Supply Chain AI: Getting the Most Bang for Your Buck

Organizations can drive significant value by making incremental efforts toward data analytics maturity

The widespread scale of the coronavirus pandemic, and the need to distribute a vaccine the moment it’s ready, has introduced new complexities to the global supply chain. These include the number of parties and jurisdictions involved, varying maturity levels in data analytics, and transportation and communication issues, among others.

Artificial intelligence (AI) and data analytics may present opportunities to more accurately predict challenges and plan an efficient rapid response while minimizing future disruptions.

The COVID-19 pandemic highlighted new and existing issues in many industries’ supply chains. Consumer goods, manufacturing and healthcare supply chain disruptions have made headlines since the beginning of the year. In addition, some logistics organizations are struggling to gather and analyze quality data while bottlenecks at any link in the chain threaten to cause cascading disruptions.

This is where AI – and specifically machine learning – can help.

AI may refer to many implementations of technology, but machine learning is the most prominent implementation of AI. It uses algorithms and applications to automate data analysis and create models of knowledge. Machine learning solutions may be used to perform predictive analysis, such as regression analysis and classification, which can be particularly helpful in predicting business issues relating to the supply chain.

Transportation applications

Transportation issues are frequently a significant component of supply chain disruption. AI solutions may help to solve these challenges by automating data collection from various points in the route and then using anticipatory shipping to a satellite location to meet consumers’ needs – sometimes even before the need is reported.

Another way to increase efficiencies in transportation could be to enable rescheduling of deliveries and trucking route modifications based on latest traffic and weather patterns. Including this data in predictive models can make those predictions more relevant and the process more efficient.

Other helpful uses include predicting inventory outages. Take the example of the eventual distribution of a COVID-19 vaccine: it would be crucial to predict not only the stock availability of vaccines themselves, but also of the peripheral supplies – such as syringes, diluent, and refrigeration supplies. All these factors could ultimately impact millions of lives. Even patient care-related predictions such staffing needs and appointment time per patient for immunization could become important.

A huge supply chain challenge

Distribution of a COVID-19 vaccine will soon be the biggest supply chain challenge facing the world. To successfully deploy a vaccine, organizations may need to predict several aspects related to the supply chain, including:

  • Timing of consumption by country, region, city, and perhaps immunization locations – Learning where vaccines will come from and where they will be distributed at their final destination can help to streamline logistics.
  • Sourcing, availability and cost of materials – Predicting shortages will be especially important to manufacturers so they can mitigate potential risks.
  • Production locations, lot scheduling and sizing – Distributors will need to account for size and availability of storage facilities along the distribution route.
  • Potential for strain on quality control resources – Several quality control points will likely be necessary to guarantee the viability of the vaccine. Overloading these “checkpoints” would create bottlenecks in the chain.
  • Spoilage probability and cascading effect – COVID-19 vaccines need to be stored in controlled cold temperatures and are susceptible to spoilage if deviations occur. Modeling space allocation and anticipating problems with stocking may reduce potential for wasted inventory.

The nature and scale of COVID-19 introduce long-tail risk scenarios that include more uncertainty than prior immunization deployment information could help solve. Many more simulations may be necessary to provide examples of alignment of multiple low-probability events and scenarios.

Automating data collection and data transformation through the use of robotic process automation (RPA) from as many sources and organizations as can be involved in the manufacturing and distribution of the vaccine may help diminish manual errors, speed up the process, and enable analysts to make more accurate predictions.

It’s all in the data

While there are many complexities involved at every junction of a supply chain, organizations can drive significant value by making incremental efforts toward data analytics maturity – without necessarily adopting a robust machine learning solution. The value from AI capabilities and improving data analytics ultimately comes in the form of better decision-making in the face of uncertainty. An organization’s data may contain indicators of risk and opportunities for new value. Most organizations can start by improving their processes around data governance, unlocking the data’s true potential.

Data for modeling may come from many sources: past and present supply and demand patterns, real-time traffic and weather updates, inventory data, market predictions, etc. As with any input-output process, more accurate data input yields more accurate predictions. Improving version control and change management practices around data management can help protect the quality of data. Furthermore, the assumptions gathered from this data and incorporated into predictive models need to be well-documented to explain the rationale and to allow ongoing monitoring of model performance and adjustment.

In addition to the data used for establishing models, data for updating and adapting models is critical. The faster that reliable data can be received back through the supply chain, the faster that other parties can respond. In the coronavirus immunization example, data from immunization sites (such as hospitals and clinics) should be shared as efficiently and accurately as possible to enable manufacturers and logistics companies to respond accordingly.

Implementing AI solutions

An automation specialist with experience implementing RPA solutions can help organizations identify and refine data sources from across various business functions. Once the relevant data and processes are identified, automation can improve the collection process and data quality.

In addition, there are automation solutions out there to help organizations establish standardized business logic, enabling improved identification and monitoring of key performance indicators (KPIs) with the use of management dashboards, and mitigating risks more quickly and competently.

The COVID-19 pandemic may continue to have disruptive impact on industries for years to come, and some of the long-term implications are not yet obvious. Nonetheless, good data analytics practices are available to organizations of all sizes and levels of sophistication.

Artificial intelligence, specifically machine learning and automation, may assist some organizations to predict events and take anticipatory action. Organizations that haven’t already done so should plan now how they will use these new technologies and the power of data to unlock potential opportunities and value, while addressing the supply chain challenges facing the world today.


Ada Cohen, CFE, CAMS, ABV, is a Risk Advisory Services Director at Kaufman Rossin, one of the Top 100 CPA and advisory firms in the U.S.