Beyond Automation: Mike Caulfield's Case for AI as a Tool for Creative Exploration

The Case for Rethinking AI Product Design

In his Substack essay “Tilling the Garden,” Mike Caulfield proposes a fundamental reframing of how developers and product teams should approach AI integration. Rather than chasing automation—the elimination of human effort through algorithmic replacement—Caulfield argues that more interesting and useful applications emerge when AI serves as an interactive tool for exploration and refinement. The distinction is subtle but consequential: automation treats AI as a substitute for human decision-making, while cultivation treats it as a partner in an iterative process where human judgment remains central.

Automation Versus Cultivation

Caulfield suggests that the automation mindset, dominant in business and technology for decades, has shaped how teams initially approach AI. According to Caulfield, this framing prioritizes efficiency metrics—reducing labor, speeding workflows, eliminating steps. But he argues this lens obscures what AI does well in domains where ambiguity, context-shifting, and creative judgment matter most. Caulfield contends that tasks involving exploration, refinement, or open-ended problem-solving might benefit from a different mental model: cultivation, where the human and the AI work in conversation, with the human maintaining editorial control and the AI offering suggestions, alternatives, and provocations.

The cultivation approach, according to Caulfield, acknowledges that many valuable outcomes cannot be fully specified in advance. Unlike manufacturing automation, which aims for predictable, repeatable processes, interactive AI tools could enable humans to discover solutions they did not anticipate when starting. Caulfield’s argument implies that product teams chasing pure labor reduction may miss opportunities to build tools that augment human capability rather than replace it.

Practical Implications for Product Teams

Caulfield’s framework suggests concrete consequences for how teams design AI features. According to Caulfield, systems built around cultivation require different interfaces, feedback loops, and success metrics than automation-first designs. Rather than optimizing for “tasks completed per user per hour,” cultivation-oriented tools might measure engagement depth, user satisfaction with outcomes, or the quality of discoveries made. Caulfield hints that this could reshape how teams prioritize transparency, control, and iterative refinement in their AI systems.

One tension worth noting: Caulfield does not fully address how to measure or scale the value of exploratory AI work, nor does he deeply explore which user segments or industries most need cultivation versus automation. His framework is deliberately metaphorical, leaving room for interpretation but also creating ambiguity about boundary cases.

Why This Matters

If Caulfield’s argument gains traction in product design circles, it could challenge the prevailing focus on cost reduction as the primary lever for AI ROI. Teams building customer-facing AI tools—whether in content creation, design, research, or decision-support—might begin asking whether they are optimizing for speed at the expense of meaningful human agency. The cultivation framing also reopens questions about user trust and control, which pure automation often sidelines. For companies struggling to justify AI investments beyond headcount reduction, reframing AI as a vehicle for better exploration and discovery offers an alternative narrative, though validating that narrative through metrics and user outcomes remains an open challenge.