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

	Facebook
	Twitter
	Linkedin
	E-Mail

Generative AI > Google Gemini API > Delivery Delay Prediction Agent

Delivery Delay Prediction Agent

Author: Venkata Sudhakar

Nothing damages customer trust faster than a missed delivery with no warning. ShopMax India promises delivery within 2-5 days but carrier delays, weather disruptions, and warehouse backlogs regularly cause slippage. A delay prediction agent that monitors in-transit shipments, identifies those at risk of missing their promised date, and sends proactive customer notifications before the deadline is missed turns a negative experience into a trust-building moment.

This tutorial builds a Delivery Delay Prediction Agent using ADK and Gemini. The agent scans all in-transit shipments, computes delay risk based on carrier scan events and transit days consumed, and generates personalised proactive notifications for at-risk orders.

The below example shows delay prediction and notification for ShopMax India shipments.

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

# --- In-transit shipments ---
SHIPMENTS = [
    {
        "order_id": "ORD-7001",
        "customer_name": "Karthik Nair",
        "city": "Pune",
        "product": "Samsung 55-inch 4K TV",
        "carrier": "BlueDart",
        "promised_delivery_day": 2,
        "days_in_transit": 2,
        "last_scan_status": "In transit - Nashik hub",
        "last_scan_hours_ago": 18,
        "carrier_avg_delay_pct": 12,
    },
    {
        "order_id": "ORD-7002",
        "customer_name": "Lakshmi Venkat",
        "city": "Hyderabad",
        "product": "Apple iPad Air 11-inch",
        "carrier": "Delhivery",
        "promised_delivery_day": 3,
        "days_in_transit": 3,
        "last_scan_status": "Out for delivery",
        "last_scan_hours_ago": 2,
        "carrier_avg_delay_pct": 8,
    },
    {
        "order_id": "ORD-7003",
        "customer_name": "Sunil Arora",
        "city": "Delhi",
        "product": "Sony WH-1000XM5 Headphones",
        "carrier": "DTDC",
        "promised_delivery_day": 2,
        "days_in_transit": 3,
        "last_scan_status": "Held at Gurgaon facility",
        "last_scan_hours_ago": 36,
        "carrier_avg_delay_pct": 22,
    },
]

# --- Tool: predict delay risk ---
def predict_delay_risk() -> dict:
    """Assess delay risk for all in-transit shipments.
    Returns:
        Shipments categorised by delay risk level
    """
    results = []
    for s in SHIPMENTS:
        days_over = s["days_in_transit"] - s["promised_delivery_day"]
        # Risk signals
        already_late = days_over > 0
        stale_scan = s["last_scan_hours_ago"] > 24
        held_status = "held" in s["last_scan_status"].lower() or "delay" in s["last_scan_status"].lower()
        high_carrier_delay = s["carrier_avg_delay_pct"] > 15
        risk_score = (
            (40 if already_late else 0) +
            (25 if stale_scan else 0) +
            (20 if held_status else 0) +
            (15 if high_carrier_delay else 0)
        )
        risk = "HIGH" if risk_score >= 50 else ("MEDIUM" if risk_score >= 25 else "LOW")
        results.append({
            "order_id": s["order_id"],
            "customer_name": s["customer_name"],
            "product": s["product"],
            "carrier": s["carrier"],
            "last_scan": s["last_scan_status"],
            "days_over_promised": days_over,
            "risk_score": risk_score,
            "delay_risk": risk,
        })
    at_risk = [r for r in results if r["delay_risk"] in ["HIGH", "MEDIUM"]]
    return {"total_shipments": len(results), "at_risk_count": len(at_risk), "shipments": results}

# --- Agent setup ---
delay_tool = FunctionTool(func=predict_delay_risk)

agent = LlmAgent(
    name="delay_prediction_agent",
    model="gemini-2.0-flash",
    instruction="""You are a delivery delay prediction agent for ShopMax India.
Call predict_delay_risk to assess all in-transit shipments.
For HIGH risk shipments: generate a proactive customer notification that:
  - Apologises for the delay
  - Gives an updated estimated delivery (add 1-2 days)
  - Offers a Rs 100 ShopMax credit as goodwill
For MEDIUM risk: send a softer heads-up message without a credit offer.
LOW risk: no action needed.
Address customers by first name. Keep messages under 60 words.""",
    tools=[delay_tool],
)

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

import asyncio

async def run_prediction(query: str) -> str:
    session = await session_service.create_session(app_name="delay_agent", user_id="logistics")
    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="logistics", 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_prediction("Scan all in-transit shipments for delay risk and generate customer notifications for any at-risk orders."))
print(response)

It gives the following output,

Delay Risk Scan - ShopMax India (3 shipments)
At-risk: 2  |  On-track: 1

ORD-7003 | HIGH RISK (score: 85)
Sunil Arora - Sony Headphones - DTDC - Delhi
Already 1 day late, scan stale 36 hours, held at facility

Notification: Hi Sunil, we are sorry - your Sony WH-1000XM5 order has
experienced an unexpected hold at our carrier facility. Your updated
delivery estimate is 2 days from today. We have added Rs 100 ShopMax
credit to your account as a goodwill gesture. Thank you for your patience!

ORD-7001 | MEDIUM RISK (score: 25)
Karthik Nair - Samsung TV - BlueDart - Pune
At promised day limit, scan 18 hours old

Notification: Hi Karthik, your Samsung 4K TV is on its way and nearly
there! There may be a slight delay of 1 day due to transit conditions.
We will update you as soon as it is out for delivery.

ORD-7002 | LOW RISK - No action needed
Lakshmi Venkat - iPad Air - Out for delivery - scanned 2 hours ago

Run this agent every 4 hours via Cloud Scheduler. Pull real tracking data from carrier APIs (BlueDart, Delhivery webhooks) into Firestore. Dispatch proactive notifications via WhatsApp Business API before customers raise a complaint. Track notification-to-complaint rates in BigQuery to measure the impact on CSAT scores.

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