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

	Facebook
	Twitter
	Linkedin
	E-Mail

Generative AI > Google Gemini API > Cash Flow Forecasting Agent

Cash Flow Forecasting Agent

Author: Venkata Sudhakar

ShopMax India's finance team needs to know three weeks in advance whether cash will be tight. Unexpected shortfalls delay vendor payments, attract late fees, and damage supplier relationships. A cash flow forecasting agent that reads confirmed inflows, scheduled outflows, and seasonal patterns produces a 30-day forward view that lets the team act before a problem becomes a crisis.

This tutorial builds a Cash Flow Forecasting Agent using ADK and Gemini. The agent aggregates receivables, payables, and fixed costs, projects week-by-week cash positions, and flags any week where the closing balance falls below the minimum operating buffer.

The below example shows a 30-day cash flow forecast for ShopMax India.

from google.adk.agents import LlmAgent
from google.adk.tools import FunctionTool
from google.adk.runners import Runner
from google.adk.sessions import InMemorySessionService

# --- Financial data (amounts in Rs lakhs) ---
OPENING_BALANCE = 42.0  # Rs 42 lakhs
MIN_OPERATING_BUFFER = 15.0  # Alert if balance drops below Rs 15L

INFLOWS = [
    {"week": 1, "source": "Online sales collections", "amount": 38.0},
    {"week": 1, "source": "B2B invoice payment - TechCorp", "amount": 12.0},
    {"week": 2, "source": "Online sales collections", "amount": 35.0},
    {"week": 3, "source": "Online sales collections", "amount": 40.0},
    {"week": 3, "source": "B2B invoice payment - InfoSys Mumbai", "amount": 18.0},
    {"week": 4, "source": "Online sales collections", "amount": 32.0},
]

OUTFLOWS = [
    {"week": 1, "desc": "Samsung India payment (PO-2025-001)", "amount": 27.2},
    {"week": 1, "desc": "Staff salaries", "amount": 18.5},
    {"week": 2, "desc": "Sony India payment (PO-2025-002)", "amount": 7.1},
    {"week": 2, "desc": "Warehouse rent - Mumbai", "amount": 4.2},
    {"week": 2, "desc": "Logistics partner payment", "amount": 9.8},
    {"week": 3, "desc": "Apple India payment (PO-2025-003)", "amount": 48.5},
    {"week": 4, "desc": "GST payment to GSTN", "amount": 11.3},
    {"week": 4, "desc": "Staff salaries", "amount": 18.5},
]

# --- Tool: build cash flow forecast ---
def build_cash_flow_forecast() -> dict:
    """Build a week-by-week 30-day cash flow forecast.
    Returns:
        Weekly cash positions with opening, net, and closing balances
    """
    balance = OPENING_BALANCE
    weeks = []
    alerts = []
    for week_num in range(1, 5):
        week_in = sum(i["amount"] for i in INFLOWS if i["week"] == week_num)
        week_out = sum(o["amount"] for o in OUTFLOWS if o["week"] == week_num)
        net = week_in - week_out
        closing = round(balance + net, 1)
        week_data = {
            "week": week_num,
            "opening_balance": round(balance, 1),
            "total_inflows": round(week_in, 1),
            "total_outflows": round(week_out, 1),
            "net": round(net, 1),
            "closing_balance": closing,
            "alert": closing < MIN_OPERATING_BUFFER,
        }
        if closing < MIN_OPERATING_BUFFER:
            alerts.append(f"Week {week_num}: closing balance Rs {closing}L is below Rs {MIN_OPERATING_BUFFER}L minimum")
        weeks.append(week_data)
        balance = closing
    return {
        "opening_balance": OPENING_BALANCE,
        "min_buffer": MIN_OPERATING_BUFFER,
        "weeks": weeks,
        "alerts": alerts,
        "month_end_balance": balance,
    }

# --- Agent setup ---
forecast_tool = FunctionTool(func=build_cash_flow_forecast)

agent = LlmAgent(
    name="cashflow_agent",
    model="gemini-2.0-flash",
    instruction="""You are a cash flow forecasting agent for ShopMax India.
Call build_cash_flow_forecast to get the 30-day projection.
Present a weekly table: Week | Opening (Rs L) | Inflows | Outflows | Net | Closing | Status.
For any week where closing balance falls below the minimum buffer:
  - Flag it as SHORTFALL RISK
  - Identify the largest single outflow that week
  - Suggest an action: delay a non-critical payment, draw on credit line, or accelerate a receivable.
End with the month-end position summary.""",
    tools=[forecast_tool],
)

session_service = InMemorySessionService()
runner = Runner(agent=agent, app_name="cashflow", session_service=session_service)

import asyncio

async def run_forecast(query: str) -> str:
    session = await session_service.create_session(app_name="cashflow", user_id="finance")
    from google.genai.types import Content, Part
    msg = Content(role="user", parts=[Part(text=query)])
    result = ""
    async for event in runner.run_async(user_id="finance", session_id=session.id, new_message=msg):
        if event.is_final_response() and event.content:
            result = event.content.parts[0].text
    return result

response = asyncio.run(run_forecast("Give me the 30-day cash flow forecast and flag any weeks we might have a cash problem."))
print(response)

It gives the following output,

Cash Flow Forecast - ShopMax India (30-Day View)
Opening Balance: Rs 42.0L  |  Minimum Buffer: Rs 15.0L

Week | Opening | Inflows | Outflows | Net    | Closing | Status
-----|---------|---------|----------|--------|---------|--------
  1  |  42.0   |  50.0   |  45.7    | +4.3   |  46.3   | OK
  2  |  46.3   |  35.0   |  21.1    | +13.9  |  60.2   | OK
  3  |  60.2   |  58.0   |  48.5    | +9.5   |  69.7   | OK
  4  |  69.7   |  32.0   |  29.8    | +2.2   |  71.9   | OK

Month-End Balance: Rs 71.9L

No shortfall alerts this month. Cash position is healthy throughout.

Notable: Week 3 has the largest single outflow - Apple India payment of
Rs 48.5L. Ensure the B2B receivable from InfoSys Mumbai (Rs 18.0L) is
collected by Week 3 as planned to maintain comfortable headroom.

Connect this agent to your accounting system via BigQuery for live receivables and payables data. Run it weekly via Cloud Scheduler and deliver the report to the CFO via email. Add a Pub/Sub alert that triggers immediately if any projected closing balance drops below the buffer, giving the finance team maximum lead time to act.

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