with provided templates, confirming effectiveness through thorough checks against fictional data.

Project Creation Steps

• To create a new project from scratch, users are guided through naming and setting up files within Claude’s project section.

AI Implementation and Testing Strategies

Core Offerings and AI Template

• The speaker discusses the importance of defining core offerings and templates that the AI will reference for generating outputs. This foundational setup is crucial for effective AI utilization.

Importance of Persistence in AI Usage

• It is emphasized that initial outputs from AI may not meet expectations, highlighting the need for persistence and effective testing to refine results. Users should be prepared to iterate on their inputs to improve outcomes.

Bridging Expectation Gaps with Iteration

• The speaker notes a common gap between user expectations of "good" output and what AI initially provides. Continuous adjustments in context and prompts are necessary to align these perceptions over time.

Achieving Convergence Between Human Standards and AI Outputs

• As users iterate on their interactions with the AI, they can gradually bring its outputs closer to human standards, eventually allowing for complete outsourcing of specific tasks to the AI when convergence is achieved.

Recap of Effective Document Creation Process

• A structured approach is outlined: start by reverse engineering an existing quality document (like contracts or proposals), embed learnings into a skill within Claude, and utilize this skill within projects for efficient document generation based on simple inputs like transcripts.

Additional Resources Offered

• The speaker mentions "Blow," a 30-day free insight series aimed at applying AI effectively in business contexts, along with an invitation for potential collaboration based on mutual fit regarding services offered.