Using AI & WMS and the Wider Supply Chain

Using AI & WMS and the Wider Supply Chain

Recent developments in artificial intelligence (AI) have created unprecedented interest. In reality, AI has been around for years, quietly evolving and being used across an array of business applications to enable new insights and higher levels of automation. What is new is the emergence of so-called generative AI to create text, speech, and images in such a way that practically anyone can use it. According to some reports, as much as $100 billion has been invested worldwide in the technology over the past couple of years. These capabilities are inevitably being integrated into supply chain applications such as warehouse management systems (WMS).

According to some reports, as much as $100 billion has been invested worldwide in the technology over the past couple of years.

One area where AI will have a major impact is across robotics and automation. It could be argued that this falls into two broad areas. First, and perhaps most obvious, there is the automation of physical infrastructure such as robotics, conveyors, lift trucks, or automated storage and retrieval systems. Previously, these technologies relied on traditional – but sophisticated – control systems that utilised conventional programs and algorithms. The most advanced now have the ability to operate autonomously with little or no human intervention. While this may appear like AI, in most cases the underlying logic is based on traditional programming although there are clearly some overlaps.

However, introducing more or better AI offers the prospect of taking these capabilities to new levels, with integrated systems adapting and responding in real time to a wider range of variables and greater volumes of data which in some cases they themselves are generating. The objective is to improve productivity, performance, and efficiency across the board, for example by increasing pick accuracy and order fulfilment, reducing stock management errors, improving material throughputs, better utilisation of space and infrastructure, and reduced reliance on physical labour among others.

However, this is not just about making the best use of the facility. AI can also help keep the infrastructure working properly, for example by enabling predictive maintenance. As always, this is not strictly new: techniques such as condition monitoring have been used for decades to identify potential machinery failures before they cause catastrophic damage. Supply chain equipment from conveyor systems to lift trucks to cleaning machines already incorporate technologies to predict breakdowns before they occur so that maintenance engineers can intervene and prevent costly downtime. In fact, 47 per cent of global manufacturers use predictive maintenance technologies to reduce operational costs. AI offers the potential to predict when machinery might fail using a wider range of real-time and historic data sources with the aim of supporting 100 per cent uptime.

A recent study found that predictive maintenance can reduce maintenance costs by up to 30 per cent and decrease unplanned downtime by 70 per cent. According to Deloitte, predictive maintenance increases productivity by 25 per cent, reduces breakdowns by 70 per cent, and lowers maintenance costs by 25 per cent(3). AI also supports new forms of support and remote maintenance. Related techniques such as augmented reality (AR) and virtual reality (VR) are already used to help diagnose common points of failure, help users carry out fixes on their own, or enable remote intervention by specialist support engineers. Generative AI is going to take this type of support to even higher levels.

According to Deloitte, predictive maintenance increases productivity by 25 per cent, reduces breakdowns by 70 per cent, and lowers maintenance costs by 25 per cent(3).

Infrastructure downtime is clearly unwanted but disruption can arise from other sources. This leads to the second, and perhaps more interesting, broad area where automation, robotics, and AI are increasingly overlapping: the way data is acquired, processed, and used to deliver performance improvements and new insights. Many WMS already incorporate analytical and reporting that offer AI-like capabilities, including the Business Intelligence (BI) component of Principal Logistics Technologies’ ProWMS application. These capabilities will become more sophisticated as AI evolves and is integrated with them.

Many WMS already incorporate analytical and reporting that offer AI-like capabilities, including the Business Intelligence (BI) component of Principal Logistics Technologies’ ProWMS application.

Supply chains exist to ensure the right items are delivered when required, whether this is a retail sale or a component at the manufacturing lineside. Disruption costs money and can result in lost business. On top of this infrastructure such as buildings and transport can be expensive so it pays to optimise utilisation and reduce wastage so that the need for additional investment is minimised or eliminated. Streamlining supply with demand is critical and AI can help by analysing data such as sales, inventory, seasonality, weather reports and forecasts, and historical trends to forecasting demand more accurately. Companies using AI for demand forecasting have seen an up to 20 per cent improvement in accuracy with resulting reductions in excess stock and improved order fulfilment rates.

Inside the warehouse, AI algorithms can dynamically assign storage locations (slotting) based on product characteristics, demand, and order frequency. This ensures efficient picking and reduces travel time for warehouse staff. Reports suggest dynamic slotting can improve warehouse productivity by up to 15 per cent while reducing labour costs by 10 per cent.

More generally, AI can improve precision in inventory management by using advanced algorithms and machine learning. Real-time tracking and minimal error margins should optimise stock levels, reducing risks of over- or understocking. Some studies suggest this could reduce prediction errors and inventory overstocking by up to 50 per cent. Elsewhere, computer vision technologies can visually inspect and identify items, enabling accurate inventory tracking without manual intervention. It can also help to identify if items from individual packs to full pallets are damaged, removing the risk of delivering something that will later need to be returned.

Real-time tracking and minimal error margins should optimise stock levels, reducing risks of over- or understocking. Some studies suggest this could reduce prediction errors and inventory overstocking by up to 50 per cent.

One of the many potential applications of generative AI in the warehouse is for training. Many warehouses suffer from high staff turnover rates or rely on temporary or short-term appointments to meet demand during seasonal peaks. Training takes time and costs money but is essential in every situation, particularly if the warehouse is subject to additional compliance regulations. The challenge is to provide appropriate training that does not take too long or otherwise overburden normal warehouse operations when new staff will have different levels of expertise and experience. Generative AI offers new methods for training in these situations, for example by adapting during interactions with staff to align content with skill levels. Users can ask the WMS how to perform tasks, receiving explanations in plain language instead of pre-programmed responses.

The WMS can also provide instructions, and understand responses, in more naturalistic language. This empowers users to learn and interact with the system effectively, including in their own first language. That might avoid the need for some system-specific training, reducing the time and cost needed, but can also allow new employees to learn as they work rather than in special sessions before they start.

Once an order is ready for despatch, AI can be used in other ways. Load and route planning are notoriously complex and difficult because there are so many potential variables. This may be particularly so in online retail and last mile fulfilment operations where the number and volumes of consignments, and their destinations, will inevitably be different for every load. Objectives will vary from business to business but will generally include the need to maximise utilisation of vehicles, optimise deliveries for maximum fulfilment rates, with minimum travel routes and so on.

Existing load and transport planning applications already offer excellent results but dealing with complex variables and datasets in real time is where AI can really come into its own. Algorithms can optimise delivery routes, considering factors like traffic, weather, and delivery windows. This reduces transportation costs and enhances delivery efficiency. Studies suggest route optimisation can lead to fuel savings of up to 15 per cent and reduce delivery time by 20 per cent.

The use of advanced supply chain applications such as WMS in conjunction with conventional and generative AI is already leading to significant performance improvements and efficiency gains. It is likely that these benefits will accelerate as the underlying technologies become even more capable and application developers integrate them into more and more of their own products.