The artificial intelligence (AI) we use today is known as weak or narrow because it can only handle the specific tasks it is programmed to do. Strong AI, or the level of AI often seen in fantastical science-fiction movies, is unlikely to be achieved in our lifetimes — if at all. Still, the technology is improving all the time.
The fact is, humans everywhere rely on AI and other information technology (IT) automation tools to accomplish all kinds of tasks. In particular, approximately 5.8 zettabytes of data were collected in 2020. To put that in perspective, one zettabyte is approximately equal to the number of grains of sand on all of the beaches in the world. That’s a lot of information to be processed, and humans can’t do it alone. Yet, computers can’t do it without human programmers. Truly, the key to success is the collaboration between humans and intelligent automation.
Intelligent automation is the combination of AI and automation, although it also can be used to describe a combination of other technologies:
- Machine learning is a method of data analysis that automates analytical model-building and is based on the idea that systems can learn from data, identify patterns and make decisions with minimal human intervention.
- Natural language processing is a branch of AI that combines computational linguistics — rule-based modeling of human language — with statistical, machine-learning and deep-learning models. It enables computers to understand the meaning and sentiment of both voice and text data conveyed in a human language.
- Robotic process automation automates routine, repetitive and predictable tasks through assisted bots, which are deployed on an individual machine and carry out cumbersome or technically complex portions of tasks. Meanwhile, a human manages other activities and unassisted bots, which are deployed on a centralized server, enabling manual control.
To emphasize the desirability of integrating entire systems, Gartner introduced the term hyperautomation, which it defines as a “business-driven, disciplined approach that organizations use to rapidly identify, vet and automate as many business and IT processes as possible.”
Hyperautomation is certainly attractive. A 2019 Harvard Business Review report found that only 10% of surveyed companies had sophisticated AI with multiple integrated applications. However, nearly 60% expected to have “sophisticated and extensive” applications in place within three years.
Still, it is important that organizations do their own research before implementation in order to ensure that new tools are compatible with existing systems. Companies that have jumped too soon have struggled with basic internet-of-things sensor connectivity issues, a lack of internal human talent to support the tools, and challenges finding the best method of data storage for future use in AI modeling. Experts currently advise organizations to store data in unstructured formats to make it easier to convert it for AI modeling. In addition, they should implement AI in chunks such that the technology can be rolled back, if needed, in a way that will not cause interruptions for the enterprise.
Gartner stresses the importance of adopting a holistic approach to hyperautomation. Experts are concerned that businesses might take a narrower view of the road to automation because of the lack of guidance for implementing AI technologies. As a result, AI-enabled intelligence often is delivered without an integrated strategy. Instead, Gartner encourages organizations to have a long-term strategy for integrated AI implementation.
Intelligent supply chain implementations
Supply chain organizations that have implemented AI applications, in whole or in part, already have achieved a variety of benefits:
- Locate interchangeable parts or substitute components, materials, formulations or ingredients
- Gather and consolidate supplier data from multiple and diverse sources to ensure understanding of different practices
- Analyze agreements, past purchases and quality trends, along with service-level agreements, in much less time than it would take a human to do so
- Manage risk by figuring out alternate suppliers and transportation routes when an emergency disrupts the supply chain
One of the most promising areas for AI in supply chain management is demand planning. AI applications tend to spot patterns and trends in both structured and unstructured data long before humans do. This can support more efficient inventory planning and purchasing.
AI applications also will make it possible for supply chain managers to evaluate present and potential suppliers for existing or future supply chain opportunities. In addition to evaluating suppliers on the basis of their product price, quality and availability, organizations will be able to assess emerging considerations such as sustainability programs, employee working conditions and the stability of management as this more nebulous information becomes available, often in unstructured formats.
In manufacturing, a study from the European Business Review discovered that AI is making advancements in the following areas:
- Defect detection: By integrating machine learning and AI into the process, these systems now are powered with self-learning capabilities.
- Quality assurance: Image processing algorithms have been developed that can automatically evaluate and establish whether an item has been perfectly produced.
- Predictive maintenance: AI can drastically cut down the relatively high costs associated with unplanned downtime while extending the remaining useful life of production robots. Predictive maintenance workers now are being trained for more advanced positions like in product design and equipment maintenance.
- Generative design: Engineers can input their design goals and parameters such as materials, manufacturing methods and cost constraints into a generative design software. The system then explores every possible configuration and provides the best design alternatives.
- Inventory management: Machine learning can be used to design solutions that promote inventory planning activities because they are better at dealing with demand forecasting and supply planning.
