When most people think about automation applications, knowingly or not, what comes to their minds is Robotic Process Automation (RPA)—the automation of tasks, such as data entry or updating records. Cognitive automation is much more than that. This sophisticated, intelligent technology aims to automate the decision-making process itself.
Although Cognitive Automation is just beginning to emerge in the marketplace, the path to its future as a mainstream application is straightforward. It follows the evolution of systems in four stages, from providing enhanced visibility to better forecasting, cross-functional optimization, and finally, the automation of decisions.
To describe this evolution, we will start with supply chain control towers.
The need for supply chain visibility was driven primarily by logistics, with control towers emerging in the 1990s. These control towers consisted of teams of in-house operators attempting to stay on top of product movement and any disruptions to the supply chain. Their objective was to react quickly to these disruptions, albeit supported by rudimentary tools, primarily Excel, phone, and fax.
Soon, technology solutions came along to support the operation—including GPS trackers that could provide more accurate visibility and systems that could centralize information and facilitate the execution at every stage of the supply chain.
These systems enabled the first giant leap forward, but they had several shortcomings. Among them, they mainly functioned in silos with poor integration capabilities. They lacked a uniform infrastructure, made the ecosystem architecture complex, had data inconsistencies and gaps, and poor user interfaces, often leaving operators to manage exceptions outside the system.
Over the years, systems improved, and a wave of new tech providers entered the scene with solutions that improved on earlier applications. Although no business has perfect end-to-end visibility, we can attribute most of the current limitations to the complexity of the supply chain more than the technology capabilities themselves.
With vast improvements in visibility, businesses started focusing on leveraging the data in a better way. Armed with much more information, leaders didn’t want to react to disruptions anymore - they wanted to predict them. A trend emerged, with analytics evolving from prescriptive to predictive.
Event forecasting within the supply chain and the need for businesses to collaborate with third parties is as old as the supply chain itself, but three unique ingredients shaped this new forecasting and collaboration stage: Big Data brought in by the visibility stage; the rise of artificial intelligence (AI), machine learning capabilities and data science capabilities; and interconnected tools that make communication possible for different parties across different channels.
Today, nearly all logistics applications come along with a set of ETAs (Expected Times of Arrival), such as expected production release, customs processing ETAs, ETA to next destination, anticipated delivery to the customer, and more. Solution providers ubiquitously mention the use of AI, machine learning, and data science in their prediction engines, though they leverage these to different degrees.
These predictions have also been enhanced by augmenting the businesses’ datasets with broader information such as commodity prices, weather forecasts, port disruptions, traffic, etc. The combination of Machine Learning and the additional external datasets have increased the accuracy of forecasts beyond what planners can build within their four walls.
At the same time, for operators to act on these predictions, they needed a set of tools to collaborate internally across functions and externally with third parties. Traditional supply chain systems were not built for workflow purposes, and operators resorted to other troubleshooting platforms, such as online meetings, email, phone, and messaging applications.
Some applications have recently started to embed workflow functionality, including alerts, task reminders, communication channels, file sharing, and reporting. With the breadth of the supply chain and the number of parties involved in the process, a collection of collaboration tools will likely continue to be the norm.
This need for efficient collaboration and, more specifically, cross-functional efficiency is at the center of the next trend in technology that businesses are currently tackling.
The challenges of the Intelligent cross-functional decision-making stage will be organizational in nature. As the supply chain grew ever more complex, not only did technology solutions become narrower and deeper in their expertise, but so did the operators, and as a result, corporate structures, just like the IT ecosystem, formed silos.
As processes evolved, expertise developed in every step of the supply chain. Visibility and forecasting were tailor-made for functions such as procurement, warehousing, and transportation, and these in turn into sub-functions such as Inbound and outbound transportation. The broader cross-functional optimization, for example, optimizing inventory levels in the warehouse based on transportation ETAs or production capacity, has been solved organically at higher echelons of the corporate structure, by leaders that have visibility into multiple functions, and from a technology perspective thanks to well-designed IT architecture with multiple system integrations and robust reporting.
The next stage will be won by businesses with corporate structures with efficient cross-functional communication and an understanding of their IT silos. In this next stage, solution providers will start expanding across functions to provide the needed value-add seamlessly. Still, the real winners will be emerging platforms that sit on top of the ecosystem. By combining end-to-end supply chain data and digitizing and modeling how the operation works, these platforms constantly search for cross-functional opportunities in a native way.
Empowered leaders will be able to make decisions more intelligently and more quickly than competitors and optimize for the enterprise at large. AI engines will provide insights in the form of recommendations. However, there will still be one constraint—time. Platforms supported by AI technology with 24/7 data visibility will produce far more recommendations than humans can effectively manage. Teams will not have enough time to go through every opportunity and will naturally focus on making sure those with the most significant impact are correct.
The final stage will be the automation of these intelligent cross-functional decisions with a cognitive automation tool, as operators gain more and more trust in the recommendation platforms. The transition will happen slowly at first, with teams setting thresholds or limited scope for the recommendations to be automated. Eventually, the Cognitive Automation tool acts as a digital workforce, completing complex tasks with little to no human intervention. Automating the recommendations will effectively let the platform make the decision—a process called cognitive automation.
Let’s take inventory transfers as a simple example. Imagine a platform that is constantly looking at inventory levels across all DCs and SKUs, looking for stock rebalancing opportunities. This platform would have end-to-end data visibility, from purchase orders to production capacity and from transportation ETAs to demand forecasts. It would have also digitized the institutional knowledge behind how the business rebalances stock.
Perhaps the teams aim to maximize truckload capacity and have modeled DC staffing constraints, or maybe they have a DC or SKU prioritization in place. Given the data, an understanding of the process and the constraints, and provided with a goal to maximize, this platform would be able to produce recommendations to rebalance stock.
Furthermore, the users could dictate that any transfer with a cost under a certain threshold can be automated. This kind of recommendation does not need user acceptance, and the platform can instruct the source systems to execute the transfer without user intervention. That, in essence, is cognitive automation—vastly more sophisticated and effective than RPA.
As the thresholds get larger and the scope of automated decisions expands, the platform will make more decisions by itself. The enterprise, in a way, adopts a "digital brain." In this future, operators are freed of their routine tasks and spend more time optimizing the process and lifting constraints, and leaders can spend more time evaluating "what-if" scenarios and working on strategy.
Cognitive automation is no longer just a vision of the future. It is the fourth, upcoming stage of our technology evolution.