Thanks to Dr. Dexin Luo (Director of AI, Linde), David Schaer (President, Computers Unlimited), Hector Villareal (President, Weldcoa) for their contributions.
The term Artificial Intelligence or “AI” has been around for close to 60 years. However, over the last 5-10 years as computing power caught up with AI’s large data processing requirements, use cases across most consumer end-markets have exploded. Voice assistants (Siri, Amazon Alexa, etc.), recommendations viewed on social media (i.e., Twitter, Facebook, YouTube), streaming services such as Netflix, and many Leading-edge eCommerce websites (i.e., Amazon) all utilize AI technology at their core. It is no longer a technology of the future…AI has arrived, and we’ve only scratched the surface. On the other side, the industrial manufacturing space (inclusive of Industrial gases) has been a bit of a later adopter, but AI has been gaining momentum over the last 3-5 years. Use cases are being developed/tested (or have been implemented) in areas including sales and marketing, plant optimization, supply chain operations, IT, risk management, and R&D to name a few. Consequently, we will take a deeper look at the progress the industry is making in AI, starting with current highlights and trends, use cases (or anticipated to be used), best practices and outlook for companies looking to get started or expand their use of AI.
The history of artificial intelligence dates back thousands of years with classical philosophers and mathematicians mulling over the idea that artificial beings had existed or could exist in some fashion. But then much later, artificial intelligence became increasingly more tangible throughout the 1700s and beyond when philosophers contemplated how human thinking could be artificially mechanized and manipulated by intelligent non-human machines. The beginnings of modern AI was not formally founded until 1956, at a conference at Dartmouth College, in Hanover, New Hampshire, where the term “artificial intelligence” was coined. Cognitive scientist and Dartmouth professor John McCarthy coined the term when he began his exploration of whether machines could learn and develop formal reasoning like humans.
Since then, AI has grown tremendously and encompasses a wide range of technologies and definitions, but the bicycle analogy by Timothy Havens of Michigan Technological University captures its essence,
…the child learning to ride falls a few times, honing their skills each time they fail…that’s AI in a nutshell…Timothy Havens of Michigan Technological University
Artificial intelligence has grown to now encompass a family of technologies including machine learning, deep learning, natural language processing, computer vision (which enables facial recognition) and many others. It is a platform of technologies that is a computerized simulation of human intelligence that can be programmed to make decisions, carry out specific tasks, and learn from its results. Typically, AI analyzes historical data to make predictions about the future, and as it collects more data from users and the public domain, it learns, and the predictions become more accurate. As a result, the field of AI has continued to grow as new capabilities have been developed.
In 2022, Gartner, IBM, McKinsey, and others have all quoted that at least 25% of their client bases have implemented AI in some form and the number continues to rise. In a McKinsey study taking a look at AI utilization across all end-markets over the last 5 years, they observed that adoption has more than doubled since 2017, and the average number of AI capabilities that organizations use, such as natural-language generation and computer vision, has also doubled. The top use cases, however, have remained relatively stable with optimization of service operations remaining in the top spot each of the past four years. With respect to revenue-based value from AI, the areas of marketing and sales, product and service development, and strategy and corporate finance reported the highest value. On the flip side, the highest cost benefits from AI was universally in supply chain management space.
Without a doubt AI has entered the commercial arena through the massive adaptation of Big Data statistical analysis and the IOT (Internet of Things). When I gave a presentation at the Dubai Gasworld conference in 2015, and shared that AI would be one of the disruptive technologies to keep an eye on, I could see that it would be one of the decade’s megatrends, but I did not anticipate how far its capability would improve in the last 7-8 years! AI was the missing link between IOT and Big Data, and now all three are developing and feeding off each other.Hector Villareal, President of Weldcoa
Consequently, AI has gradually been adopted by a small but growing portion of many global businesses over a wide range of end-markets, and behind these existing use cases, there is a host of other AI-based technologies that are being piloted that will disrupt and change markets even further.
Real-world cost effective applications of AI are now a reality in the industrial gas and welding supply industry that will help transform existing businesses processes and drive new levels of efficiency.David Schaer, President of Computers Unlimited
Outside our industry unmanned vessels that can stay at sea for months, robotic combat vehicles, enhancements to supply logistics, intelligence gathering, and many more use cases are being explored by the Defense industry. In the US, the Pentagon plans on spending upwards of $1B on AI related technology. The promise of self-driving or autonomous vehicles has been with us for over a decade and continues to be a focus of many startups and global car companies, although they’ve had some setbacks in recent years that have extended their timelines. But probably one of the most exciting AI developments in recent years is Generative AI. So, what is it?
Generative AI takes data, a user prompt/question, ongoing interactions with users (that help it “learn”) to generate entirely new content. It can produce original content in response to queries, develop blogs, sketch package designs, write computer code, or even theorize on the reason for a production error.
