Digitizing Chemical Supply Chains with Cognitive Automation

Many businesses dealing with chemical supply chains are “mature,” meaning that they evolved their processes well before the age of digital natives. The first challenge, then, is these companies must join or even build a digital platform ecosystem before any automation is possible.

The core processes typically used in their supply chain planning and execution tend not to be geared for the digital age. The pain points in their production are issues like delivery time, stock inventory, and ability to deal with various company representatives—all of which are ripe for digitization to improve outcomes.

End customers report their key buying factors are short delivery time, price transparency, and available tech support. Therefore, the goal is always faster speed and higher reliability.

A Four-Step Process For Applying Cognitive Automation

Applying cognitive automation to a supply chain starts with digitizing and harmonizing the applicable data from both within the enterprise systems and the outside world, and then using it in a four-step process to drive decision-making rooted in data and best outcomes.

1. Sensing and predicting where the harmonized data is compared to either deterministic or scenario-based rules to determine when a problem is likely to occur.
   

2. Solution exploration is accomplished through a variety of methods (decision trees, heuristics, advanced optimizers).
   

3. Option evaluation uses option ranking and trade-off optimization.
   

4. Action and learning occur when a human is presented with the scenario and then chooses to go with the system’s recommendation or not. The system learns from each of these decisions and adjusts accordingly.
   

A Cognitive Automation Case Study: European Chemical Brand

Let’s take a deeper look at how intelligent automation technologies can support companies focused on transforming their business processes for the digital era.

Prior to implementing cognitive automation, a European Top 10 player faced multiple business challenges. Relevant data was siloed in more than 100 sliced dashboards and spreadsheets, necessitating a huge portion of human workforce hours being squandered on data processing rather than on solving problems or innovating.

Their goals with cognitive automation were to harmonize their systems/data, shift worker focus to triaging issues through automated risk assessment, and to increase efficiency and overall speed of business.

The data was brought together using continuous crawlers targeted on the client’s goals. This information was then fed into a “digital workbench” where the data was normalized via processing and blending rules.

Finally, the normalized data fed into the cognitive data layer, which was targeted to key areas such as sales risk, cost management, and shipment issues. The system is scalable across many use-cases.

The cognitive layer was developed by using worker experience to fuel the decision science; starting with current best practices and subject experts is how the process begins. Using current business rules, risk categories are generated, then classification models are built which can utilize the available data.

It very quickly became apparent that cognitive automation surpassed human ability to foresee production issues. For example, there was previously a 75 percent false-positive rate of predicting delayed delivery, necessitating rush orders, spot-buys, and increased expense. Machine learning reduced that false-positive rate by 60 percent.

The data shows that human reactions can sometimes exacerbate supply chain issues, due to the instinct to over-correct.

Intelligent Technologies Improve Accuracy, Foster Collaboration

When we look at this case study through five separate indices, significant improvement was seen across the board.

• Automation aggregated and harmonized the necessary data in real time, allowing for revamped risk classification and recommendations from the system.
• Coverage of the supply chain was no longer dependent on humans making connections between siloed processes, and problems stopped falling through the cracks.
• Accuracy of the machine learning recommendations was 60 percent better than human rules-based predictions.
• Frequency of data-gathering was twelve times a day via the crawlers, giving a real-time snapshot and enabling immediate business responses as needed.
• Granularity of problem-solving improved, with action on 100% of transportation orders vs. just 40 percent prior to automating.
  

In addition to service levels going up and rush orders going down, automation drives business processes, enabling collaboration and change.

Workers are shifting from dashboards to storyboards. While automation was initially met with wariness, workers quickly realized it freed up time spent on “boring” tasks and allowed them to work in more creative and impactful ways, achieving the ultimate goal of cognitive automation—freeing humans to do the work that humans do best.

‍‍

Get social with us!

Follow us on LinkedIn: Cognitive Automation Community
Follow us on Twitter: CognitiveAutomation

‍