June 4, 2020 Business Challenges for Industrial Application of AI

Business challenges for industrial applications of AI and how to overcome them.

New artificial intelligence solutions are entering the markets at a fast pace. They create value and generate revenue at a still staggering growth rate. For the heavy industry, including oil&gas, discrete manufacturing, mining and metallurgical companies, new technology adoption is still a big challenge. These sectors are just starting the adoption of Artificial intelligence (AI). Compared to others, the heavy industry is lacking behind.

But what are the reasons and causes that have led to the current situation? The heavy industry is a very cost-driven sector. In its mature market reducing costs and optimizing processes are the top priorities set by almost all managers and directors. AI is now offering new approaches to do things. It seems to work differently in its way to create value when compared to more traditional methods. The biggest obstacle for starting AI projects is the high total cost of solution ownership.

Further, there is a problem in the absence of AI specialists with all the necessary skills. The main barrier for graduates entering these industries is outdated proprietary technologies, which demands time-consuming and expensive training and limits future job prospects. A successful AI project also needs the right set of tools as well as already-made templates, algorithms and implementation methodologies for the specific sector.

In general, the heavy industry is making long term investments. Huge investments are needed for new production plants and mining setups. Once started operation, a continuous optimization and costs reduction effort are in place. Here starts the interest in AI. The new use cases emerge to fulfill companies’ needs. Optimization leads to growing requirements on the speed of business processes changes. It increases pressure on the decision-making speed and exempt people from taking part in routine processes.

A lot of these narrow use cases, like defect detection, seem easy to solve with AI. However, it’s still hard to move on to organize the whole business processes, such as monitoring a whole production line and controlling the entire value chain. Critical decisions in solution architecture are difficult to pass on to outside people. Companies are not always ready to cooperatively develop solutions, expecting already implemented examples and staying reluctant to experiment and invest.

Traditional values are hindering costs savings

Organizational structure in the heavy industry follows, in most cases, very traditional values. This does not match the need for a more agile and flexible operation of a modern enterprise. Today, market development is led by digital companies. Their projects and new methods control the way development is heading. These companies are followed by large financial organizations and retailers, telecoms, eCommerce and service providers sectors. They are now more likely to experiment with AI technologies, but the experiment is becoming more widespread as they get real results from the implementation. Companies, except the oil and gas industry, mostly speak about the prospect of using AI technologies, rather than adopting them.

At the same time, we see the rise of cognitive services, which propose expert-level support for a broad range of areas. Cognitive AI systems and applications assist workers with data-driven guidance, partially addressing the shortage of data analysts and programmers. The use of AI for processes optimization gives a quick, measurable result and the main projects are in this area.

AI solutions are available and can be deployed now

All over the heavy industry, some pioneer companies implement AI solutions and do a lot of experimenting. AI startup Tulip Interfaces, a spinout from Massachusetts Institute of Technology, offers an IIoT-enabled manufacturing app platform which is in use by smaller manufacturers. Fanuc Corporation, the industrial robotics heavyweight, has created an artificial intelligence-based tool to reduce the robotics training process by hours. Or, Freeport-McMoran, a Copper miner, has successfully tested AI technology at its smart mine in Arizona, USA. They plan to increase the output by 90,000 metric tons of copper with very little capital investment.

AI technology is ripe but also continuously evolving. Data analytics allows companies to extract knowledge from a large volume of diverse information sources. AI technologies enhance human behaviour and are capable of identifying patterns, making recommendations while analyzing the results. The distinction of artificial intelligence compared to more common information technologies is its growth in capability over time. The use of AI processes and technology creates further data to keep the improvement process going. Examples of successful projects and business gains from the use of AI technologies will be a convincing argument for AI projects implementation.

Why is it not done?

Industrial use cases take into account domain specifics and business processes. These are currently used less frequently but are considered by companies for use in the next 1 to 2 years. 40% of companies plan to automate predictive maintenance, 48% are considering supply chain and logistics, 34% are prioritizing quality monitoring and management. These scenarios reflect the plans of manufacturing companies to digitize the business as a whole.

