The initial hurdle of getting teams to use AI is over. The more difficult problem is ensuring they use it correctly to produce high-quality work, not just 'slop'—low-effort content generated by outsourcing taste and judgment to models like Claude.
'AI Slop' flourishes when leaders don't explicitly define what 'good' looks like. This failure, combined with decentralized tool usage and a lack of a central source of truth, allows low-quality, AI-generated work to become the default standard within an organization.
AI is bifurcating managers into two groups: 'deep craftspeople' who scale their expertise, and 'context carriers' who just move information. AI automates the latter's role, revealing their lack of craft and making them the primary source of low-quality 'AI slop.'
AI implementation isn't neutral; it creates a feedback loop. Using AI to teach skills makes teams and systems smarter. Conversely, allowing teams to outsource judgment to AI creates a 'slop doom loop' where cognitive rot makes both people and systems worse over time.
To ensure consistent, high-quality AI output, centralize context into a 'taste profile.' This document should go beyond demographics to include deep emotional and psychological data about your customer, plus core brand narratives, serving as the AI's instructional foundation.
The solution to 'AI slop' is not a new management technique. Instead, AI's power simply makes foundational leadership principles—setting clear expectations, defining quality, and providing context—more critical than ever before. Good management is the core solution.
AI reframes the nature of work. An employee's primary role shifts from manual execution to holding ultimate accountability for the quality of the final product, even if an AI performs most of the labor. The human is the final quality check and owner of the outcome.
To codify your expertise into a custom AI tool, use AI to discover your own subconscious standards. Feed a model 'before' and 'after' examples of your work (e.g., a raw draft and your edited version) and ask it to identify the recurring patterns and principles you apply.