AI Strengthens Smart Factories To Overcome The Challenge Of Product Complexity

Bottom Line: 2020 roadmaps are dominated by AI-based, configurable products that are the foundation of new services-based business models, making smart factories’ flexibility and speed the most vital link in any manufacturer’s’ future.


Manufacturing CEOs I’ve spoken with are in unanimous agreement that the best way to drive new revenue growth is by transitioning to more services-based revenue models based on next-generation products. Their product roadmaps include configurable, customized products capable of delivering data back to manufacturers they can monetize as services. Manufacturers are looking to get beyond relying on transaction revenue alone. They’re mostly focused on how they can use configurable products to launch higher-margin outcome-based business services, and their 2020 roadmaps reflect this goal. The following graphic from McKinsey’s Leveraging Industrial Software Stack Advancement For Digital Transformation (50 pp., PDF, no opt-in) explains why manufacturers’ 2020 roadmaps are dominated by more configurable products capable of driving new services-based business models.


How Amazon designed and launched the Alexa product family has useful lessons for manufacturers selling customized, configurable products. These lessons include the following:

  • Designing the base product or platform from widely available components to avert backorders and ensuring electronics are upward compatible from one product model to the next.
  • Finding innovative ways to turn customer interactions with products into valuable feedback.
  • Having a product design flexible enough to add security controls in response to customers’ privacy concerns.
  • Starting with the design goal of having Alexa drive services revenue, expecting minimal profit from the device itself.
  • Define data governance, data structures, and reporting to capture intelligent products’ real-time insights.
  • Give every customer the freedom to customize the product to their unique preferences and stay current as their needs change.
  • Amazon’s Alexa is a prime example of what happens when a configurable, customized product begins generating more intelligent data than the production process used to create it. By integrating PLM, ERP, MES, and CRM systems with a real-time view of each Alexa product, Amazon has averted the “dumbing down” of their production operations and enabled manufacturing to scale with the increased intelligence of each successive product generation.
  • Refusing to allow their production systems to be “dumbed down” by a lack of integration across PLM, ERP, MES, and CRM systems, Amazon aggressively takes a lifecycle view of Alexa. Their goal is to provide value to customers over the lifetime they own the device. The recent Oxford Economics study, Smart, Connected Products Manufacturing’s Next Transformation reflects three tiers of progress manufacturers are making concerning device-level intelligence. Tier 1 is the most prevalent type of product intelligence configurable, and connected products are capable of today.
  • Giving customers the freedom to use APIs and apps to customize and scale products can enable network effects not attainable by a single manufacturer alone. Alexa was designed for a low production cost per unit with services revenue driving the business case. As a result, there are over 90,000 skills available on Amazon Alexa from external developers. Amazon introduced health and wellness capabilities for developers and announced a HIPAA-eligible environment for developers with new healthcare skills from Express Scripts, Cigna, and Boston Children’s Hospital. Alexa’s product-based services strategy is configurable enough to scale across multiple vertical markets with success, as is happening in the automotive industry today.


Smarter, more configurable, customized products dominate manufacturing roadmaps for 2020. Manufacturers need to bridge the gap between manufacturing operations data and the exponentially increasing intelligence of the products they produce. The following are initial strategies they are undertaking to close the gap, relying on AI and machine learning to accomplish each:

  • Capitalize on machine learning’s strengths to solve the constraint challenges of improving production efficiency while defining the best possible production workflows for highly customized, engineer-to-order (ETO) products. Supervised machine learning algorithms excel at solving constraint-based problems, making their use ideal for maintaining production efficiency while producing highly configurable, customized products. 
  • AI and machine learning need to be the glue that enables PLM, ERP, MES, and CRM systems to share real-time intelligence on product models and across product lifecycles. Instead of allowing each of these systems to run at their clock speed or cadence, consider how AAI and machine learning can better synchronize them, creating a more accurate, actionable product model and view of product life cycles than ever before.
  • Instead of relying on traditional and often brute-force MES production scheduling, machine learning can optimize production schedules down to the work center, operator, and material. In 2020 and beyond, manufacturers are not going to have the time to rely on brute-force production schedules that can become trial-and-error fast for ETO products. The more configurable a product, the more a manufacturer needs machine learning to optimize their production scheduling.
  • Machine learning shows the potential to manage ETO products’ Bill of Materials (BOM), workflows, variations in work instructions and quality levels better than traditional manufacturing systems. Only by having a system that can track a lifecycle view can manufacturers hope to get to the level of efficiency they need to make ETO strategies financially successful.
  • All manufacturers need the next generation of metrics that provide more insights into why production efficiency, quality, and yields vary, and machine learning can deliver them. Overall Equipment Effectiveness (OEE) is useful for stabilizing machinery by evaluating availability, performance, and quality. Yet OEE can also be a constraint on manufacturing by hiding more significant quality and yield problems. Amazon and the contract manufacturers they work with are using machine learning to troubleshoot efficiency, quality and yield issues to understand why each varies. They’re getting beyond OEE to see root cause analysis with machine learning.
  • As Amazon’s success shows, the more PLM can capitalize on the intelligent data products produce, the faster new product launches can happen, and the more synchronized engineering, manufacturing, marketing, and sales are. Products are delivering more intelligent data than manufacturers generate when producing them. Finding new ways to improve efficiency, quality, and yields need to start by integrating Information Technology (IT) and Operational Technology (OT)-based systems, so customer and production data yield new insights on how to improve.

originally posted on by Louis Columbus