Lucky for you, we sat down with Global VP of Industrial Manufacturing at Mendix, Raffaello Lepratti. Raffaello knows a thing or two about evolving manufacturing technology. He has worked in manufacturing for 25 years, serving in various roles in automotive manufacturing, manufacturing operations management, and business development. He holds a degree in electrical engineering as well as a PhD in Advanced Human-Machine Collaboration.

Here’s what we discussed:

• AI prerequisites your product engineering department, manufacturing, shop floor, and supply chain need to consider
• Industrial use cases for AI
• What real-time decision should look like with AI

Getting started with AI

When getting started with AI, it’s important to consider your specific needs and your company’s readiness.

When it comes to AI in the industrial space, where do you start?

Raffaello Lepratti (RL): It’s all about focus. There is a lot of appetite around AI and how AI can be adopted in a way that, for manufacturers, leads to the purpose of technology. AI is one potential lever to achieve better results. In a certain industry, better results means faster time-to-market. For example, if you’re in electronics, you want to be the first to introduce your product innovation into market.  AI can help there. Streamlining processes, assisting design, development and operations to prevent or even predict issues. It lets you better leverage historical data and correlations across them.

But that thinking changes in another manufacturing industry, where being first to market isn’t necessarily the primary goal, but rather to do the right thing and provide its customers with the best quality and compliant product. Take aircraft manufacturing for instance. Quality and compliance are a must. There’s zero tolerance for failure. It’s a completely different approach to AI than, say, launching a new phone, which could have a few bugs that you resolve via updates. At the end of the day, it’s about knowing which types of processes you need and then figuring out how can AI support that.

So when you think of AI, you need to slice and dice what kind of priority you’re talking about, and make sure you are focused on your differentiating aspects.

What are the prerequisites industrial organizations should consider when beginning an AI initiative?

RL: There are three.

1. Data access and integration
2. Data quality and governance
3. A willingness to change

This is across all manufacturing industries.

Consider your systems: Is the data accessible? Do your core systems speak to each other?

You need to make sure that the data you need to focus on is consumable. AI has no meaning if the right data isn’t there. You need to have historical data, you need to have real-time data. Manufacturing execution system (MES) data or enterprise resource planning (ERP) data or product lifecycle management (PLM) data.

And it’s not just about the amount of data, it’s about the quality of data. Is it structured and contextualized? If it came from this sensor or this piece of equipment, for example. At which time did it come in and with what tools?

Finally, consider your organization. Is your company as a whole ready to introduce AI-based approaches? Will there be resistance because it takes employees out of their comfort zone? You need to consider if AI-decision-making in a certain area is appropriate. Consider specific areas of responsibility. That’s a lot you need to take your organization through.

How are manufacturers tackling willingness, data readiness, and data access?

RL: To be clear, you don’t need all three everywhere, all at once.

There are a lot of investments being made to investigate how AI can create an impact through higher productivity, better quality, lower costs, etc.

AI initiatives can start in a specific domain. So you don’t have to have every department in your organization immediately using AI. Knowing that every domain will not have the same level of data quality, the same availability, to begin with is important too.

AI use cases to start with

When considering use cases for AI, it’s good start with something that can bring immediate value and has a low technical complexity.

What is a good starter AI use case?

RL: Here’s an example of a starter AI use case: using AI to assist, in certain industries, shop floor operators. The operator is under time pressure. To do their job, they may need to consult documentation. For instance, safety instructions or instructions for an operator’s tool. That can take time. AI should assist this person in searching and providing the required information reliably and ideally faster. An AI copilot would be a great place to start here.  

It’s also a great place to start, because you can leverage the data of an MES for instance. It’s one tool. You can use your MES and have a copilot ask for information about what happened in shift X, what is the material for Y. Where is the material being used that I got from supplier Z?

And because there’s the genealogy of the product, the data is already structured in a way that a copilot can use it. It’s a good use case. And valuable.

AI and real-time decision-making

Successful real-time decision making is dependent on your specific industry. So be realistic with your expectations.

How is AI going to change the shop floor from a place of execution to a place of real-time decision-making?

RL: AI is often talked about with speed. But I think it’s important to set expectations. Having a reliable real-time decision-making process is probably a vision for many companies. But keeping the human in the loop is always important. Because there are also risks attached to that. In my opinion, technology evolves, but the level of trust people have will not evolve at the same speed. So I don’t think it’s going to happen in the way that probably we define real-time.

In a highly regulated industry, you might not want an AI agent deciding what to do. But having a real-time suggestion because you can foresee a delay helps the human in the loop make the right call at a time that’s appropriate for your industry.

How Mendix and Siemens Xcelerator help you get started in AI

20% of organizations have cited data, integration, and quality and availability are the biggest challenges in leveraging AI. Another 13% cite cost and resourcing. 11% say employee training and change management are their biggest challenges.

How do you see Mendix and Siemens Xcelerator helping address AI challenges?

RL: Mendix being part of Siemens Xcelerator already provides an integrated value proposition based on a rich set of capabilities and connectivity. These give developers the ability to access high-quality data from Teamcenter, Opcenter, Capital, or Polarion and use them to create smart industrial applications. With Mendix, we help manufacturers define their roadmap to create a Digital Thread of applications that support the specific requirements of their processes, in engineering or manufacturing.  

This makes it easier to depict the roadmap to really implement AI in a broader scale.

See how Mendix customers are creating as much value as they can with AI.