How Microsoft is prepping manufacturers for Industry 4.0

At a contract manufacturing facility in China that makes Microsoft consumer hardware, production lines are connected to the Azure cloud to give the technology giant full visibility over the movement of raw materials and finished products on the factory floor.

After the products – including Surface devices and Xbox gaming consoles – have been shipped, Microsoft tracks their movement across its supply chain, including the containers and ships they have been loaded onto. And while the goods are en route to their final destinations, weather data is being pulled to predict shipping delays and reallocate inventory on the fly.

The back-end system, along with artificial intelligence (AI) capabilities, that powers Microsoft’s supply chain from factory floor to store was developed in just eight weeks, according to Scott Hunter, regional business lead for manufacturing at Microsoft Asia. “It’s no longer a three-year implementation timeline for an enterprise resource planning system,” he said.

Hardware makers like Microsoft are the envy of manufacturers that have been struggling to keep costs low amid shrinking profit margins. With more visibility over the factory floor, the theory goes, manufacturers can reduce material wastage, manage inventory levels better and fix factory equipment before it breaks down to minimise disruption.

Although the use of AI is key to realising the benefits of these so-called smart manufacturing processes in what is known as Industry 4.0, 59% of manufacturers in Asia-Pacific have not adopted AI as part of their business, according to an IDC study commissioned by Microsoft.

“This is a worrying sign for the industry that needs to thrive on innovation,” said Hunter. “To achieve supply chain excellence, and even develop new business models to address changing customers’ needs, integrating AI for their business is a must. Organisations that fail to adopt an AI-first strategy risk being left behind in today’s competitive market landscape.”

Indeed, Asia-Pacific manufacturers lag behind those from other industries in the region in AI strategy, according to IDC’s study. Hunter said that although retail and financial services firms are embracing AI to keep pace with growing consumer demands, manufacturers have been shielded from such pressure for some time.

But that is changing with the likes of Nike, which is enabling individuals to design their own sneakers, prompting manufacturers to look into ways to handle “a lot size of one” to be competitive, said Hunter.

Building on its manufacturing and technology knowhow, Microsoft has teamed up with industrial firms to support the growing demand for smart manufacturing capabilities.

For example, renewable energy giant Siemens Gamesa is streamlining and automating how technicians inspect and maintain industrial wind turbines by migrating its autonomous drone-digital process to Microsoft Azure and infusing it with Azure AI.

Integrating Azure AI services will help to speed up the process, with image recognition reduced to 34 seconds, rather than four to six hours using the manual method, which also was prone to errors, according to Microsoft.

In the optical space, Zeiss teamed up with Microsoft to develop a spectroscopy measurement and data-sharing capability that uploads production data from spectroscopic analysis to Azure.

With this capability, food manufacturers, for example, can get insights about their products, such as the amount of fat, moisture, salt content, directly in the production line, enabling production managers to react immediately to process variations.

Acknowledging the importance of ensuring that AI models factor in safety considerations since any wrong decision can be devastating on the factory floor, Hunter pointed to one of Microsoft’s customers, Toyota Materials Handling Group, which is training intelligent forklifts to navigate a virtual warehouse rather than a physical one before deploying them on the ground.

Another way Microsoft is addressing safety concerns is through a recent acquisition, Bonsai, which uses machine teaching to abstract the low-level mechanics of machine learning, so that subject matter experts, regardless of AI aptitude, can specify and train autonomous systems to accomplish tasks.

“When tuning AI models, we find that the bigger the dataset and the more we can train, the safer we’ll be,” said Hunter. “There are parameters to ensure we don’t come off the rails and do something crazy.”