Internal Audit – Payment Process at ABC HealthCare Leveraging Data Science and AI to Strengthen Payment Oversight

ABC HealthCare is a global pharmaceutical company specializing in cancer research. It operates with:

• 206 independent research teams across universities and research organizations worldwide
• Centralized legal and finance services managed by ABC HealthCare
• Payments team based in Thailand, handling over 200,000 payment requests annually

• Fraudulent invoices submitted by research teams
• Invoices not matching claims
• Employee fraud within the finance team

A preliminary investigation recommends a comprehensive audit using advanced data analytics.

You are provided with a .zip file containing 4 datasets:

Includes 2000 sample payment records with fields:

• Date received: When the invoice was received.
• Date of invoice: When the invoice was issued.
• Date of authorisation: When the invoice was approved.
• Payment due date: When the payment is due.
• Date of payment: When the payment was made.
• Research team: Team responsible for the expense.
• Submitted by: Person who submitted the invoice.
• Authorised by: Person who approved the invoice.
• Payment authoriser: Person who approved the payment.
• Invoice number: Unique identifier for the invoice.
• Description of spend: What the invoice is for.
• Invoice value: Amount on the invoice.
• Payment amount: Amount paid.
• Payment Status: e.g., Paid, Pending.
• Type of expense: e.g., Lab suppliers, Consulting, etc.
• Company: Vendor or service provider.
• phone_number: Contact number for the company or submitter.
• email: Contact email.

Includes:

• Research team: Name or identifier of the research group.
• Director: Person leading the team.
• Location: Geographic or institutional location.
• Affiliation: Associated institution or department.
• Research type: Field of research (e.g., Basic Research', 'Translational', etc.)
• Annual budget: Total yearly funding allocated to the team.
• Item type: Specific budget item (e.g., Lab suppliers, Consulting, etc.).
• Item budget: budget associated with the item.
• Comments: Additional notes or context.

Includes:

• Team: The name or identifier of the team.
• Location: Where the team or individual is based.
• Name: The name of the team member.
• Role: The position or function of the person within the team.

Same structure as Payments Master, with an added Fraud_flag column:

• 1 = Fraudulent
• 0 = Not fraudulent

• Identify teams with highest spend
• Track volume/time trends by category and location
• Analyze average payment time by expense type, value, and location
• Compare team spend vs. budget

Hypothesis: All invoices take ~same time ±1 day
Task: Predict time to payment using regression or similar modeling

Hypothesis: Invoices are random; no fraud pattern
Task: Use Z-scores or similar to detect statistical outliers for audit sampling

Build and explain:

• Supervised model to predict fraudulent payments
• Unsupervised model to detect anomalies

Include evaluation metrics like:

• Confusion matrix
• Precision, recall, F1-score
• ROC-AUC or perplexity

Demonstrate how an AI tool (e.g., GPT) can:

• Summarize findings from Sections 1 & 2
• Generate audit insights and recommendations

• Presentation: Max 10 slides for Group Internal Audit Leadership Team
• Files: Include scripts, notebooks, apps, or tools used
• Documentation: Explain your process, assumptions, and conclusions