What is Prompt Engineering?
Prompt engineering is the practice of designing natural language inputs — prompts — to get better, more reliable outputs from AI models. It includes techniques like few-shot examples, chain-of-thought reasoning, role assignment, and structured formatting.
The discipline emerged alongside large language models and is genuinely valuable. A well-crafted prompt produces dramatically better results than a vague one.
Where prompt engineering excels
Prompt engineering is the right tool for interactive, conversational AI use cases:
- Content generation. Writing, summarizing, translating — where the output is consumed immediately and the prompt captures the full context.
- Analysis. Asking a model to reason about data, code, or text within a single session.
- Exploration. Brainstorming, prototyping ideas, getting a first draft of anything.
In these cases, the prompt is the specification. The interaction is ephemeral by design, and that's fine.
Where it falls short
Prompt engineering breaks down when applied to product development — building software that teams maintain, users depend on, and AI agents execute autonomously.
Prompts are ephemeral. You write it, the agent consumes it, and it's gone. The next developer, sprint, or agent session starts from zero. There's no institutional memory.
Prompts mix concerns. A single prompt tries to carry the objective, constraints, edge cases, and design intent in unstructured natural language. The agent has to guess which parts are requirements and which are suggestions.
Prompts lack verification criteria. A prompt tells the agent what to build. It doesn't define what "done" looks like. Without explicit success criteria, there's no way to evaluate the output except subjective review.
Prompts don't compose. You can't hand a prompt to a different team member, a different agent, or a CI pipeline and expect consistent results. The context is too implicit, too tied to the original author's mental model.
From prompts to intent
The gap between prompt engineering and production software is the same gap that intent engineering fills. Where a prompt says "build a password reset flow," an IntentSpec defines the user problem, measurable outcomes, constraints, edge cases, and verification criteria.
The prompt optimizes a conversation. The spec defines a contract. Both have their place — but when AI agents are writing production code, the contract is what matters.