Chapter 09 · Sustainability & AISustainability & AI

Prompt engineering

Meaning statusEstablishedSource recordDirect document linkedWhy these are different

Definition

The practice of designing inputs — instructions, context, examples, constraints — to steer the behaviour and output quality of generative AI systems.

References

arXivSahoo et al. — A Systematic Survey of Prompt Engineering in Large Language Models: Techniques and Applications (2024)

This source provides part of the technical or institutional basis for the definition.

NeurIPS 2020 / arXivLewis et al. — Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks

This source supports the explanation of how the term is applied, measured or governed in practice.

Overview

What it means

Because large language models respond to natural-language instructions, the phrasing, structure and context of a prompt materially change the result. The practice has been systematised into documented patterns and techniques, surveyed in the academic literature since 2024.

How it is used

Practitioners use prompting to extract structured data from sustainability reports, control tone and citation behaviour in drafting tools, and constrain models to provided sources — often combined with retrieval rather than prompting alone.

Why it matters

Prompting is the cheapest governance lever over generative AI outputs, but also the most fragile: it guides rather than guarantees. Workflows that matter — disclosure, compliance, claims — need verification layers beyond well-crafted prompts.