How AI is (and isn’t) meeting manufacturers’ supply and demand

July 31, 2020

AI in manufacturing is growing more prevalent, yet tremendous untapped potential remains.

Artificial intelligence in manufacturing is growing more prevalent, but there remains tremendous untapped potential. In a May 2020 report, McKinsey identified four main trends that are transforming the manufacturing industry and highlighting the “need to augment workers with technology”: retiring baby boomers, regionalization, the proliferation of shop-floor data, and the ongoing COVID-19 pandemic.'s Pedro Alves

Currently, many manufacturers have a strong desire to adopt AI and other automation tools, but lack the preparedness to utilize them to their full potential.

In evaluating progress on AI deployment up to this point, here is a summary of some of the more promising AI areas where manufacturers can now jump in, as well as some of the key areas needing major improvement where manufacturers should wait or exercise caution.

Promising AI applications: an opportunity for manufacturers to enter now

Manufacturing is well suited for computer-vision applications, like inspecting machinery for defects or optimizing products. Manual testing is slow and expensive compared to testing using automation. Employing the proper sensors, a robust AI platform collects data and offers the company real-time readings to help limit stoppages and inefficiencies. By expanding their AI capabilities, manufacturers can uncover additional use cases (like adding machine learning to smart robots) to improve performance, integrating consumer data into product development choices, and uncovering bottlenecks in production lines.

For manufacturers starting the AI journey, data should be a guiding point throughout the process. Manufacturing firms that continually build their data sets without throwing any of the data away are the ones best positioned to improve and grow at scale. Manufacturers should also carefully consider their operational needs before hiring a data-science team. New personnel is costly, and some AI efforts can be achieved more cheaply with the help of proven software tools. That said, whenever a project is covering uncharted territory and has higher stakes for the organization, involvement from a human AI team—in conjunction with software tools—may be required.

AI applications in need of improvement: an opportunity for manufacturers to watch and wait

Predictive modeling and maintenance, critical areas where manufacturing AI is required, needs an array of sensors and a full understanding of past conditions and the likelihood of future occurences in order to be successful. This can be problematic because sensors were built for human monitoring, not for machine-based monitoring. Making improvements to the data and design of sensors enables manufacturers to better transition into AI.

Error detection is another area for AI improvement. It is a critical, yet difficult, undertaking, especially when manufacturing parts that require precisions. A July 2019 McKinsey report found that AI quality testing could raise detection rates by up to 90%, as compared to human inspections. With continual fine-tuning, AI models have the capacity to spot irregular sensor signals that are not visible to the naked eye. 

Forecasting efforts also need to improve, and AI/machine learning has the potential to prompt this improvement. Currently, pricing and demand forecasting are typically completed through simple statistical models with limited data. Using AI technology, firms can augment their forecasting models with more data and significantly improve their forecasting accuracy for greater efficiency and simpler scaling.

Integrating AI into the future

While there is some initial success worth acknowledging, manufacturers are still taking on a more reactive stance to company and industry problems. If manufacturers don’t shift to a more proactive, AI-integrated stance, they will see more challenges in meeting supply and demand, and face limitations to their overall success and profitability in the long term.

When there is a technological advancement, the market leaders and early adopters willing to take on risk overcome the biggest hurdles. The first groups adopting are the ones that will reap considerable benefits and help to define the industry, potentially for decades to come.

Pedro Alves is CEO of