Newsletter Idea Curation + Publish Workflow Refinement with Claude

This workflow illustrates the newsletter creation process (Recent post, upcoming posts, Nivationist announcements, and newsletter element idea curation). The workflow involves MailerLite (the newsletter editor/sender), Notion (the Claude-generated single-truth idea-curation library), and Google Keep (the fast idea-jotting app).

<aside> 🎯

The goal of this newsletter workflow refinement is three-fold:

• Workflow Refinement (Primary): Reduce writer’s block and perfectionism (the bottleneck that leads to 3+ hours from newsletter creation to “schedule-ready”) to limit the workflow completion time to under an hour.
• Anticipate the Effectiveness for Another Workflow: Claude offered a similar refinement approach to two workflow components: blog publishing & newsletter curation/publishing (this document). Since the timing of the test aligns closer with the newsletter creation, the results of the newsletter revamp test will be a great predictor of the effectiveness of the blog publishing workflow.
• Mental Offloading for Business Growth: Make space to begin social media marketing initiatives to increase subscriber count and generate testimonial/case study material through freelance work. </aside>

<aside> 💡

Why Do I Choose to Revamp the Newsletter Curation + Publish Workflow?

• Semi-Established Consistency of Habit —> The newsletter publishing workflow is well established, but the methods of content curation are scattered without a central referenceable hub.

• Clear Distinction Between Human-Controlled vs. AI-Assisted —>

  • The content (personal reflections, raw notes, experiences, skills, decision-making, etc.) is free from AI, which retains ownership, integrity, and ownership to my audience, brand, and myself.
  • The presentation (do the words fully reflect what I wish to articulate, ways to better help my audience, etc.) is a collaborative effort with AI.
  • Operational processes (such as repetitive formatting, transfer of information, element checklists, etc.) could be laid out to be fully automated in the future.

• Immediate Impact —> One of the biggest bottlenecks right now, and it is easy to validate the level of satisfaction. </aside>

Step 1: Current Process Evaluation with Claude

Step 1 was conducted on Friday, May 15, 2026.

<aside> 📖

Claude was given the context that this newsletter curation + publish workflow is a bigger content management system.

The newsletter curation/publish workflow (this doc) and the blog publish workflow (another doc) were both targeted to be revamped with a similar level of approach.

For more information about the AI prompts I used and the prompt findings I’ve discovered during the process, see the blog publish workflow case study.

• 1 Google Docs artifact on the step-by-step instructions on how to implement and use the workflow with tailored AI prompts.
• Generated the [Public] Newsletter Content Bank on Notion as a fillable template to jot down weekly reflections, insights, third-party inspirations, etc., that could be used to populate and refine for the next newsletter. </aside>

Step 2: Human Review of Claude Artifacts + Final Refinements Before Deployment

Step 2 was conducted on Friday, May 15, 2026.

<aside> ➡️

AI responses tend to hallucinate, so I would need to review and revise the artifacts that Claude had given me before its initial test deployment.

This step also serves as an opportunity for me to test the synergy between Claude and built-in AIs within my existing blog management tools (Google Drive —> Gemini, Notion—> Notion AI, Elementor —> Elementor AI)

Below is a ChatGPT-generated workflow diagram that was made from my audio-to-transcription process of how I review the Claude Artifacts, give feedback to Claude, and keyword pull comments into Google Keep and Google Tasks using Gemini.

</aside>

Step 3: Real-Life Workflow Testing

Step 3 was conducted on Wednesday, May 20, 2026.

A newsletter is due this week, which is a great opportunity to test the recommended workflow revamp in real-time.

<aside> ✅

Part 1: Claude AI Content Curation + MailerLite Email Setup

Previous Workflow Time: ~2 hours

Claude’s Target Time: 10 minutes

Current Workflow Time (Actual): 25 minutes

Part 1 Workflow Description

Claude splits the newsletter curation + publishing workflow into two parts. Part 1 consists of the following main deliverables:

• Selecting 1-3 rotating prompts to write about in the newsletter
• Writing a short description for each selected rotating prompt
• Compiling and uploading context files
• Let Claude generate a post description of the recent blog, 1-3 insights pulled straight from the next upcoming post, and further elaborate on rotating prompt responses based on my brand voice from my blog and past newsletters (uploaded as context files) to reduce writer's block and use as a starting reference for human revision. See Part 2.
• Let Claude generate a list of email subject lines and preheader descriptions to use as a reference for human revision. See Part 2.

In-Process Friction Points / Quick-Fix Solutions

1. I have to update the [Public] Newsletter Content Bank which is a single-truth hub for all notes, reflections, experiences, insights, etc., for newsletter content curation on the day of the workflow test.

