While Artificial Intelligence (AI) has been prevalent in industries such as the financial sector, where algorithms and decision trees have long been used in approving or denying loan requests and insurance claims, the manufacturing industry is at the beginning of its AI journey. Manufacturers have started to recognize the benefits of embedding AI into business operations—marrying the latest techniques with existing, widely used automation systems to enhance productivity.
Antony Bourne, President, Industries, IFS, explains manufacturers must construct ethical systems as AI becomes pervasive in the industry. He outlines three major ethical problems facing the industry as manufacturers ramp up their AI investments and embark upon this new technological direction.
Manufacturers are heavily investing in AI. A recent international IFS study polling 600 respondents, 383 of which were major manufacturing decision makers, working with technology including Enterprise Resource Planning (ERP), Enterprise Asset Management (EAM), and Field Service Management (FSM), found more than 90 percent of manufacturers are planning AI investments.
Combined with other technologies such as 5G and the Internet of Things (IoT), AI will allow manufacturers to create new production rhythms and methodologies. Real-time communication between enterprise systems and automated equipment will enable companies to automate more challenging business models than ever before, including configure-to-order or even custom manufacturing.
Despite the productivity, cost-saving and revenue gains associated with AI, the industry is now seeing the first raft of ethical problems come to the fore as it becomes more widespread. Here are the three main ethical considerations companies must weigh-up when making AI investments.
AI will make its strongest contribution towards top-line growth by creating net new product and service offerings that were previously unachievable. For manufacturers executing aftermarket service contracts, the use of chatbots equipped with natural language processing (NLP) capabilities are the obvious choice.
These machines can automate a high percentage of customer interactions, helping manage simpler service needs that do not require specific escalation, thus allowing support staff to focus on more complex issues. Further, tying AI systems into connected devices has the potential to make remote resolutions more efficient and far less labor intensive.
But there will also be significant implications for the bottom line as the number of labor hours required to produce value drops, and this has led to various opinions and predictions on how exactly AI will impact the workforce.
According to a 2017 National Council on Compensation Insurance (NCCI) study, since 1990 U.S. manufacturing output increased by over 70 percent while employment fell over 30 percent in the same period. This means that in 2016 the U.S. produced at least 70 percent more goods than in 1990 with only about 70 percent of the workforce, and largely explains why during this time U.S. manufacturing labor productivity grew by a staggering 140 percent.
As the NCCI study points out, automation in manufacturing has reduced production costs, making U.S. manufacturers less expensive and the role more competitive because of the reduced requirement for labor—particularly if new jobs created by AI are short-lived during a transition period. Important questions are now being asked—will we need a shorter work week, new business models or economic systems to deal with this shift?
Optimists suggest AI technology may replace some types of labor, but that efficiency gains will outweigh transition costs. They believe AI will come to market at first as a guide-on-the-side for human workers, helping them make better decisions and enhancing their productivity. In addition, the investment in this technology has the potential to upskill existing employees and increase employment in business functions or industries that are not in direct competition with AI.
Furthermore, recent IFS research points to an encouraging future for a harmonized AI and human workforce in manufacturing. The IFS AI study revealed that respondents saw AI as a route to create, rather than cull, jobs. Around 45 percent of respondents stated they expect AI to increase headcount, while 24 percent believe it won’t impact workforce figures.
It is easy for organizations to say they are digitally transforming. They have bought into the buzzwords, read the research, consulted the analysts, and seen the figures about the potential cost savings and revenue growth.
But digital transformation is no small change. It is a complete shift in how you select, implement and leverage technology, and occurs company-wide. A critical first step to successful digital transformation is to ensure that you have the appropriate stakeholders involved in your digital journey from the very beginning. This means manufacturing executives must be transparent when assessing and communicating the productivity and profitability gains of AI against the cost of transformative business changes to significantly increase margin.
When businesses first invested in IT, they had to invent new metrics that were tied to benefits like faster process completion or inventory turns and higher order completion rates. But manufacturing is a complex territory. A combination of entrenched processes, stretched supply chains, depreciating assets and growing global pressures makes planning for improved outcomes alongside day-to-day requirements a challenging prospect. Executives and their software vendors must go through a rigorous and careful process to identify earned value opportunities.
Implementing new business strategies will require capital spending and investments in process change, which will need to be sold to boards of directors, investors and other stakeholders. As such, executives must avoid the temptation of overpromising when discussing AI. They must distinguish between the incremental results they can expect from implementing AI in a narrow or defined process as opposed to a systemic approach across their organization.
There can be intended or unintended consequences of AI-based outcomes, but organizations and decision makers must understand they will be held responsible for both. We have to look no further than tragedies from self-driving car accidents and the subsequent struggles that followed as liability is assigned not on the basis of the algorithm or the inputs to AI, but ultimately the underlying motivations and decisions made by humans.
Executives therefore cannot afford to underestimate the liability risks AI presents. This applies in terms of whether the algorithm aligns with or accounts for the true outcomes of the organization, and the impact on its employees, vendors, customers and society as a whole. This is all while preventing manipulation of the algorithm or data feeding into AI that would impact decisions in ways that are unethical, either intentionally or unintentionally.
Margot Kaminski, associate professor at the University of Colorado Law School, raised the issue of ‘automation bias’—the notion that humans trust decisions made by machines more than decisions made by other humans. She argues the problem with this mindset is that when people use AI to facilitate decisions or make decisions, they are relying on a tool constructed by other humans, but often they do not have the technical capacity, or practical capacity, to determine if they should be relying on those tools in the first place.
This is where ‘explainable AI’ will be critical—AI which creates an audit path so both before and after the fact, there is a clear representation of the outcomes the algorithm is designed to achieve and the nature of the data sources it is working form. Kaminski asserts explainable AI decisions must be rigorously documented to satisfy different stakeholders—from attorneys to data scientists through to middle managers. “A lawyer may be interested in different kinds of explanation than a computer scientist, like an explanation that provides insights into whether a decision is justified, whether it is legal, or allows a person to challenge that decision in some way,” says Kaminksi.
Manufacturers will soon move past the point of trying to duplicate human intelligence using machines, and towards a world where machines behave in ways that the human mind is just not capable. True digital transformation through AI will see its influence across all processes within an organization, automating systems and making repetitive tasks a distant memory.
While this will reduce production costs and increase the value organizations are able to return given limited inputs, this shift will also change the way people contribute to the industry, the role of labor and civil liability law. There will be ethical challenges to overcome, but those organizations who strike the right balance between embracing AI and being realistic about its potential benefits – alongside keeping workers happy and, in turn, contributing to society – will usurp and take over. Will you be one of them?