This just in: A new global forecast on the market for artificial intelligence (AI) in manufacturing predicts nearly a 40% compound annual growth rate during the period from 2019 to 2027, reaching $27 billion by the end of the forecast period.

While the road to the smart factory might seem more readily traversable for major manufacturing companies, small-to-medium enterprises (SMEs) take note: There are multiple on-ramps, particularly in an era where advanced technologies are available in forms such as software as a service (SaaS).

Figure 1: The road to the smart factory contains multiple on-ramps.

Without question, the growing deployment of the industrial internet of things (IIoT) is a foundational technology for incorporating AI into manufacturing. Factory floor devices capable of collecting, sharing and responding to actionable data can be seen as the body for which artificial intelligence is the brain.

This interconnected relationship can manifest itself in numerous aspects of enterprise resource planning (ERP), and one example is asset management. Utilizing a digital platform to monitor the health of assets unlocks the potential for predictive maintenance to extend their life cycles. Integrating machine learning and AI into this system, as IBM does through its Watson IoT tool Maximo, enables it to become self-learning over time.

There are also software tools focused on the supply chain, such as Microsoft’s Dynamics 365. The platform provides a high-level digital overview of production — allowing a supervisor to view material availability before releasing an order to the factory floor, for instance, or a planner to adjust schedules to work around material or resource shortages. Within this context, machine learning can be configured to provide guidance. In the bigger picture, overall production can then be optimized through a digital feedback loop that responds to market fluctuations.

AI can be used in manufacturing for the product as well as the process. An example is machine vision, which can be used on the production line to spot microscopic defects too small for the human eye to perceive. A visual inspection tool from Landing AI, the Silicon Valley firm started by Google Brain co-founder Dr. Andrew Ng, can distinguish between true material defects and acceptable anomalies such as scratches and air bubbles.

Among the most advanced applications for AI-based manufacturing is the digital twin, which is essentially a data-driven virtual representation of a physical system. Digital twins can be used for maintenance, repair and operations (MRO) of equipment beyond the physical proximity of human operators. NASA was an early adopter in the 1970s, using the concept to make corrective decisions for spacecraft based upon measurements designed to mirror the inevitable transformations caused by harsh environmental exposure. More recently, the digital twin has been embraced for supply chain optimization — and also for its potential to transcend the atmospheric danger of exposure to COVID-19.

This short list of possibilities for AI in manufacturing is really just the beginning. It is important to note, however, that the daunting prospect of a total manufacturing process overhaul need not be a deterrent to making progress in small stages. As IBM suggests in its business operations blog for those just getting started: “Start s