Generative AI technology is being viewed as impinging on the realm thought to be unique to the human mind…creativity. This latest class of generative AI systems has emerged from large-scale, deep learning models trained on massive, broad, unstructured data sets (such as text and images) that cover many topics. Developers can adapt the models for a wide range of use cases, with little fine-tuning required for each task. As a result, products like ChatGPT and GitHub Copilot, as well as the underlying AI models that power such systems as Stable Diffusion, DALL·E 2, GPT-3, to name a few, are taking technology into these creative realms. So much so, Microsoft in late January announced a “multiyear, multibillion dollar” investment in ChatGPT maker OpenAI, and stated it plans to integrate the technology into its Azure cloud service, and many of its products that touch consumers and companies around the world. If you haven’t experienced AI in action visit ChatGPT and ask it a question…you will definitely be amazed at the completeness of its response.
Narrowing our focus on the chemical industry which encompasses Industrial Gases, more than 80% of executives surveyed by IBM admit that artificial intelligence (AI) will have an immense impact on their business within the next three years. However, only 4 out of 10 chemical companies widely implement AI in their operations with the use cases employed including:
- R&D – support acceleration of the discovery of new chemicals and materials by analyzing large amounts of data to identify promising new compounds.
- Process Optimization – monitor and control chemical production processes in real-time, leading to increased efficiency, lower costs, and improved product quality.
- Predictive Maintenance – analyze data from equipment and predict when maintenance is required before equipment failures occur.
- Supply Chain Management – optimize the distribution and delivery of chemicals by predicting demand, optimizing routes, and managing inventory levels.
- Environmental and Safety Compliance – monitor production processes to ensure compliance with environmental and safety regulations.
Specific to Industrial Gases, the Tier One players have been incorporating some of the same AI capabilities into business processes for some time.
- Plant optimization – monitor, optimize and control industrial gas production processes in real-time. For example, in 2020, Air Products announced that it was implementing Aspen Technology’s AspenONE software solutions (which incorporates AI technology) to optimize its plant operations and improve process efficiency.
- Predictive maintenance – analyze data from equipment to predict when maintenance is required. Linde has been applying machine learning algorithms to years of historic sensor data from industrial plant equipment to predict malfunctions early and prevent costly downtime.
- Distribution Scheduling – AI algorithms are being used to optimize the distribution and delivery of industrial gases, reducing costs and improving customer satisfaction.
The Tier one players view AI as both evolutionary by augmenting existing processes and making them more efficient, as well as disruptive and transformational by capturing large amounts of data and using it as a strategic asset that provides competitive advantage. Air Liquide has shared that leveraging its data from 3.5 billion data points collected each day in 600 production plants across the globe, from 10,000 trucks and 24 million gas cylinders in circulation, plus the data collected from customers, along with AI, allows them to achieve both operational and service excellence for their customers and patients.
At Linde we continuously adopt emerging and mature AI technologies, to make better and faster decisions, improve processes and productivity, make our operations safer and more sustainable, and enhance customer experience.Dr. Dexin Luo Director of AI at Linde
Other segments of the industrial gas industry are also leveraging the power of Data and AI for the some of the above use cases on a smaller scale as well as some of the ones below:
- Salesforce Automation – via Salesforce, the #1 CRM company in the world, and its native AI platform called Einstein AI, it enhances sales processes including lead scoring, predictive opportunity forecasting, and automates workflows.
- ERP Systems – multiple platforms from companies including SAP, Oracle, and Microsoft Dynamics have integrated AI capabilities/features such as financial forecasting, demand prediction, supply chain optimization, predictive maintenance, fraud detection, advanced analytics, and real-time language translation, to enhance many enterprise business processes.
- Customer service – multiple companies offer AI enhanced platforms providing features such as automated task support, automated case routing, predictive next best actions, sentiment analysis and provide personalized recommendations to customers in real-time.
As a leading provider of integrated software solutions to the industrial gas industry, AI can help boost a developer’s coding productivity, improve cylinder production forecasts with machine learning models, or deploy new ways to monitor assets with computer vision.David Schaer, President of Computers Unlimited
Our history is built on leveraging emerging innovations that create greater productivity, safety, reliability and returns for our clients. We have leveraged technology to introduce some of the first commercially available automated systems used in the industry, and AI has the potential to provide additional ways to augment our automation. We are fortunate to have Mitsubishi as one of our technology partners, and recently, Gary Schueman, our Director of Automation hired his first AI programmer. Our automation team has been in communication with Mitsubishi to develop possible solutions that leverage emerging AI technologies that could assist with Product Life Cycle Management and possibly Predictive Maintenance on our gas cylinder filling systems and controls. It’s too early to go into detail but that’s what we are currently exploring.Hector Villareal, President of Weldcoa
Many of the companies across the industrial gas industry have leveraged these platforms to various degrees, and now have even greater and more cost-effective access to AI capabilities to improve and enhance their business processes.
