Revolutionize Supply Chain Operations with Generative AI powered Smart Demand Forecasting Solution
This blog explores how Generative AI powered demand forecasting systems analyze complex patterns, enterprise data, and market trends using agentic workflows to deliver unprecedented accuracy in forecasting demand.
The automotive industry is undergoing rapid transformation, with the global transition to clean energy driving a surge in demand for electric vehicle (EV) batteries. While the growth of battery demand and production offers significant environmental and social advantages, numerous challenges lie ahead for both producers and purchasers. In 2025, companies are expected to face continued supply chain risks, disruptions, potential delays, and elevated costs due to geo-political factors such as tariffs, transportation, labour challenges and more throughout the value chain.
In a recent study, 78% of manufacturers indicated that they have implemented or are planning to invest in supply chain planning software. Manufacturers traditionally invested in on-prem, hybrid cloud, and IIoT 4.0 technologies to unlock value in end-to-end manufacturing operations. However, many of these initiatives failed to deliver impact due to various factors such as:
Fragmented data sources: Structured, semi-structured, and unstructured data across ERP, CRM, and supply chain management systems.
Heterogeneous data formats such as semi-structured and unstructured data (image, audio, video), and siloed business processes.
Lack of real-time Insights as traditional forecasting models fail to incorporate external macroeconomic and environmental factors effectively.
Although this blog focuses on the supply chain challenges for EV batteries, the proposed solution equally applies to other manufactured products as well.
Gen AI Opportunity in Supply Chain
According to a study by McKinsey, Generative AI opens a new frontier to accelerate, augment, and automate manufacturing and supply chain operations that was never possible before. By leveraging generative AI, companies can now achieve unprecedented accuracy in predicting battery demand patterns, enabling seamless coordination between supply chain stakeholders. This technological integration delivers impressive business outcomes: maximized revenue through smart inventory management, stronger supplier relationships, enhanced customer satisfaction, and improved strategic planning capabilities. To help the automotive company develop a robust supply chain strategy to meet its 60% EV target by 2028, the value chain needs to address numerous environmental, social and governance challenges as depicted below.
Figure 1
Use Case : Gen AI driven Demand Forecasting
Supply chain management (SCM) encompasses a series of interconnected steps designed to efficiently produce and deliver vehicles and components such as batteries. This blog will specifically focus on the Analysis & Forecasting challenges (Figure 2). This is a critical component within supply chain management, enabling manufacturers to predict future demand for products and services. Accurate forecasts are essential for maintaining optimal inventory levels, efficient resource allocation, and enhanced customer satisfaction. Forecasting future demand is inherently complex due to unpredictable factors such as geopolitics, weather conditions, shipping route disruptions, and global events. Generative AI can be a game changer as Supply chain analysts can generate near real-time insights more effectively than ever by establishing a robust data foundation that integrates internal and external data sources—including macroeconomic indicators and environmental conditions.
Figure 2
Solution Overview
With that background, let's dive deep into the solution now. Karini AI’s smart demand forecasting solution built on Amazon Bedrock can integrate seamlessly with diverse enterprise systems. For example, Karini AI offers built-in data connectors for diverse data sources, such as –
Amazon S3, Amazon Q Retriever Index
Enterprise systems such as ERP, CRM
Web sources for market trends, weather, geopolitical factors, social media sentiments etc.
Using the sophisticated Agent 2.0 framework, customers can effortlessly build advanced and complex generative AI applications. These comprehensive capabilities can enable automaker & manufacturers to optimize inventory, enhance supply chain resilience, and scale operations efficiently.
Figure 3
Data set:
This smart demand forecasting solution leverages a comprehensive set of synthetic datasets generated by Karini AI to provide deep insights into product order patterns and supply chain dynamics. This solution ensures that each forecast not only reflects historical data but also aligns with broader economic, environmental, and geopolitical dynamics, empowering businesses to make well-informed decisions amidst complex global challenges.