Despite these intentions, the heavy industry faces management and decision-making problems. Often, they’re stuck in their traditional separation of concerns. Business is less frequent driven by technology and IT key players. More often, directors and officers focus solely on financials. The situation is very different from technology-driven sectors like commerce, finance, or hospitality & leisure.

The involvement of business representatives in AI projects is crucial for gaining financial returns on investments in AI. The positive results of the project depend not only on the correctly chosen scenarios but also on the support of the management. The market can grow even faster, given the low base effect. The growth rate will be positively affected by:

•     Undergoing digital transformations in companies and public bodies
•     Local players with expertise in AI and implementation experience
•     Positive pilot projects result and proven economic benefits
•     Initiatives to roll out projects that have already begun
•     Availability of government programs

The excuses to not change are that it is not hurting enough. Continuous business is often still running at good levels. And some marginal improvements could be achieved with traditional methods. As the massive effect of AI is not a sudden shift. The impact of this disruption is spreading more slowly, and it is hard to catch up once you’re left behind the Big Data acquisition.

The specific problems of the industry

The majority of industrial use cases boost the workers’ productivity, increase the efficiency of key business processes and control quality. The most significant potential is locked in narrow AI applications for prediction and making recommendations, pattern recognition systems and video streaming analysis.

The general assumption of industry decision-makers is that the heavy industry is “special”. The size and complexity of the production and plants lead to a focus of thinking only inside their own sector. AI requires thorough knowledge of these business processes and industrial AI specifics. So, industrial companies rarely use external applications of consulting services in this area. This is still the case, despite the lack of specialists in this field. Also, poor pricing strategies and business models lead to the creation of industrial tools rather than turnkey solutions.

In production companies, digital assistants are capable of giving recommendations on choosing the optimal parameters and technological modes to achieve maximum productivity. Recommendations are based on historical data and predictive models which eliminate human error and increase process efficiency. Further, it is necessary to take into account the interaction between man and machine. Because productions, especially discrete ones, are a place where people and machines can be located and influence each other.

Ways to start

When a decision to implement an artificial intelligence solution is settled up, the detailed way of getting started has to be elaborated. The cost sensitivity and the absence of AI experts with the needed skills and understanding inside most enterprises could make that difficult.

One tested and viable option is to team up with capable and experienced big IT guys. Google, Microsoft, IBM, Huawei, Amazon, Baidu have a lot of enterprise experience and can handle the unique needs of the heavy industry. They provide the needed trust and resources.

The lack of data might be overcomed with cross-industry data sharing. Companies see the risk of losing competitive advantage in the information exchange because of IP protection. Data have monetary but intangible value, which companies do not want to share with others. The industry has yet to figure out how large data sets can be valuable to all participants without violating privacy laws and contractual agreements.

As a result of the lack of expertise, customers’ expectations regarding the capabilities of computer vision systems, as well as their cost and payoff period, are overstated. So, it’s recommended to start a small project to learn and improve the knowledge of AI applications. Training and investing in employees is a key part, as AI is not IT. Unfortunately, there is an actual fear fueled by negative AI media coverage, powering employee resistance on AI adoption. AI projects are focused around a lot of data and need a more holistic approach.

The digital transformation officer unites AI projects

Organizations should start working on strategy and roadmaps for AI deployment. AI is a core part of digital transformation and should be delivered by the business in close cooperation with IT, in which the business department determines the desired outcome. The IT department is in charge of the required solutions and technologies. The initiative to implement projects goes from top to bottom, and there is a need in a strong leader, often a CEO or director of digital transformation, to help models become a part of business processes and prove their value.

Ekaterina Lyapina, AI and IIoT Consultant at Zyfra

Zyfra develops turnkey solutions for the mining, oil and gas, machinery and metallurgy industries that drive digital transformation and enhance operational efficiency for businesses across the globe.

Subscribe to Industry Today