Ending AI Slop Through Better Management and Context

The Root Cause of AI Slop

AI slop—the proliferation of low-quality, generic, or irrelevant content—is fundamentally a management problem, not a technology problem. Hilary Gridley identifies three primary drivers: decentralized and inconsistent usage of AI tools, a lack of a "central brain" or shared canon of information, and the absence of a clearly defined quality bar. When managers fail to articulate what "good" looks like, employees default to using AI as an easy button, leading to a "slop doom loop" where poor inputs lead to poor outputs, further eroding team judgment.

Craftspeople vs. Context Carriers

Kipp Bodnar distinguishes between two types of workers: craftspeople (those with deep domain expertise) and context carriers (those whose primary value was moving information around). AI has largely obsoleted the latter. The slop epidemic is most prevalent among context carriers who lack the deep domain knowledge to verify AI outputs or set standards. To break the cycle, managers must shift from being information conduits to being architects of quality, ensuring their teams are building skills that would remain valuable even if AI tools were removed tomorrow.

Defining the 'Taste Profile'

To combat slop, teams need a "taste profile"—a repository of non-demographic data, core emotional brand stories, and strategic narratives. This context allows both humans and AI agents to make decisions that align with the company's specific voice and goals. Gridley emphasizes that this is simply good management: providing the right information at the right level of depth is the core challenge of leadership, whether managing humans or agents.

Working Backward from AI-Native Workflows

Instead of incrementally adding AI to existing, outdated processes, leaders should envision what an "AI-native" version of their team's work looks like one year from now. By working backward from this future state, managers can set clear expectations and constraints. Gridley advocates for a "30-minute prototype constraint" to prevent teams from wasting time on elaborate AI tools that solve no real business problem, ensuring that all AI efforts remain tethered to actual value creation.

Building Executive Editor Tools

Rather than building "a second you" (which often fails to capture nuance), managers should build tools that codify their standards. Gridley demonstrates using an "Executive Editor GPT" that acts as a rubric-based feedback loop. This allows the manager to scale their judgment and standards without becoming a bottleneck, effectively teaching the team how to self-correct and maintain quality independently.