Loan Scoring Workflow - Usage Examples¶
Example 1: Basic Step-by-Step Workflow¶
Scenario: New loan inquiry for existing customer¶
Step 1: Initiate Workflow
User: "I want to check if customer 456 qualifies for a loan"
AI Response:
{
"step": "search_customer",
"success": true,
"customer_id": 456,
"customer_name": "Jane Smith",
"message": "Customer found: Jane Smith. Now select a loan product.",
"next_step": "select_product",
"prompt": "Which loan product would they like?"
}
Step 2: Select Product
User: "What products do we have?"
AI: (calls list_loan_products)
1. Personal Loan - 12% annual interest, Min: 10k, Max: 500k
2. Business Loan - 15% annual interest, Min: 50k, Max: 2M
3. Home Loan - 8% annual interest, Min: 100k, Max: 5M
User: "Let's use the Personal Loan"
AI Response:
{
"step": "select_product",
"success": true,
"customer_id": 456,
"product_id": 1,
"product": {
"id": 1,
"name": "Personal Loan",
"code": "PL001",
"interest_rate": 12.0,
"min_amount": 10000,
"max_amount": 500000,
"tenure_days": 365
},
"message": "Product selected: Personal Loan. Now enter the loan amount.",
"next_step": "enter_amount",
"prompt": "Enter the loan amount (min: 10000, max: 500000)"
}
Step 3: Enter Amount
User: "They want 75000"
AI Response:
{
"step": "enter_amount",
"success": true,
"customer_id": 456,
"product_id": 1,
"loan_amount": 75000,
"message": "Loan amount 75000 confirmed. Now calculating loan score...",
"next_step": "calculate_score",
"prompt": "Ready to calculate the loan score"
}
Step 4: Calculate Score
User: "Go ahead"
AI Response:
{
"step": "calculate_score",
"success": true,
"customer_id": 456,
"customer_name": "Jane Smith",
"product": "Personal Loan",
"loan_amount": 75000,
"score_result": {
"score": 820,
"risk_level": "LOW",
"recommendation": "APPROVED",
"breakdown": {
"created_on": "2026-02-02T14:30:00",
"risk_level": "LOW"
}
},
"message": "Loan scoring completed. Score: 820 (LOW risk). Recommendation: APPROVED ✓",
"workflow_complete": true
}
Example 2: All-in-One Request¶
Scenario: Fast-track scoring with all parameters provided¶
User: "Run loan scoring for customer 789, product 2 (Business Loan), amount 200000"
AI Response (after all steps):
{
"step": "calculate_score",
"success": true,
"customer_id": 789,
"customer_name": "Bob Johnson",
"product": "Business Loan",
"loan_amount": 200000,
"score_result": {
"score": 650,
"risk_level": "MEDIUM",
"recommendation": "REVIEW",
"breakdown": {
"created_on": "2026-02-02T15:45:00",
"risk_level": "MEDIUM"
}
},
"message": "Loan scoring completed. Score: 650 (MEDIUM risk). Recommendation: REVIEW - Additional documentation recommended",
"workflow_complete": true
}
Example 3: Error Scenario - Customer Not Found¶
Scenario: User provides invalid customer ID¶
User: "Do loan scoring for customer 99999"
AI Response (Step 1):
{
"step": "search_customer",
"error": "Customer with ID 99999 not found",
"prompt": "Enter a valid customer ID",
"suggestion": "You can search for customers by name using list_customers"
}
User: "Search for customers named John"
AI: (calls list_customers)
Found 3 customers:
- John Doe (ID: 123)
- John Smith (ID: 456)
- Johnny Brown (ID: 789)
User: "Use customer 123"
AI Response:
{
"step": "search_customer",
"success": true,
"customer_id": 123,
"customer_name": "John Doe",
"message": "Customer found: John Doe. Now select a loan product.",
"next_step": "select_product"
}
Example 4: Error Scenario - Blacklisted Customer¶
Scenario: Customer is on blacklist¶
User: "Check loan scoring for customer 555"
AI Response (Step 1):
{
"step": "search_customer",
"error": "Customer 555 is blacklisted and cannot apply for loans",
"prompt": "Enter a different customer ID or contact support",
"details": {
"blacklist_reason": "Loan default",
"blacklist_date": "2025-06-15",
"status": "active"
}
}
User: "Can we check if they're eligible for removal?"
AI: "You'll need to contact the compliance department.
Check the blacklist_status of customer 555 for more details,
or use get_blacklist_status tool for full information."
Example 5: Error Scenario - Amount Out of Range¶
Scenario: Requested amount exceeds product limits¶
User: "They want to borrow 1000000"
AI Response (Step 3):
{
"step": "enter_amount",
"error": "Loan amount must be between 10000 and 500000",
"prompt": "Enter a valid loan amount within the product limits",
"product_limits": {
"product_name": "Personal Loan",
"minimum": 10000,
"maximum": 500000
}
}
User: "What about 400000?"