The dataset includes the following –
Detailed historical product orders and supplier data, such as product codes, categories, suppliers, purchase dates, and demand levels.
Specific supplier and supply chain details, including pricing, locations, availability, product sales volume, shipping carriers, costs, and transportation modes.
Time series forecasting capabilities of the Nixtla. This is a Generative pretrained transformer for time series trained on over 100B data points and claims to accurately predict forecasts across various industry domains.
Karini AI’s Smart Demand Forecasting solution leverages a suite of specialized agents, each powered by Amazon Bedrock models, tailored to enhance various business operations from demand forecasting to order management. Following diagram shows the Agent architecture in Karini AI no-code platform.
Figure 6
These agents employ sophisticated tools to streamline tasks and boost operational efficiency.
Supervisor agent: The Supervisor agent plays a pivotal role in directing user requests to the appropriate specialist agents. Utilizing the Amazon Bedrock Claude Sonnet 3.7 model and Routing tools, this agent not only forecasts product demand by analyzing historical sales data, market trends, and supplier details but also assesses risks influenced by global events and climate factors. Furthermore, it handles order placements and sends confirmation notifications.
Demand Forecast agent: Focused on demand forecasting, this agent analyzes historical sales data and translates user inquiries into actionable data for demand forecasting. It adeptly translates user inquiries into actionable data for the Nixtla client, which is triggered via an AWS Lambda function. The agent also translates AWS Lambda outputs for user comprehension, and creates visual data representations.
Model: Amazon Bedrock Claude Sonnet 3.5
Tools:
AWS Lambda: Invoked to run the Nixtla client, generating demand prediction reports that aid inventory planning and sales strategy, especially in the retail and supply chain sectors.
Supply chain agent: This agent retrieves supplier information from the AWS PostgreSQL supply chain database. It's instrumental for supply chain management and vendor analysis, providing procurement teams with essential supplier data to aid in efficient decision-making.
Model: Amazon Bedrock Claude Sonnet 3.5
Tools:
Database: Utilizes AWS PostgreSQL to query the supply chain database, providing essential supplier data to facilitate efficient procurement decisions.
Logistics, Delivery and Risk management agent: Responsible for analyzing delivery and logistics risks, this agent combines data from various sources to provide a comprehensive risk assessment.
Model: Amazon Bedrock Claude Sonnet 3.5
Tools:
Knowledge Base: This tool retrieves relevant electric vehicle (EV) market information from a knowledge base. Useful for analyzing trends, statistics, and insights in the EV industry.
REST API - Performs web searches to assess risks and impacts on supply chain delivery and fulfillment.
Order management agent: This agent handles the end-to-end order placement process. It gathers order details, creates purchase orders, updates databases, and sends email confirmations through AWS Lambda and email messaging tools. This agent is essential for automating procurement processes, ensuring efficient order processing and direct communication with stakeholders.
Model: Amazon Bedrock Claude Sonnet 3.5
Tools:
AWS lambda: This tool processes component-related information to place an order. It takes component details as input, executes the order placement, and returns the S3 path where the order information is stored. Primarily used for automated inventory management and order processing systems.
Messaging (Email): This tool sends an email containing a purchase order. Ideal for automating order confirmations and communications, ensuring recipients receive purchase order information directly in their email inbox.
Check the demo video here on the end to end workflow for the above solution.
Try it Yourself:
To sign up for Karini AI, please reach out to support@karini.ai.
You can access the source code, including prompts, recipes, and agents, in this source code repository.
Conclusion:
Automakers and manufacturers face significant challenges in EV battery supply chain management, from inventory optimization to risk mitigation, as they push toward ambitious EV production targets. To address these challenges, a Generative AI-powered demand forecasting solution can help increase the forecast accuracy and strengthen supply chain resilience. Built on Amazon Bedrock and utilizing Karini AI’s agentic workflow, the platform seamlessly integrates internal and external data sources with time series forecasting capabilities. This comprehensive solution may enable manufacturers to make data-driven decisions while efficiently managing their supply chain operations.
