With AI at the Wheel, Allcargo Steers Logistics into the Future

The Challenge

Allcargo Logistics is a leading integrated logistics company headquartered in Mumbai, India. Founded in 1993, the company operates across 180+ countries offering a comprehensive range of logistics solutions including less than container load (LCL) consolidation, full container load (FCL) forwarding, air freight services, import and export handling and specialised cargo handling services.

“For optimising logistics operations, there is a pressing need to forecast demand across various routes, enabling more efficient planning. Additionally, implementing real-time price prediction can enhance decision-making and responsiveness in dynamic market conditions. Furthermore, establishing a reliable proof of delivery system is crucial for improving cash collection outcomes, ensuring that transactions are completed smoothly and efficiently. To address these challenges, Allcargo opted to harness the capabilities of artificial intelligence,” says Kapil Mahajan, Group and Global CIO, Allcargo Logistics. 

LCL Capacity Management

As a major LCL logistics provider, Allcargo faced the challenge of efficiently managing its container capacity across various trade lanes.

The company pre-commits to container capacity on different trade lanes, such as China to Singapore, to ensure availability for its customers. This committed capacity needs to be filled by consolidating smaller shipments from various customers.

Accurately predicting the demand for container capacity is crucial to avoid under-utilisation or over-commitment. Factors like seasonal trends, global events, and supply chain dynamics can significantly impact the company’s ability to fill the committed capacity.

The company leveraged demand and forecasting models and advanced analytics to address these challenges and optimise its LCL capacity management.

By analysing historical data, the AI models identified patterns and trends in customer demand, accounting for seasonal fluctuations and global events. This enabled the company to make more accurate capacity commitments and better predict the volume of shipments that could be consolidated to fill the containers.

The AI-powered models also support dynamic pricing and spot rate offerings, allowing the company to quickly adjust rates based on real-time demand and capacity utilisation. This has helped the company recover costs and maintain profitability, even in scenarios where the committed capacity cannot be fully utilised.

“We continuously refine the AI models and integrate new data sources to enhance the accuracy of capacity forecasting and pricing strategies. Such a data-driven approach enables the company to optimise its LCL operations, reduce excess capacity, and improve overall profitability,” says Mahajan.

Optimising Network Efficiency

The logistics network at Allcargo operated like an airline’s route system, with multiple pickup and delivery points. The company had around 1,500 daily route departures, with some routes not being fully utilised on the return leg, especially in consumption-heavy regions like the Northeast region of the country.

Mahajan analysed historical data using machine learning and big data to identify opportunities to streamline the network and reduce the number of daily departures while maintaining the same overall capacity.

“The AI-driven optimisation model helped the company reduce daily departures by 8-10 per cent without impacting service levels. The recommendations were thoroughly validated by the company’s own experts to ensure the model was not biased and the changes were feasible,” says Mahajan.

IBRS advocates companies to organise a team that will conduct independent quality assurance on training data and its results. This will include data scientists and engineers and data labellers who can contribute their expertise to predictive models.

After consolidating routes and adjusting touchpoints, the company was able to save on fuel, time, and transportation costs while maintaining its service level agreements.

Improving Cash Flow

The company implemented computer vision projects to automate the proof of delivery (POD) process for end-customer deliveries.

The POD documents often had varying signatures and stamps, making it difficult to verify them manually. The manual process of reviewing POD images and approving the delivery was causing delays in the cash flow cycle as customers would not pay until they received the physical POD, impacting the company’s outstanding receivables.

A traditional training model would not be able to handle this variability effectively. The solution needed to work offline, as internet connectivity could be unreliable in some delivery locations.

Mahajan and his team developed a neural network-based model that could accurately detect and validate signatures and stamps, even with variations.

“The model was deployed as an edge computing solution on the company’s mobile app, allowing it to function offline without relying on cloud-based services. The automated POD verification process enabled instant approval, improving cash flow by reducing outstanding receivables by 15-20 per cent,” he says.

IBRS has highlighted the advantages of utilising smaller, specialised machine learning models and large language models (LLMs) for enterprise applications. These benefits include the ability to operate on edge devices, reduced costs, a smaller hardware footprint, and the flexibility to fine-tune models for specific tasks.

IBRS believes AI’s impact on the future of work will be profound, ushering in new forms of automation that will fundamentally alter how tasks are performed and how people allocate their time. This transformation will unfold over the coming decade. The key to effectively navigating this change is to consistently evaluate our work processes, asking whether they could be performed better, faster, and cost-effectively.