“So, how does your platform work?,” you ask the artificial intelligence (AI) Industry 4.0 vendor.
“Well, our agnostic AI technology feeds on all of your data sources in real time, and generates predictions, warnings, prescriptions, and recommendations that help you optimize your yield and throughput, while minimizing your losses.”
He continues with a slick visualization showing streams of bits flowing gracefully into a cloud, with valuable business insights coming out the other end.
“But what exactly are you going to solve for us that we were not able to solve with our existing models?”
“Most of our data isn’t really that good, how are you going to predict and optimize the process without understanding the fundamentals of the process?”
“Our process changes quite a bit, rarely running the same way for more than weeks at a time. How can your platform handle that?”
“Can this integrate with our existing process control or monitoring systems?”
“How can our engineers and operators build trust and accept this platform if they don’t understand the logic behind it?”
“Have you actually done this on a facility such as ours? What are the use cases?”
These are just some of the questions that arise as you dive deeper and begin making decisions about your path forward. If this AI vendor serves a wide spectrum of industries, you might not be satisfied with the answers. The vendor offers a demo of the AI technology on your historical process data: “Just an offline data exercise, what have you got to lose?”
In case you do move forward with the demo, chances are you will find yourself or your busy colleagues spending hours on end teaching the vendor application engineer the intricacies of your specific manufacturing process. When the vendor circles back with their findings, to your disappointment, the results do not provide you with significant new insights. The results might not exceed the capabilities of your current linear empirical models or rigorous kinetic models. Some industries might find this vendor useful for simpler applications, where analyzing process data and building models is not very common, and low hanging fruit are abundant. But that’s not your industry, and it’s certainly not your company. You are looking for a much more accurate and deliberate approach to employing AI to optimize your complex process. What are your options?
Build it yourself
One option is to build AI capabilities internally and combine them with your internal manufacturing process understanding, building tailored applications to solve specific problems. While this may be done with homegrown data science groups, even the largest companies will try and avoid building their own AI products. It is difficult for a large corporation to compete with an external data science company that doesn’t employ thousands of people who run processing plants. Most in-house products lack a long-term roadmap and support when the developers switch roles.
A dedicated partner
Another option is an AI vendor who has dedicated itself to you and to your industry. This vendor understands that in order to solve the most difficult problems that you and your team left unsolved, it takes more than state-of-the-art AI. It takes a carefully curated combination of state-of-the-art AI and world-class expertise uniquely focused on your specific process.
This vendor has built a dedicated team of application engineers who worked in your industry for years, in companies and plants just like yours. They will understand exactly where your most valuable and difficult challenges are. What tools and models you have at your disposal, and what problems are worth going after. These folks speak your language and know your process like the back of their hand. They know how to align the right stakeholders in the plant around the right common objectives. They have done it before. This vendor knows how to treat your noisy and missing data, integrate the process first principles into the model, and verify the correctness of relationships that the model has learned between different variables. Since this vendor is focused on your industry, you are benefiting from the massive experience they bring, and will continue to bring, from applying the AI platform on plants just like yours.
You are no longer working with just a vendor. You now have a partner, that will guide you with confidence through a structured and methodical procedure to develop models, and safely deploy them while fully respecting and understanding your process control environment. They will work with you on the management of change (MOC) documents you need and provide you with operator and engineer training materials. Your partner will work with you to commission your application, prove its economic value, and stay by your side to help you sustain the value for years to come.
Yes, it is actually possible to navigate the sea of Industry 4.0 AI vendors and ambiguous answers, to find the right solution for your company. You will know when it’s the right fit when you find a partner that is dedicated to your domain and speaks the same technical language as you.