• Writer’s block: Having to think back to what happened this week. It takes me 13 minutes to write three raw inputs, one for each newsletter topic category.

[Quick-Fix Solution] New Habit Developed: I started jotting down ideas and reflections via audio and text on Google Keep, reducing this friction in the future.

1. Manually searching and pasting the context needed for each Claude AI prompt.
   • Claude’s workflow refinement instructions present each step as a standalone prompt with no cross-referencing. Claude doesn’t recommend its own time-saving features, such as using Projects and/or Skills. This led to context redundancy.

[Quick-Fix Solution]: In-the-Moment Initiative: Instead of repeatedly copying/pasting the same file as instructed, I uploaded all the files needed for each prompt by consistently using the referencing shorthand (ref: “file name”) immediately after the context. I jot down each reference file name on paper for quick/consistent referencing for subsequent prompts and to prevent redundant uploads.

1. Claude AI became inoperable in the final prompt after the free limit was hit. I was unable to do another prompt until hours later.

[Quick-Fix Solution]: In-the-Moment Initiative: I proceeded with the test using ChatGPT instead of Claude to assist in creating the subject line and preheader description for this week’s newsletter on MailerLite.

1. Claude’s estimated target time of 10 minutes for Part 1 is too short for real-world practicality and execution.

[Quick-Fix Solution]: Realistic Discretion: Claude’s estimated target time is the “ideal,” a theoretical benchmark that may be impractical because it doesn’t have past data references to gauge a realistic time of completion.

• The metric, Claude’s Target Time, could be retained as a long-term goal that could be worked towards in each workflow iteration.
• The metrics, Previous Workflow Time and Current Workflow Time (Actual), will be given greater weight to calculate the realistic productivity improvement in this case study. See how it’s done.

Part 2: MailerLite Assemble

Previous Workflow Time: ~1.5 hours

Claude’s Target Time: 20 minutes

Current Workflow Time (Actual): 55 minutes

Part 2 Workflow Description

Part 2 consists of the following main deliverables:

• Transfer generated content into the newsletter for final human review, editing, and refinement.
• Written introduction letter to the audience + disclosure of AI-assisted workflow revamp.
• Ensure all links, formatting, brand colors, etc. are accurate.
• Newsletter test email + email edits (if applicable) + schedule-to-publish.

In-Process Friction Points / Quick-Fix Solutions

• Although the content will always originate from me, how the content will be presented is a collaborative effort with AI. The newsletter elements generated from my week’s experiences, highlights, and notes are too wordy and sometimes create hallucinations in an attempt to add more context to my story.

[Quick-Fix Solution]: Alternative Use & Human Review: I use the generative content as example prompts to help me overcome writer’s block and as a starting point for me to rewrite, refine, and correct the content elements. I also do text searches within the referenced files to see if the insights are pulled straight from the files instead of accepting summarized or hallucinated responses.

</aside>

Step 4: Calculating Results + Next Steps Analysis

Step 4 was completed on Tuesday, May 26, 2026.

<aside> ✅

Results

The results of this process revamp attempt do not meet the 1-hour time limit goal. However, the new workflow does improve the previous workflow's completion time significantly.

• It takes ~62% less time to complete the newsletter creation process (idea curation + publishing).

  *Calculation Method: % Change Between Total Previous Workflow Time and Total Current Workflow Time (Actual): (210-80)/210 100 = 61.9%

• The current process reduced the gap to the ideal from 180 minutes to 50 minutes, eliminating ~72% of the original gap.

  As discussed previously, Claude’s Target Time could serve as a long-term goal and be used to calculate the % margin of improvement towards the ideal.

  *Calculation Method: Gap Difference Between Claude’s Target Time (Ideal) and Previous Workflow Time: (210-30=180) - Gap Difference Between Claude’s Target Time (ideal) and Current Workflow Time (Actual): (80-30=50) = 130 —> % Improvement Towards Ideal (130/180) 100 = 72.2%

• It takes ~79% less time to complete Part 1 of the workflow (content curation + content generation). This is largely attributed to the added content curation system (Google Keep + Notion), a reduction of writer’s block due to content generation, and an outflow of human review/editing to Part 2.

  *Calculation Method: % Change Between Part 1 Previous Workflow Time and Part 1 Current Workflow Time (Actual): (120-25)/120 100 = 79.2%

• It takes ~64% less time to complete Part 2 of the workflow (newsletter assemble + human editing/review). The human review/editing (content rewriting, non-AI-generated introduction letter to the audience + disclosure of AI, AI accuracy review, etc.) was previously in Part 1 of the Previous Workflow, but due to the inclusion of AI, it has now been moved to Part 2. The large deviation between Part 1 of the Previous Workflow and the Current Workflow should be at the discretion of the reviewer. </aside>