# I Don't Fear AI Slop. I Fear Untold Stories - Zack Proser, WorkOS

## Executive summary

The talk argues that instead of fearing 'AI slop,' the greater concern is the number of valuable stories and insights that remain untold. The speaker, Zack Proser, proposes the concept of building a 'factory'—a codified, automated system or process—for a person, team, or niche domain. He demonstrates this by detailing the creation of Blogbot, a system built on Cloudflare that provides 'asymmetric leverage' by allowing colleagues to generate high-quality, on-brand drafts (85-90% complete) from raw inputs (like event transcripts) using LLMs, while still requiring human oversight to maintain craft and quality.

## Key takeaways

- The Focus Shift: Untold Stories over AI Slop: The speaker is more concerned with the loss of valuable, unshared knowledge than with the quality of AI-generated content. He argues that with proper intention, accountability, and process design, AI can improve storytelling and processes, allowing more people to share their work effectively.
- Asymmetric Leverage via Blogbot: Blogbot, built on Cloudflare, allows users to input raw data (e.g., event transcripts) via a simple Slack interface and receive a draft that is 85-90% complete. This system automatically sources context and framing, significantly increasing publishing velocity while preserving the need for human review and final polish.
- The Craft Metaphor: Drawing on a blacksmithing metaphor, the speaker asserts that 'craft' runs through the entire production line. Even when machines handle the heavy lifting, the quality and dedication to the final product must remain human-controlled (the 'spark,' 'bellows,' and 'quench').
- Building Factories for Niche Audiences: The concept was applied to a personal project: building a bedtime story app for two toddlers in four hours. This demonstrated that the factory pattern is applicable even for extremely small, highly specific audiences, proving the speed and efficacy of the pattern.

## Technical details

- Blogbot System Architecture: Blogbot is a system built on Cloudflare that automates the drafting process. It accepts inputs (e.g., event transcripts) and uses internal logic to produce structured, on-brand blog posts. It features a data connector to multiple company information sources and includes a special endpoint for image generation that mimics the company's design team output.
- LLM Integration and Workflow: The system leverages LLMs to accelerate content creation, allowing the user to move from initial prompt to a substantial draft quickly. The process requires human intervention for final review, fact confirmation, and personal sign-off, ensuring quality remains paramount.
- Development Speed and Iteration: The speaker recounts building the bedtime story app from zero to a working prototype in four hours, highlighting the rapid development cycle enabled by the factory pattern, even for complex, personalized artifacts.

## Practical implications

- Identify internal processes or specialized knowledge within your team that are currently undocumented or underutilized.
- Codify these processes into a tool or workflow ('factory') that can be accessed via a simple interface (e.g., Slack integration).
- Use AI/LLMs to handle the tedious, high-volume parts of the process (e.g., drafting, summarizing, formatting), while mandating a human review step to ensure quality and brand integrity.
- Apply the 'factory' concept not just to large teams, but to small, specific, or highly niche audiences.

## Topics

Process Automation, Generative AI, Content Strategy, Build Engineering, Workflow Design, Asymmetric Leverage, Zack Proser

Source: https://www.youtube.com/watch?v=1et3Lb_KLQo