AI and Machine Learning Bring CE-Grade Manufacturing and Supply Chain Management to Any Industry

The bitterness of poor quality remains long after the sweetness of low price is forgotten. But the reality is that there are deadlines. It could be a holiday, a design refresh calendar, or an initial release; there are a number of motivating factors that drive companies to get products out the door at the time they do so. Sometimes this means outsourcing key functions like manufacturing—even if it entails concessions on cost, process visibility, and possible delays. Other times, technology fills the gap. This is where companies like California-based startup Instrumental come in: Instrumental offers an Artificial Intelligence- (AI)-driven solution to help find and fix problems hidden to the human eye—as they occur.

Major manufacturing firms like Foxconn, Flex, Pegatron, and Goertek are already on board.

Instrumental’s solution comprises of camera-equipped inspection stations that take images of products at key stages of production for remote inspection in order to improve and automate anomaly detection. The company refers to this capability as their “Detect” feature and attributes its enablement to the use of machine learning. A logical evolution from here is to include “Inspect”- and “Act”-type features, though doing so would require deeper integration with supplier networks; whether or not this will happen remains to be seen.

In the hardware market, the design of a manufactured product is generally demarcated by three phases of development after the completion of conceptual and/or prototype work:

• Engineering Validation Test (EVT): Can several units that operate as expected and that meet all functional requirements be built?
• Design Validation Test (DVT): Can several units that operate as expected, that meet all functional requirements, and that are aesthetically acceptable be built?
• Production Validation Test (PVT): Can many units that function as expected, that meet all functional requirements, and that are aesthetically acceptable be built with industry-leading manufacturability?

Software feature engineering is a bit different: teams often build an app, deploy it to a limited number of servers, and then check for obvious issues before proceeding with a larger rollout. This introduces the chance for human error and the potential for bias (regardless of intent). Although most companies run automated tests to check code for glaring errors, the sprawl and complexity of today’s connected endpoint infrastructure makes the jump from pilot to production environment that much bigger. Here is where the advent and layering of microservices can end up interacting with each other in unexpected ways. Here is also where AI can be a big help.

Closing the feedback loop is a multifaceted endeavor. There is internal feedback from individuals, teams, departments, regions, and business units; direct external feedback from a client, customer, or partner; and indirect external feedback, such as from market observation. The cloud brings the internal-external and digital-physical feedback loop(s) full circle by replacing reactive workflows with their proactive better half. However, there are times when it can be unexpectedly useful to disseminate different information to different groups, teams, or individuals.

A new product launch is a great example; different teams have different priorities, and cross-functional coordination is key. Marketing needs to know the public feature release date, since that is when all associated marketing collateral needs to be ready; engineering needs to know the feature deployment date, since this milestone precedes a public launch; while accounting only needs to know whether to record bookings vs. sales.

Early production is one of the most critical stages of the product life cycle. This is generally when the quality of the product (e.g., Ingress Protection [IP] Code rating) and delivery schedule is determined. Solutions from companies like Instrumental have broad applicability across a number of vertically oriented Industrial, as well as general-purpose, Internet of Things (IoT) applications, but to see the benefits of AI in action, industrial, and manufacturing firms must:

• Articulate the governance model: The nature of work performed and the needs of the situation (urgency, ability, availability) are vital determinants of success in the Robotic Process Automation (RPA) planning and prioritization process. The role of governance is to assess such opportunities and provide guidance around the assignment of tasks, roles, and responsibilities (accountability, field risk control, enablement). The nature of IoT projects and dependencies generally puts this function between business and Information Technology (IT), where there is a growing need to wrangle change management issues.
• Skip the Bureaucracy: Pitch money-saving apps and devices directly to the people that use them to bypass skepticism with buy in.
• Find alignment: Aligning on key performance indicators (KPIs) beforehand, regardless of AI maturity, is essential. If an app or pilot project achieves its target KPIs/Return on Investment (ROI), then the company should commit, scale, and implement it companywide. It is this commitment—even, if not especially, in the face of polarity—that keeps businesses moving.