This project focuses on designing structured workflows for integrating generative AI into organizational decision-making. The work emphasizes prompt engineering, evaluation frameworks, documentation, and human oversight to improve consistency, transparency, and reproducibility in AI-assisted processes.
Rather than treating large language models as standalone decision-makers, this work demonstrates how prompt design, evaluation criteria, and human review can be combined into repeatable workflows that improve efficiency while maintaining accountability.
Overview
This project includes the development of:
- AI-assisted workflow design
- Prompt engineering strategies
- Evaluation frameworks for AI outputs
- Human-in-the-loop review processes
- Technical documentation and training materials
The workflow begins with structured user inputs, automatically retrieves relevant opportunities through Python scripts and external APIs, applies prompt templates using large language models, and concludes with human review and evaluation before results are finalized.
Prompt Engineering Workflow
Key components of the workflow include:
- Defining search criteria using structured keywords
- Developing relevance scoring criteria
- Designing and refining prompt templates
- Evaluating prompt performance against baseline results
This process supports consistent AI outputs while allowing prompt templates to be iteratively improved through evaluation and documentation.
Outcomes
This work resulted in:
- Standardized prompt engineering guidance
- Structured evaluation criteria for AI outputs
- Training materials for new users
- A repeatable workflow supporting consistent human review