Manufacturers worldwide spend hundreds of billions of dollars annually on quality inspection, with estimates for quality-related costs reaching a trillion dollars when prevention, appraisal and cost of failure are included. For U.S. manufacturers, the bill for inspection and related activities is plausibly around $100 billion. Factor in other supply chain nodes, and aggregate cost only grows. So, what happens if most of that cost goes away?

Will quality inspection be the first area to undergo major artificial intelligence-driven transformation? In recent years, we’ve witnessed remarkable improvements in visual analysis. AI’s capability in this respect far exceeds that of humans. In fact, human error is often why we need inspection to begin with. Many manufacturers minimize manual inspection to avoid introducing fresh problems. Achieving Six Sigma quality levels with purely manual inspection is virtually impossible.

In the late twentieth century, manufacturers learned from Japanese quality thinking to build quality into processes, not just test at the end. Diligent supplier management was followed by detailed incoming inspection to ensure that all product in stock was conforming. In the future, this will not be done by adding human checkpoints, but by using systems to ensure compliance. In manufacturing, inspection can be done by inserting electronic measurements in the process. In fulfillment, systems already check that addresses are valid upon entry, or that the correct number of serial numbers are scanned; yet more proactive system-initiated inspected is possible. In receiving, inspection can be done electronically.

But if humans aren’t the primary line of defense, the question becomes: Where, exactly, should inspection occur? The answer isn’t simple. Using sensor and imaging technologies, we can look inside closed boxes, for instance. Moving to 100% AI-driven inspection throughout fulfillment could, in theory, bring error rates to a minimum and dramatically reduce returns and customer dissatisfaction.

AI opportunities go beyond quality inspection. By inserting preventive measures in manufacturing, yield will increase, waste will decrease and rework will mostly disappear. In production, AI can assess the product, but equal value comes from assessing and refining the production system, such as line set-up, tool wear, temperature drift and material inconsistencies. This leads to stronger prevention. The system can adjust setting and process parameters, fundamentally eliminating waste.

Rethinking quality could yield huge savings. The cost of quality, broken into preventive, appraisal and failure categories, is generally assumed to be a whopping 5% to 20% of revenue. Appraisal cost includes inspection. By extrapolating appraisal cost percentage of manufacturing cost, one can estimate aggregate cost. If we aggressively drive down appraisal cost, it demands the question: Can AI-aided inspection drive a massive 4% out of cost of goods if deployed throughout the supply chain?

More exciting from a supply chain standpoint is the application for one-off-processes, including fulfillment, stacking, loading, slotting, cycle counting and a multitude of other functions. Historically, inspection to achieve higher accuracy in these areas ran on a manual, “throw more bodies at it” effort. Today, our answer should be to throw more compute power at it.

AI inspection will represent a dramatic staffing shift, moving from inspectors to facilitators. No longer looking for random errors, the inspectors will be tasked to spot systematic errors and prevent programming errors. Quality teams will need to mitigate new risks: over-reliance on algorithms, bias in AI data and cybersecurity vulnerabilities. This will demand a dramatically different skill set. We’ll still require quality inspectors to meet regulatory requirements, which are always slow to change. Yet as 90% of quality inspectors go the way of typesetters and lamplighters, we must offer upskilling and reskilling.

What does this mean for you, if your competition is already deploying these strategies, but you and your suppliers are not?

Hannah Kain is president and chief executive officer of ALOM.