What Happened
Jock Busuttil published an essay arguing that the quality gap between vague and specific AI direction is larger than most practitioners appreciate - and that it produces differences in business outcomes, not just output aesthetics or surface-level quality.
The essay draws on his experience running directed AI experiments in business contexts and compares results from open-ended prompts versus prompts with explicit constraints, context, and defined success criteria. The framing he uses is "vibe business" versus directed AI work: treating AI as a tool you point at a general problem versus a collaborator you brief carefully.
Why It Matters
Prompt quality is often treated as a nice-to-have skill rather than a core competency that affects results. This framing understates the actual impact. When the difference between a vague and specific prompt produces outputs with different accuracy, different applicability to the actual task, and different downstream utility, that difference compounds across all the work that depends on the AI output.
For teams deploying AI at scale - where hundreds or thousands of AI-assisted tasks happen per week - the aggregate cost of low-quality prompting is significant. A 20% reduction in output quality across all AI-assisted tasks translates directly into rework, missed requirements, and decisions made on unreliable information. Prompt discipline is quality control.
The essay also makes a structural point: vague prompts extract vague value. When organizations deploy AI broadly without investing in prompt standards, they get diffuse, uneven results that are hard to attribute to the AI deployment. When they specify what good looks like - explicit context, defined output format, clear success criteria - results become predictable and comparable.
This has practical implications for how teams structure AI onboarding and training. The question is not just "how do we get employees using AI" but "how do we get them using it with enough precision to produce reliable outputs."
Our Take
Teams spend significant time evaluating which AI model or product to use and comparatively little time on the systematic quality of their prompting practices. The ROI on prompt discipline is often higher than switching to a more expensive model. A well-structured prompt on a mid-tier model frequently outperforms a vague prompt on a frontier model. Start with your prompting before blaming the model, and treat prompt quality as an engineering discipline rather than an individual skill.