Industry 4.0 and the digital transformation of process engineering and manufacturing technology have kicked off a development that is making today’s production markedly different from that of times past. While product life cycles have become shorter, the degree of individualization is rising—in other words, there is greater variation and in some cases, customers can even customize products themselves. At the same time, the production of a “batch size of 1” should not cost more than mass production; this represents an enormous challenge for machine manufacturers and automation specialists.
In order to meet this challenge, machines must become more and more flexible. Artificial intelligence makes machines “smart”: It provides them with additional capabilities and allows them to react to new requirements and conditions with greater flexibility by learning from the accumulated data. In addition, AI can help simplify machine operation, optimize processes, and increase product quality.
This increases the competitive pressure and accelerates the trend towards the use of AI in control and automation engineering— and it means that those who want to prepare their business for the future must take on the issue of artificial intelligence.
To help you get started with the new technology, this article presents a structured approach based on our experiences (and failures) with AI projects at KEBA.
Step 1: Develop the big picture
The first step is to define the initial situation. How is the market changing? What are the technology trends that will dominate the next two to five years? What are the machine manufacturers and automation specialists doing? And above everything: What do the customers want? How are their strategies and business models evolving, and what does that mean for their requirements? Talk to your customers and ask them about their plans and strategies for the coming five to ten years in order to gain a good overview:
- Are there changes to the business model?
- How will production change from their point of view?
- What (new) requirements does this entail for their machines and plants?
- What are their actual pain points, and which of these can (only) be addressed through AI?
- Which trends do they consider important?
It has in fact become more difficult to assess upcoming market shifts because the rate of change has accelerated significantly. But even with the uncertainty of assumptions, it is still worthwhile to systematically work out a “big picture”.
Step 2: Identify the influence of AI
The next step is to figure out where in the big picture the effects of artificial intelligence will become relevant. What role does AI play in these trends? What areas are relevant for AI? For example:
- Machine operation: simplification through assistant systems
- Intelligent enhancements of machine control and functionality
- Cloud solutions for data analysis
Anyone still without in-house AI resources should use this step to acquire basic knowledge of AI, for example by attending industry events on the subject, by studying best-practice examples, or by obtaining support from more experienced partners.
Step 3: Define your own position
The next important step is to consolidate the knowledge gathered so far and apply it to your own situation. Use the big picture and the identified AI trends as a basis for finding answers to questions: How will the market changes affect your own business? What role can you—or do you want to—fill in the future?
What are your own capabilities with regard to data availability? What data is available within your company, what data can be generated for applications, and what data is definitely out of reach? Since AI is data-driven by definition, the answers to these questions define an important, even essential component of your own position.
Put in more practical terms: What can you do to support the changing technologies and strategies of your customers? How can you use digitization and AI to adapt your products and services to better fulfill your customers’ requirements?
Step 4: Identify AI potential and action fields
The next question is inevitable: Which AI aspects are relevant to you? In what areas can you—or do you want to—become active? The answers depend, at least partially, on what in-house competencies are available. You should take into account that AI solutions need both hardware and software.
Completing this step successfully requires uncompromising focus. AI technology in its totality offers an enormous wealth of options. But individual applications will be successful only if they solve specific problems.
Flexible machines and plants rely more and more on AI, and this development continues in the move towards the Smart Factory
Step 5: Develop your business case
Becoming active in a field is one thing—achieving business success is another. And so the next step is to develop the business model and define the future business strategy. The following questions can be helpful:
- How can AI applications be integrated into your own business model?
- Is it possible to offer new, additional data-driven services?
- What are the benefits with regard to cost?
- Where are there opportunities for additional revenue?
Sometimes, the use of AI solutions is simply unavoidable in order to remain competitive and not lose existing revenue!
Step 6: Create the right conditions
Now the time has come to turn plans into realities. What do you need in terms of skills and resources? How can you create AI competence in-house? Data scientists in particular are needed in order to develop data-driven applications and services.
A surprise may await you at this point: Frequently, the sought-after competence does exist within the company—but nobody knew! Younger employees in particular frequently have had the necessary training, but they have been assigned other tasks because previously, there was no need for AI expertise. Another important requirement are employees with in-depth domain knowledge who can function as the liaison, supporting the work of the AI and data experts by contributing subject-matter expertise on processes and workflows. Such interface positions, too, can frequently be filled with existing in-house personnel.
However, if no such experts can be found in-house or if the ensuing gaps cannot easily be filled with new hires, another consideration could be to gain the necessary skills through partnerships and cooperative ventures. A word of warning though: Never rely on external support alone! Rather, use the cooperation as a learning opportunity; foster and train your own employees, and use joint projects to build up the practical know-how you need.
Step 7: Find a cooperation partner, if needed
Working with external partners can strengthen your company by providing additional resources. Possible cooperation partners include other companies, but also for example research institutes. When choosing a partner, think about the following questions:– What competencies do you need? In addition to product development, also take into account product maintenance and subsequent customer support.
- What competencies can be integrated best?
- What competencies does the partner contribute in terms of hardware and software?
- Does the partner also come with application competence?
- In addition to best-practice examples, does the partner also provide examples of projects that did not work out? What changes were made as a result?
The AI sector is brimming with newcomers and start-ups whose innovative solutions are driving the market. While they do represent interesting cooperation partners from a technical perspective, the question must be considered whether they are in a position to provide a competent and reliable long-term partnership with a high level of quality. Start-ups come with a twofold risk: A lack of success can cause them to fold, and good success can result in an acquisition. KEBA for example lost an interesting cooperation partner after just half a year because that company was acquired by a competitor.
Furthermore, experience has shown that newcomers might be able to build impressive demos and present brilliant ideas, but the implementation does not live up to the requirements of customers in the industrial sector. The latter are looking for solutions that can be supplied for at least five to ten years, that will be maintained and further developed during this period, and that are based on the original hardware installed in the field.
This last issue in particular is a frequent sticking point in AI projects. Whenever neuronal networks and algorithms become hardware-embedded, each new generation of the solution frequently entails a change of the technical basis—an absolute no-no for industrial customers. This is one of the reasons why the machine-embedded use of AI has been slow to gain momentum.
With regard to potential cooperation ventures, this is a criterion that leaves little wiggle room for compromise: Either the partner acquiesces to the conditions and restrictions of the industrial sector—or machine manufacturers and automation specialists themselves must take responsibility and create solutions that work for industrial uses.
The same applies to data analysis cooperation ventures, whether with large cloud providers or with platforms offered by automation specialists or a machine manufacturer consortium. Each AI solution should be based on an independent architecture so that users can freely select their platforms; there is already a broad range of options to choose from: public clouds or private clouds, on-premises installations or edge computing, which has met with great interest in particular in AI and 5G.