AI and the supply chain of the future

In an economic and geopolitical landscape where uncertainty is the only certainty, pharma companies must secure all available control tools. Among these, artificial intelligence undoubtedly plays a crucial role.


According to digitisation experts, AI can play a significant role in the pharmaceutical supply chain context by contributing to optimised transport routes, real-time inventory management and more accurate and timely production control. Not surprisingly, companies such as Pfizer, Amgen, GSK, Merck and Roche already leverage artificial intelligence solutions to predict supply chain disruptions, optimise inventory levels and improve production processes.

Predictive analytics and drug shortages

One of the most common issues facing the market is drug shortages: even before Covid, 95 per cent of respondents in a survey conducted in 39 European countries cited drug shortages as an obstacle to the optimal treatment process.

In 2020, with the pandemic, a third of European countries reported a shortage of more than 400 drugs, and in a period of only 9 months, five states published a total of more than 5,000 shortage reports (Finland, Sweden, Norway, Spain and the USA).

According to a study in the USA, even there ‘current shortages are the highest in a decade’. Of course, dependence on third countries weighs on this factor, but with a view to optimising the optimisable, predictive analytics can make its contribution. For example, it can be used to predict product demand in greater detail, allowing companies to calibrate inventory systems and maintain adequate stock levels while reducing the risks of under- or over-stocking.

According to one digital service provider, the introduction of cloud-based machine learning systems from AWS (Amazon’s web services) enabled a big pharma to reduce its forecasting error by about 15%, saving 2.5% in costs and increasing revenues by 1.5% (about USD 600 million).

In its simplest form, artificial intelligence predicts which items will be stored the longest and positions them accordingly. Forbes reports the case of a food supplier in the cold supply chain that increased its productivity by 20% in this way.

Smart routes

Failure to optimise transport means and routes can also be costly: a study by McKinsey revealed that inefficiencies in the pharmaceutical supply chain increase costs by up to 30%. AI algorithms make it possible to identify the most cost-effective and efficient means and routes of transport, taking into account factors such as geographical characteristics, weather conditions and logistical patterns. According to Analytics Insight, the global AI market in transport is growing at a compound annual rate of 15.8 per cent and is expected to reach $3.8 billion by 2025.

Managing billions of data

Pharmaceutical supply chains also generate huge volumes of data, which can grow exponentially as information sources increase. This is an industry-wide problem. The International data corporation (IDC) predicts that the global ‘datasphere’ will quadruple from 2019 values to 175 ZB by 2025 (one Zettabyte, ZB, is equivalent to one trillion Gigabytes).

Using AI-based analysis tools, such as machine learning and natural language processing, companies can identify hidden relationships in this magma of data and use them to make predictions.

An application for organising, labelling, cleaning and analysing biomedical and healthcare data, for example, takes only 12 minutes to analyse 1.2 million variants associated with a disease in 155 patients.

And the performance of these machines improves exponentially. To train a neural network at the level of AlexNet (one of the best known) today requires only 2% of the computing power needed in 2012, a progression that surpasses Moore’s Law (formulated by the co-founder of Intel, it predicts a doubling of computer performance every two years, at the same cost).

Optimising production

Another interesting contribution of AI concerns the production stages, where the ability to analyse data in real time allows immediate decisions to be made regarding activity scheduling, inventory levels and logistics. Algorithms can analyse data from sensors on equipment to predict when maintenance is due, reducing downtime and improving equipment efficiency, but they are also able to analyse product images to detect defects, similar to the analysis of radiological images for diagnostics.

The contribution of a predictive maintenance system can be significant: a McKinsey study estimated that these technologies can increase productivity by 50 to 100 per cent and even 150 to 200 per cent in ‘average performance’ laboratories. Automation can reduce manual errors and variability, ensuring better quality and compliance, with a 65% reduction in overall deviations and 90% faster turnaround times.

According to an industrial software provider, predictive analytics has reduced disruptions in a pharmaceutical company’s supply chain by saving 60% in maintenance costs and 50% in capital expenditure.