Banking customer-onboarding teaching case with AI-constrained fraud-screening agent

Design · All designs
Evidence · 2 claims (2 for) · 1 study (1 design), q2 · 0 of 1 report an effect size · 2 claims rest on one study

Description

A fictitious retail-bank customer onboarding process (registration, fraud assessment, approval) facing a scalability crisis, deliberately rich in regulatory constraints, knowledge-intensive work, and heterogeneous data sources. Document B proposes AI agents, prioritizing email-based registration extraction and RAG-based fraud screening; the fraud-screening agent is "deliberately constrained by a known class-imbalance problem (the data science team reports high recall but imperfect precision)". Case materials, module instructions, and datasets are openly available in a GitHub repository.

Design Implications

Context

Requirements

• Two main documents (Document A as-is process with seven roles; Document B redesign brief) distributed at different course points

Constraints

• The fraud agent's class-imbalance problem forces students to reason about when output can be trusted autonomously and when a human must remain in the loop

Target Learners

• Master-level students with no prior BPM or AI knowledge assumed

Learning Goals

• designing, implementing, and evaluating AI-augmented process variants across the full lifecycle
• governance concerns such as human-in-the-loop routing and fallback handling

Claims

• At the implementation checkpoint, several groups added an explicit human-approval step after encountering local-LLM structured-output reliability limitations [+W]
• Checkpoint artifacts prompted Master-level student groups to revise initial AI overestimates after baseline simulation evidence accumulated [+W]

Related Designs

• Evolving-artifact course organization: each module's output becomes the next module's input

Examples

-

Key Sources

• Amin Jalali. (2026). AI-Decision Checkpoints for AI-Augmented Business Process Management: Framework and Educational Instantiation. https://arxiv.org/abs/2610.06207

Page metadata

Field	Value
type	design
id	banking-onboarding-teaching-case
title	Banking customer-onboarding teaching case with AI-constrained fraud-screening agent
description	A fictitious retail-bank customer onboarding process (registration, fraud assessment, approval) facing a scalability crisis, deliberately rich in regulatory constraints, knowledge-intensive work, and heterogeneous dat...
status	review
generated	by process:settle-candidates on 2026-10-10
trust	unverified — no one has confirmed this page's content
sources	1 — amin-jalali-2026

Frontmatter is the page's machine-readable record — see the Schema & Guide.