Margin Compliance AI Agent (Oracle)
Project Summary
The Margin Compliance AI Agent is a conversational AI agent designed for Oracle Fusion Cloud that gives sales representatives real-time floor pricing and margin compliance guidance during active deal negotiations, without requiring pricing expertise or manager escalation.
Problem Statement
Sales reps closing deals have no way to know how low they can discount while keeping the company margin compliant. Target margin thresholds live in policy documents reps cannot access mid-negotiation. The result is margin leakage, delayed deals, and lost confidence at the most critical moment in the sales cycle.
The Solution
The agent was designed in Oracle AI Studio using a topic-and-tool architecture. It retrieves live pricing data including list price, cost, discount, and customer pricing strategy via REST endpoints, and interprets the company's uploaded margin policy document using RAG AI. The AI handles natural language understanding, tool orchestration, and policy interpretation. Margin calculations are handled by a separate deterministic service to ensure reliable, accurate outputs. The result is a rep who can negotiate confidently in plain language and get a real answer in seconds.
Sample Interaction
Rep: "My customer wants 50 units of Item X but their budget is $4,000. How low can I go?"
Agent: "The list price is $90 per unit, which is $4,500 for 50 units. Based on your margin policy, the floor price for this item category is $78 per unit, which keeps you compliant. At 50 units that is $3,900, under their budget and you are still protected. Want me to check if a volume discount applies if they move to 75 units?"
Responsibilities
(See my Core Responsibilities across all projects)
• Independently conceptualized the agent use case and authored the full product specification including agent architecture, topic triggers, tool call sequencing, and data source mapping
• Built hands-on fluency in Oracle AI Studio and AI Prompt Lab, testing and iterating on prompt logic and output formatting
• Facilitated working sessions with the Product Manager to validate the architecture and resolve key design decisions
• Presented the specification to senior leadership and incorporated feedback into subsequent iterations
• Made deliberate architectural decisions including separating arithmetic from the AI and using RAG instead of a hard-coded rules UI
Outcome
• Full product specification completed, reviewed, and validated by senior leadership
• Specification queued for future engineering implementation