Although AI adoption has gradually risen across all end-markets, there have been implementation challenges across the board as you would expect. Although, the exact failure rate of AI and data science projects is difficult to quantify accurately, it has been estimated that between 40-60% of the projects do not deliver the expected results or fail altogether. The reasons for failure can vary, but common causes include lack of clear goals or objectives, poor data quality, insufficient resources, and lack of technical skills have been the major contributors. These are some of the same causes that affect any transformational effort…so no surprise. But fortunately, lessons learned are being shared more and more and below are some of those best practices to employ:
- Strategy is Key – as with any transformation project/program it starts with business strategy and answering some fundamental questions. What problem(s) or pain point(s) are we trying to solve, what metrics are you trying to improve, what value does this capability bring to the business (efficiency, productivity, revenue growth)? Then start with a clear and well-defined use case that targets the specific problem you are trying to solve, using the benefits AI can provide. For some companies this may be a difficult step in “seeing the potential of AI,” so there may be a need for an external consultant/company to assist and support this step.
- Preparation & Readiness – educate yourself – you don’t need to know the execution level specifics, but you do need to know the business aspects such as recognizing opportunities and measuring success. Perform an AI Readiness assessment to help identify any foundational gaps in data, storage, culture, skills and budget. In the book, “The Business Case for AI,” by Kavita Ganesan, she calls these gaps the “5 pillars in preparing for AI.” You will probably need to revisit or refine your business case and AI roadmap during this step.
- Data is Crucial – The old adage “garbage in yields garbage out” is the same for AI projects. Getting your data in shape is one of the most critical steps in preparing your organization for AI. It starts with understanding what data you have now, the data your systems/teams are generating, defining what data is needed for your specific use case, and how do you get it and cleanse it. More often than not, a company does not have the data they need or it is not in the right condition for use, so either new data needs to be captured or existing data needs to be transformed and warehoused for analytics. High-quality, well-defined data is critical for training accurate AI models.
Capture high quality data from a business process, choose the right AI model or algorithm based upon the problem being solved, and integration into the underlying software application are some of the key elements behind a successful AI project.David Schaer, President of Computers Unlimited
- Training Patience – you arguably need 3 ingredients to train AI models effectively, high-quality data, accurate data definition and a culture of experimentation and patience. Small amounts of poor-quality data will quickly cause problems, so high-quality data is key. In addition, the data needs to be accurately defined, otherwise, the model will have no contextual guidance to help it properly interpret the data, let alone learn from it. For example, consider the importance of correctly defining images in a self-driving AI model (and the ramifications if done incorrectly) to tell the difference between signs, people and animals. However, errors are actually a valuable and normal part of the AI training process, and the key is to analyze the results for insights that will help you get to the bottom of what happened and why.
- Exchange & Communication – if the “what” is seeing the potential of AI and recognizing business opportunities, then communication and exchange around the topic is the “how.” Provide the means for employees to upskill and add to their subject matter expertise via learning materials and each other. Linde, for example, runs a digital exchange network called “dX” which offers a wide range of live sessions covering everything from AI demystification to showcasing projects and sharing best practice – from both internal and external speakers. It helps shine light on successful AI use-cases in one part of the business that could be replicated in another.
As shared above, education is key for any executive exploring new technology, and in my case if I wanted to understand the power of Generative AI, I had to start playing around with ChatGPT and understand its potential and limitations first hand. I also spoke to friends of mine who were educators and I asked them about their concerns with ChatGPT. Their solution to the potential threat of students cheating through their use of ChatGPT for assignments was wonderfully simple. They now have students write short essays during class.Hector Villareal, President of Weldcoa
Consequently, the technology is forcing us to make some adjustments to longstanding education and business “processes” as well…
Because of the where AI is in its maturity curve, there is a lengthy list of additional “best practices” that could be included such as the importance of selecting the right tech partner, use case selection, IT infrastructure options, ethical considerations, and many others. However, there are some key steps that are needed independent of the transformation being executed…executive leadership, strong project management, tracking progress, celebrating wins, and having a focus on continuously improving the output.
We have learned a lot on our AI journey so far, mostly importantly that we stay focused on answering the question, ‘what business value will be obtained from this use case?Dr. Dexin Luo Director of AI at Linde
Closing Comments & Outlook
AI has arrived and continues to make inroads into our personal lives and continues to gain traction in optimizing business processes as well. Many experts view AI as one of the technology inflection points similar to the birth of the internet and the advent of smartphones. As adoption gradually rises across all end-markets and a wide variety of use cases, we are starting to see the beginning of an exponential rise in AI impact on everything we do. If the frenzy created by ChatGPT and the Generative AI space is a precursor, the best is yet to come. The question is do you want to wait or are you ready “to stick your toe in the water” and start gaining a competitive advantage in the marketplace?