- Demand prediction: Predictive analytics tools are used to estimate market demands by looking for behavioral patterns, linking key factors including location, socioeconomic and macroeconomic factors, and weather patterns.
- Customer service: AI-powered solutions can analyze the behaviors of customers, identify patterns and then predict future outcomes.
Because of these skills in analyzing customer behaviors and delivering the results consumers want, there is significant interest in intelligent automation in the retail industry. IBM, in conjunction with the National Retail Federation, conducted a survey of 1,900 leading retail and consumer products companies in 23 countries. They found that the application of intelligent automation is expected to increase from 40% to more than 80% within the next three years. In retail companies, the major areas of interest will be in supply chain planning, demand forecasting, customer intelligence, marketing and advertising, store operations, and pricing and promotion.
As an example of retail intelligent automation in action, Nike, Beaverton, Ore., has designed a system that enables customers to design their own shoes by donning sample shoes and, using a voice-activated system, selecting the preferred fabric and color. The system uses augmented reality, object tracking and projection to display the custom shoes to the consumer. The finished shoes are ready within two hours.
Another application that holds promise for retailers is more extensive analysis of customer buying habits to transmit targeted ads or suggest additional products for consumers. Many consumers will recognize and appreciate the care that retailers take to provide relevant recommendations and buying opportunities for them.
Weather forecasting is another prediction area in which intelligent automation can assist supply chain professionals, particularly those involved in inventory planning and transportation and distribution. Today’s mathematical forecasting models incorporate about 100 million pieces of data each day — a level of complexity comparable to simulations of the human brain or the birth of the universe. Analyzing this amount of data exceeds the ability of conventional data analysis methods. AI involves the use of neural networks to detect patterns that may portend future weather occurrences. For example, AI applications show promise for increasing the speed and precision of short-term severe weather forecasts, such as for tornados and hail storms. While not replacing traditional weather forecasting, AI will augment and strengthen existing methods.
AI use in the form of increasingly intelligent robots is expanding in the logistics arena. Warehouses have been using robotics for some time in the form of automated guided vehicles (AGVs). However, these robots are confined to prescribed routes. Today’s warehouse robots are more flexible and can maneuver themselves through a warehouse by using built-in cameras and sensors that provide navigation and check for safe maneuvering conditions. Another promising application is using picking robots to pick products from shelves and bring them to a specified location, such as the station of a human picker or a loading dock. Some robots are equipped with sight capabilities that enable them to vary their picking arms to accommodate different product sizes. Going forward, robots also can be used for sortation and order fulfillment. As more logistics companies invest in the technology, the overall cost of the solutions decreases and the capabilities of the solutions increase.
There are both technical and ethical obstacles to the successful implementation of AI. One of the major technical issues is the immaturity of the technologies. As a result, for now, AI is more prevalent in the largest and most capable organizations. Another obstacle is that implementation of AI requires process changes and therefore changes in employee skills. Finally, to be completely effective, AI applications must be fully integrated with other systems within the organization, which can be tricky.
IBM raises some ethical points regarding AI, such as:
- Liability issues, in terms of who is responsible for accidents caused by autonomous vehicles
- Impacts on human jobs, including reducing the number of human jobs or requiring human workers to learn more complex skills and change roles
- Privacy related to the protection and use of personal data
- Discrimination caused by facial recognition programs
- Accountability, particularly because of a lack of industry standards and a means of enforcing AI rules
Michael Totty of The Wall Street Journal offers this illustration of the dichotomies of AI: “To some experts, an AI world means more jobs, and more interesting ones; to others, it means a devastating loss of employment opportunities. To some, it means a deadly threat to human existence; to others, it means better health and longer — perhaps much longer — lives. To some, it means a time when AI can help us make smarter decisions; to others, it means the destruction of our privacy.”
Regardless of these challenges, AI solutions most likely are the future. Tom Davenport of Babson College points out that, in the past, interest in AI has waned and waxed and then been followed by another burst of enthusiasm and hype. Right now, there are thousands of AI startups making enormous technology progress along multiple fronts. Similarly, universities, research institutions and other organizations are studying the possibilities and implications of AI use. However, it is important not to expect too much from AI too soon. Davenport writes for MBR Journal: “It will no doubt become a revolutionary force in the fullness of time, but right now it is largely evolutionary. As Amara’s Law suggests, we are likely to overestimate AI in the short run and underestimate it in the long run.”
For now, humans are still in control, and we’ll be in control of where AI goes next.