AI Response:
{
"step": "enter_amount",
"success": true,
"customer_id": 456,
"product_id": 1,
"loan_amount": 400000,
"message": "Loan amount 400000 confirmed. Now calculating loan score...",
"next_step": "calculate_score"
}
Example 6: Multiple Customers Sequential Scoring¶
Scenario: Staff needs to score multiple customer applications¶
User: "Score customer 111 for product 1, amount 50000"
AI: Completes workflow for customer 111
Score: 700 (LOW) → APPROVED
User: "Now score customer 222 for product 2, amount 150000"
AI: Completes workflow for customer 222
Score: 580 (MEDIUM) → REVIEW
User: "And customer 333 for product 3, amount 300000"
AI: Completes workflow for customer 333
Score: 620 (MEDIUM) → REVIEW
User: "Summarize all three"
AI:
Loan Scoring Summary:
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
1. Customer 111 - 50,000 Personal Loan
Score: 700 (LOW) → APPROVED ✓
2. Customer 222 - 150,000 Business Loan
Score: 580 (MEDIUM) → REVIEW ⚠
3. Customer 333 - 300,000 Home Loan
Score: 620 (MEDIUM) → REVIEW ⚠
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Example 7: Score Investigation¶
Scenario: Need to understand why a score is low¶
User: "The score of 450 for customer 777 seems low. What's the issue?"
AI: "The LOW score (450, HIGH RISK) is likely due to:
Let me check the score history for customer 777:
(calls get_loan_score_history)
Recent Scores:
• 6 months ago: 580 (MEDIUM)
• 3 months ago: 520 (MEDIUM)
• Today: 450 (HIGH)
Possible factors:
1. Recent loan defaults or missed payments
2. Increased debt exposure
3. Change in income status
Recommendation: Review customer's recent transaction history
and request updated documentation before proceeding."
User: "Check their recent transactions and blacklist status"
AI: (calls get_transaction_by_id, list_transactions, get_blacklist_status)
Results show:
- Multiple overdue loan payments in last 3 months
- High transaction volume (possible financial stress)
- NOT currently blacklisted but at risk
Example 8: Product Comparison¶
Scenario: Customer eligible for multiple products¶
User: "Can we compare products for customer 444 with amount 80000?"
AI: "Let me score customer 444 with different products:
Product 1: Personal Loan (12%)
Score: 750 (LOW) → APPROVED ✓
Monthly: ~7,200 over 12 months
Product 2: Business Loan (15%)
Score: 680 (LOW) → APPROVED ✓
Monthly: ~8,000 over 12 months
Product 3: Home Loan (8%)
Score: 790 (LOW) → APPROVED ✓✓ (Best)
Monthly: ~6,800 over 12 months
Recommendation: Home Loan offers best terms and highest approval score"
Example 9: Bulk Scoring Report¶
Scenario: Generate report for multiple customer applications¶
User: "I have 5 loan applications to process. Can you score them all?"
AI: "I'll score each application. Please provide customer IDs and amounts:
Provide data like:
1. Customer 111, 50,000
2. Customer 222, 75,000
3. Customer 333, 100,000
4. Customer 444, 80,000
5. Customer 555, 120,000"
User: (provides list)
AI: (Runs workflow for each, times vary by scoring complexity)
Processing Results:
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
✓ Customer 111 - Score: 750 (LOW) - APPROVED
✓ Customer 222 - Score: 680 (LOW) - APPROVED
⚠ Customer 333 - Score: 580 (MEDIUM) - REVIEW
⚠ Customer 444 - Score: 620 (MEDIUM) - REVIEW
✗ Customer 555 - Score: 420 (HIGH) - REJECT
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Summary:
- 2 Approved
- 2 Require Review
- 1 Rejected
Next Steps for Review Cases: Request additional documentation"
Example 10: Workflow with Interruption & Resume¶
Scenario: User needs to check something mid-workflow¶
User: "Score customer 666 with product 1 for 120000"
AI: Starts workflow, completes Step 1-3
User: "Wait, let me verify that product. What's the current rate?"
AI: (answers the question)
"Product 1 (Personal Loan) has an interest rate of 12% per annum"
User: "OK, proceed with scoring"
AI: (continues with Step 4, calculating score)
Score: 710 (LOW) → APPROVED ✓
Common Workflow Patterns¶
Pattern 1: Quick Approval Check¶
User: "Will customer 100 get approved for 60000?"
AI: (runs full workflow)
Result: Score 780 (LOW) → APPROVED ✓
Pattern 2: Risk Assessment¶
User: "Is customer 200 high risk?"
AI: (runs workflow)
Result: Score 420 (HIGH) → REJECT ✗ (Yes, high risk)
Pattern 3: Maximum Loan Amount¶
User: "What's the max customer 300 can borrow?"
AI: "Based on score of 750, the product max is 500,000.
Recommend: 350,000-400,000 for better terms"
Pattern 4: Comparative Scoring¶
User: "Which product is best for customer 400?"
AI: (scores with multiple products, recommends best)
Tips for Effective Usage¶
- Have Data Ready: Have customer ID and desired amount before starting
- Verify Information: Always confirm customer name and amount
- Check Limits First: Review product limits before entering amount
- Understand Risk Levels: Know what each score range means
- Use Score History: Check past scores for trends
- Review Blacklist: Always check blacklist status for high-risk
- Document Decisions: Keep notes on why you approve/reject
- Follow Up: For REVIEW cases, get additional documentation
These examples demonstrate the flexibility and power of the loan scoring workflow in various real-world scenarios.