Agentic Workflow Builder for Banking Security Scenarios

Banks rehearse their responses to fraud, suspicious activity, and compliance failures the way pilots rehearse emergencies, but writing those response playbooks is slow expert work. This project builds a system that drafts them: given a security scenario, it generates the workflow, the recommended actions, and the explanatory documentation. Working in python, the intern creates a collaborative tool where users select or describe scenarios such as fraud detection, suspicious activity reporting, or compliance checks. A langgraph graph orchestrates the generation process through stages for scenario interpretation, workflow construction, action recommendation, and report writing, while an agentic-ai layer coordinates the stages and loops back when a draft is incomplete. Generative AI through openai models produces the substance: detailed step-by-step scenario workflows, prioritized recommended actions, and clear reports. Outputs are delivered as scenario maps and text-based security documents that a review team could critique and adopt. This is a hard, portfolio-worthy build. The intern learns security process automation, multi-agent workflow design, and the craft of getting language models to produce structured operational documents rather than loose prose.