Domain Aware Prompt Engineering

Taravarthi, Subhash and Koilakonda, Raghunath Reddy and Bikkavolu, Venkatasatyaravikiran and Tarakampet, Saikrishna (2025) Domain Aware Prompt Engineering. World Journal of Advanced Engineering Technology and Sciences, 16 (1). pp. 569-574. ISSN 2582-8266

Introduction: In this world of Generative AI, it is very important to understand the thought process for generating accurate results. Implementing a Generative AI application using different models has become a very common practice using prompt engineering. Generic Prompting often results in hallucinations, compliance risks and lack of actionable intelligence. To overcome these problems, we introduce Domain Aware Prompt Engineering which is an emerging technique to tailor Gen AI outputs on enterprise specific knowledge, semantics and decision making. Objectives: In this world we have multiple domains like Finance, Marketing, Supply Chain, Telecom and many more... Providing the insights of each domain in the prompt engineering often gives excellent solutions. Need to understand the cosmetics of each domain and put the flow in your prompt during the implementation of our enterprise Gen AI applications. This article demonstrates how to operationalize domain aware prompt engineering to strategically enable the building of accurate, secure and contextually intelligent Gen AI applications in enterprise environments. Methods: The proposed architecture caters with multiple functional domains like Finance, Supply Chain, Marketing, Telecom for a Text to SQL implementation where it requires domain specific to understand the functional aspects of underlying data. Give business users an area of self-service implementation on UX/UI where they utilize their domain knowledge to leverage high accuracy. This domain knowledge emphasizes their functional thought process. Results: The Azure-based Gen AI Text-to-SQL architecture with domain aware prompt engineering has emerged as the enterprise applications. It increased the accuracy up to 95% based on domain knowledge of SMEs, and empowered non-technical users to craft precise SQL queries using natural language, which means less dependence on IT. The ability to seamlessly query across databases like Snowflake, Databricks, and Oracle has really sharpened decision-making. Plus, with Azure AD and RBAC in place, security compliance is a solid 100%. Thanks to Azure services, deployment is scalable and boasts a reliability rate of 99.9%. A case study in the supply chain sector revealed that query times plummeted from days to mere minutes, leading to a 50% boost in analyst productivity and an 80% drop in errors, all of which enhances strategic agility. Conclusions: A scalable GenAI Text-to-SQL setup that leverages Azure OpenAI, Fast API, and various Azure services is transforming the way enterprises approach Decision Intelligence. It allows users to make natural language queries, gain insights across different databases, and maintain strong governance. This not only lessens the dependency on IT but also enhances agility through effective change management.