The Economics of Generative AI: Prompt Engineering, Turnaround Time, and Value Creation
Year: 2026
Type: article
Abstract: Prompt engineering is an economically consequential human capability that can mediate the relationship between generative artificial intelligence (AI) access and productive performance. This paper develops three contributions. First, it proposes a five-layer framework linking workers' domain-specific human capital, access to AI tools, prompt quality, organisational capability and economy-wide distributional outcomes. Second, it develops a task-level model in which the proportional reduction in task turnaround time depends on the interaction between prompt quality, model capability and task–technology fit. Under an explicit multiplicative specification, prompt quality and model capability are complementary inputs over the interior of the feasible domain; the paper states the conditions for this result and identifies circumstances in which saturation or poor task fit can weaken or reverse it. Third, it uses five transparent, illustrative scenarios to examine how alternative distributions of AI access, prompt-training investment and domain expertise may shape productivity and inequality. These scenarios are not empirically estimated or calibrated counterfactuals: the available studies provide external empirical anchors for plausible mechanisms and task-level magnitudes, but they do not identify the model's structural parameters or economy-wide wage-distribution effects. A sensitivity exercise therefore varies the assumed annual depreciation of domain expertise and the elasticities of the turnaround-time function rather than presenting unsupported point estimates for aggregate productivity or wage Gini changes. The analysis suggests that broad-based capability development may improve the distribution of AI-enabled opportunities, while concentrated training and uncritical cognitive offloading may reinforce existing disparities. Universal access to effective AI-use education is consequently framed as a testable productivity-and-equity proposition, not as a quantitatively established policy effect.
Author: Akinmade, Tunde
Language: English
Subfield: Economics and Econometrics
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7225906412
