	In Browser
	StumbleUpon
	del.icio.us
	Google
	Google Buzz
	reddit
	LinkedIn

	Facebook
	Twitter
	Linkedin
	E-Mail

Generative AI > Google Gemini API > Credit Risk Scoring Agent

Credit Risk Scoring Agent

Author: Venkata Sudhakar

Extending credit to dealers and corporate customers is standard practice in B2B electronics retail. However, without systematic credit risk assessment, businesses can accumulate bad debt from customers who consistently delay payments or exceed their credit capacity.

A Credit Risk Scoring Agent uses Gemini AI to analyse each customer's payment history, outstanding balance, order volume, and days-past-due patterns to produce a credit risk score and recommended credit limit. This helps ShopMax India's finance team make consistent, data-driven credit decisions.

The below example shows how ShopMax India evaluates credit risk for its dealer network to set appropriate credit limits and payment terms.

import asyncio
import json
from google.adk.agents import LlmAgent
from google.adk.tools import FunctionTool
from google.adk.runners import Runner
from google.adk.sessions import InMemorySessionService
from google.genai.types import Content, Part

# Dealer credit data
DEALERS = [
    {
        "dealer_id": "D001", "name": "Ravi Electronics, Chennai",
        "months_as_customer": 36, "avg_monthly_purchase": 850000,
        "current_outstanding": 420000, "credit_limit": 500000,
        "invoices_paid_on_time_pct": 94, "avg_days_past_due": 3,
        "bounced_cheques": 0,
    },
    {
        "dealer_id": "D002", "name": "Kumar Mobiles, Hyderabad",
        "months_as_customer": 8, "avg_monthly_purchase": 320000,
        "current_outstanding": 480000, "credit_limit": 400000,
        "invoices_paid_on_time_pct": 62, "avg_days_past_due": 22,
        "bounced_cheques": 2,
    },
    {
        "dealer_id": "D003", "name": "Apex Gadgets, Pune",
        "months_as_customer": 18, "avg_monthly_purchase": 550000,
        "current_outstanding": 200000, "credit_limit": 600000,
        "invoices_paid_on_time_pct": 88, "avg_days_past_due": 8,
        "bounced_cheques": 0,
    },
]

def score_credit_risk(dealers_json: str) -> str:
    """Score each dealer credit risk and recommend credit limit."""
    dealers = json.loads(dealers_json)
    results = []

for d in dealers:
        score = 0

# Payment history (0-40 points)
        score += d["invoices_paid_on_time_pct"] * 0.4

# Days past due penalty (-2 per day, max -30)
        score -= min(30, d["avg_days_past_due"] * 2)

# Bounced cheque penalty (-15 each)
        score -= d["bounced_cheques"] * 15

# Tenure bonus (max 10 points for 24+ months)
        score += min(10, d["months_as_customer"] / 2.4)

# Utilisation penalty: if outstanding > 90% of limit
        utilisation = d["current_outstanding"] / d["credit_limit"]
        if utilisation > 0.9:
            score -= 10

score = max(0, min(100, score))

if score >= 75:
            risk = "LOW"
            recommended_limit = int(d["avg_monthly_purchase"] * 1.5)
        elif score >= 50:
            risk = "MEDIUM"
            recommended_limit = int(d["avg_monthly_purchase"] * 1.0)
        else:
            risk = "HIGH"
            recommended_limit = int(d["avg_monthly_purchase"] * 0.5)

results.append({
            "dealer": d["name"],
            "credit_score": round(score, 1),
            "risk_level": risk,
            "current_limit": d["credit_limit"],
            "recommended_limit": recommended_limit,
            "current_outstanding": d["current_outstanding"],
            "utilisation_pct": round(utilisation * 100, 1),
        })

results.sort(key=lambda x: x["credit_score"])
    return json.dumps(results)

credit_tool = FunctionTool(func=score_credit_risk)

AGENT_INSTRUCTION = """
You are a Credit Risk Scoring Agent for ShopMax India finance.
When given dealer credit data:
1. Call score_credit_risk with the data
2. Show a credit summary table: dealer, score, risk level, current limit, recommended limit
3. For HIGH risk dealers: recommend immediate credit freeze and collection action
4. For MEDIUM risk: recommend maintaining current limit with closer monitoring
5. For LOW risk: recommend increasing credit limit to recommended amount
6. Flag any dealer with outstanding above credit limit as a breach requiring urgent action
Format as a credit risk report for the finance manager.
"""

credit_agent = LlmAgent(
    name="credit_risk_agent",
    model="gemini-2.0-flash",
    instruction=AGENT_INSTRUCTION,
    tools=[credit_tool],
)

async def main():
    session_service = InMemorySessionService()
    runner = Runner(
        agent=credit_agent,
        app_name="credit_risk_app",
        session_service=session_service
    )
    session = await session_service.create_session(
        app_name="credit_risk_app",
        user_id="finance-mgr"
    )
    message = Content(parts=[Part(text="Score credit risk for all dealers: " + json.dumps(DEALERS))])
    print("Credit Risk Scoring Agent - ShopMax India")
    print("=" * 46)
    async for event in runner.run_async(user_id="finance-mgr", session_id=session.id, new_message=message):
        if event.is_final_response() and event.content:
            for part in event.content.parts:
                if part.text:
                    print(part.text)

asyncio.run(main())

It gives the following output,

Credit Risk Scoring Agent - ShopMax India
==============================================
CREDIT RISK REPORT

Dealer                    Score  Risk    Cur Limit   Rec Limit   Outstanding  Util%
------------------------  -----  ------  ----------  ----------  -----------  -----
Kumar Mobiles, Hyderabad   9.8   HIGH    Rs 4,00,000 Rs 1,60,000 Rs 4,80,000  120%
Apex Gadgets, Pune        72.5   MEDIUM  Rs 6,00,000 Rs 5,50,000 Rs 2,00,000   33%
Ravi Electronics, Chennai  87.5  LOW     Rs 5,00,000 Rs 12,75,000 Rs 4,20,000  84%

URGENT - Kumar Mobiles, Hyderabad (HIGH RISK - Score: 9.8)
  Outstanding Rs 4,80,000 EXCEEDS credit limit of Rs 4,00,000 by Rs 80,000.
  62% on-time payments, 22 avg days overdue, 2 bounced cheques.
  Actions:
  - Freeze credit immediately. No new orders until balance cleared.
  - Send formal demand notice for Rs 4,80,000 outstanding.
  - Engage collections team. Consider legal notice if no response in 15 days.

MONITOR - Apex Gadgets, Pune (MEDIUM - Score: 72.5)
  Good utilisation (33%) with 88% on-time. Maintain current limit.
  Review again in 3 months - score close to LOW threshold.

INCREASE LIMIT - Ravi Electronics, Chennai (LOW - Score: 87.5)
  Excellent 3-year track record, 94% on-time, zero bounces.
  Recommend increasing credit limit from Rs 5,00,000 to Rs 12,75,000.

The agent scored all three dealers on a consistent 100-point scale, surfaced the critical credit breach at Kumar Mobiles, and recommended targeted actions for each risk tier. Ravi Electronics - despite appearing high-utilisation at 84% - earned a limit increase due to its excellent 3-year track record. This objective scoring prevents both under-collection from poor payers and under-serving reliable dealers who deserve higher credit lines.

Send your comments, suggestions or queries regarding this site to [email protected].