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

	Facebook
	Twitter
	Linkedin
	E-Mail

Generative AI > Google Gemini API > Payroll Anomaly Agent

Payroll Anomaly Agent

Author: Venkata Sudhakar

Payroll errors cost Indian companies millions of rupees annually through duplicate payments, miscalculated overtime, and unauthorised salary changes. For ShopMax India's 2,000-person payroll, manual auditing is impractical. An anomaly detection agent can scan every payroll run and surface exceptions before payment is released.

This tutorial builds a Gemini ADK agent that compares the current payroll run against the previous month's baseline, flags salary changes above a threshold, identifies duplicate employee entries, and highlights overtime claims that exceed policy limits.

The below example shows a payroll anomaly detection agent for ShopMax India in a business context.

import asyncio
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 import types

# Previous month payroll baseline
PREV_PAYROLL = {
    "EMP-301": {"name": "Suresh Joshi",  "gross_rs": 65000, "overtime_hrs": 4},
    "EMP-302": {"name": "Meena Pillai",  "gross_rs": 48000, "overtime_hrs": 0},
    "EMP-303": {"name": "Vikram Das",    "gross_rs": 92000, "overtime_hrs": 2},
    "EMP-304": {"name": "Ananya Singh",  "gross_rs": 38000, "overtime_hrs": 6},
}

# Current month payroll run
CURR_PAYROLL = [
    {"emp_id": "EMP-301", "name": "Suresh Joshi",  "gross_rs": 65000, "overtime_hrs": 4},
    {"emp_id": "EMP-302", "name": "Meena Pillai",  "gross_rs": 58000, "overtime_hrs": 0},
    {"emp_id": "EMP-302", "name": "Meena Pillai",  "gross_rs": 58000, "overtime_hrs": 0},
    {"emp_id": "EMP-303", "name": "Vikram Das",    "gross_rs": 92000, "overtime_hrs": 38},
    {"emp_id": "EMP-304", "name": "Ananya Singh",  "gross_rs": 38000, "overtime_hrs": 6},
]

SALARY_CHANGE_THRESHOLD_PCT = 10  # flag changes above 10%
MAX_OVERTIME_HRS = 20             # policy limit per month

def detect_payroll_anomalies() -> dict:
    anomalies = []
    seen_ids = {}
    for row in CURR_PAYROLL:
        eid = row["emp_id"]
        # Duplicate check
        if eid in seen_ids:
            anomalies.append({"type": "DUPLICATE", "emp_id": eid, "name": row["name"],
                              "detail": "Duplicate payroll entry detected"})
            continue
        seen_ids[eid] = True
        prev = PREV_PAYROLL.get(eid)
        if not prev:
            anomalies.append({"type": "NEW_EMPLOYEE", "emp_id": eid, "name": row["name"],
                              "detail": "No prior month record - verify is genuine new hire"})
            continue
        # Salary change check
        change_pct = abs(row["gross_rs"] - prev["gross_rs"]) / prev["gross_rs"] * 100
        if change_pct > SALARY_CHANGE_THRESHOLD_PCT:
            anomalies.append({"type": "SALARY_CHANGE", "emp_id": eid, "name": row["name"],
                              "detail": f"Salary changed by {round(change_pct,1)}% (Rs {prev['gross_rs']} -> Rs {row['gross_rs']})"})
        # Overtime check
        if row["overtime_hrs"] > MAX_OVERTIME_HRS:
            anomalies.append({"type": "OVERTIME_EXCESS", "emp_id": eid, "name": row["name"],
                              "detail": f"Overtime {row['overtime_hrs']} hrs exceeds policy limit of {MAX_OVERTIME_HRS} hrs"})
    return {
        "payroll_month": "April 2026",
        "total_entries": len(CURR_PAYROLL),
        "anomalies_found": len(anomalies),
        "anomalies": anomalies,
        "status": "HOLD FOR REVIEW" if anomalies else "CLEAR FOR PAYMENT",
    }

payroll_tool = FunctionTool(func=detect_payroll_anomalies)

agent = LlmAgent(
    name="payroll_anomaly_agent",
    model="gemini-2.0-flash",
    description="Detects anomalies in ShopMax India payroll runs before payment release.",
    instruction=(
        "You are a payroll audit agent for ShopMax India. "
        "When asked to audit the payroll, call detect_payroll_anomalies. "
        "Present each anomaly clearly with type, employee, and recommended action. "
        "State whether the payroll should be held for review or cleared for payment."
    ),
    tools=[payroll_tool],
)

async def main():
    session_service = InMemorySessionService()
    runner = Runner(agent=agent, app_name="payroll_app", session_service=session_service)
    session = await session_service.create_session(app_name="payroll_app", user_id="finance1")
    msg = types.Content(role="user", parts=[types.Part(
        text="Audit the April 2026 payroll run for anomalies"
    )])
    async for event in runner.run_async(user_id="finance1", session_id=session.id, new_message=msg):
        if event.is_final_response() and event.content:
            print(event.content.parts[0].text)

asyncio.run(main())

It gives the following output,

Payroll Audit Report - April 2026 | ShopMax India

Total Entries : 5  |  Anomalies Found: 3
Status        : HOLD FOR REVIEW

[DUPLICATE] EMP-302 - Meena Pillai
  Duplicate payroll entry detected.
  Action: Remove one entry before payment. Investigate data entry error.

[SALARY_CHANGE] EMP-302 - Meena Pillai
  Salary changed by 20.8% (Rs 48,000 -> Rs 58,000)
  Action: Verify HR-approved increment letter before processing.

[OVERTIME_EXCESS] EMP-303 - Vikram Das
  Overtime 38 hrs exceeds policy limit of 20 hrs.
  Action: Obtain manager and HR approval for excess overtime.

Payroll is on HOLD. Resolve all 3 anomalies before releasing payment.

ShopMax India can run this agent automatically each month when the payroll file is uploaded to the HR system. Anomalies are routed to the payroll manager and the relevant department head for sign-off before the payment batch is released, preventing overpayments and duplicate transactions from reaching employee bank accounts.

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