© 2019 SourceMedia. All rights reserved.

5 key questions when applying AI to industrial processes

Industrial processes stand to benefit from artificial intelligence and machine learning to an equal or even higher degree than the data-driven processes in other parts of the organization. But there is still much confusion as to how these technologies should be implemented and exactly what outcomes can be expected.

A good way to view these developments in relation to industrial processes is that AI is the goal and ML is the means to get there. ML introduces the capacity to learn how to monitor the performance of key tasks and then to continually improve it over time. Once that is done, you have the makings of an intelligent process that can leverage a range of tools, such as predictive analytics and big data capture, to optimize the entire system.

One of the key objectives of industrial operators is to connect asset data with financial metrics. This results in the ability to achieve predefined business outcomes, such as increased productivity, lower unit costs, and other means to foster steadily increasing profit margins.

industrial processes.jpg
An attendee working on an Apple Inc. laptop computer participates in the Yahoo! Inc. Mobile Developer Conference Hackathon in New York, U.S., on Tuesday, Aug. 25, 2015. The Hackathon is an opportunity for mobile developers to come together and hack around the Yahoo! Inc. Mobile Developer Suite. Photographer: Victor J. Blue/Bloomberg

But getting there will not be easy. For a smooth transition leading to successful results, executives should be ready to answer the following questions:

What are the specific business issues I want AI to solve?

For far too long, businesses have been deploying technology for technology’s sake. Quite often, this leads to a situation in which business goals were then defined according to the limits of that technology. AI, ML and other intelligent technologies have the propensity to impact processes in all kinds of ways, so to avoid losing control altogether, it helps to know exactly what you hope to accomplish.

What is the path to a successful implementation?

Once you know what you want AI to do, you have to plan how you want it to do it. If the goal is to increase production, what aspects of the pipeline are creating the limitations? Is it materials sourcing? Bottlenecks on the assembly line? Distribution? Establishing concrete, meaningful goals goes a long way toward implementing the broader transition to overall system intelligence.

Do I have access to the data needed to support the proposed project?

Without the right data, even the most intelligent system cannot deliver meaningful results. A properly functioning analytics engine requires three key attributes to the data it receives: quantity, quality, and access. The more data sources present in a system – whether they are live-stream sensors, historical data stores, or cross-sets of multiple points – the more certainty it can apply to the decision-making process.

Do I have a combination of data scientists and subject matter experts in-house?

Few AI/ML roll-outs are successful without strong collaboration between data scientists and the subject matter experts who understand the target processes. This is a rare combination in today’s industrial enterprise. It’s probably something that should be addressed right away, and it should be made clear that both are responsible for creating successful outcomes.

Do I need a solutions provider to assist with project execution?

In-house talent should not obscure the need for an independent voice to help evaluate and execute the deployment. This can range from a specialized analytics firm or a consulting engineer, to an all-in-one provider capable of end-to-end support. The key is to find someone who has experience with cutting edge technologies and can provide services within a reasonable budget.

It should also be noted that AI and ML lend themselves to a more continual roll-out than traditional systems that are deployed once and then upgraded periodically over time. This requires a different managerial mindset in which constant evaluation and refinement take the place of hands-on operation and maintenance.

In this way, AI and ML represent more than just new technology, but a cultural change that will lead to new job responsibilities, new relationships, and perhaps an entire re-imagining of the business model itself – all of which are necessary to compete in the 21st century economy.

For reprint and licensing requests for this article, click here.