MxD’s 22,000 square-foot manufacturing floor

ML Journal

ML Journal

AI in Manufacturing: 11 Focus Areas to Consider

Accelerating the use of AI can result in substantial benefits, but manufacturers must proceed with a practical understanding of the best applications.  

The main challenges to increased AI adoption and deployment largely revolve around data and the need to capture, organize, and analyze growing volumes in a timely manner.
Organizationally, manufacturing companies should stand up an internal team with expertise in AI, data science, and data engineering to handle all AI-related activities and investments.
AI’s single biggest impact in manufacturing could be helping companies address the future workforce gap.

Artificial Intelligence adoption and deployment seem to be less extensive and mature in industrial manufacturing than in most other industries. So far, there have been fewer big AI success stories in manufacturing and thus less competitive pressure to take immediate action. Although most manufacturing companies generally acknowledge the importance of AI—and see it is an essential and disruptive capability that could greatly affect their ability to operate and compete in the future—most efforts to date have been limited to small-scale pilots and proofs-of-concept projects focused on narrow parts of the business.

The main challenges to increased AI adoption and deployment largely revolve around data. Unlike many other industries where digital data plays a central role, manufacturing still revolves around physical work and physical assets, with many of those assets geographically scattered and disconnected from digital networks. Widespread deployment of IoT-related technologies is starting to fill this data void. However, in order to be useful, the resulting data needs to be organized, captured, and analyzed in a timely manner. Also, edge computing and edge AI technologies should be harnessed to enable timely processing and analysis of data in dispersed locations at the edge of the network.

“The main challenge to increased AI adoption revolves around data.”


For most manufacturing companies, the immediate and important next step is to establish an internal team with expertise in AI, data science, and data engineering to serve as a focal point for all AI-related activities and investments. This team would coordinate AI activities across the company’s business ecosystem, while providing a core set of internal AI resources and capabilities that can be supplemented from the outside as needed. Also, the team would provide a broad, balanced, and informed perspective on using AI across the enterprise.1

Emerging AI Use Cases

Using sensor data and AI to create and analyze digital models of real-world machines and factories can enable operations to be improved without disrupting production. Trying to optimize a manufacturing operation without disrupting production can be like trying to change the tires on a race car while it’s zooming around the track at 200 miles per hour. The solution? An AI-enabled digital twin.

A digital twin is a virtual representation of a physical device or system that mirrors its exact elements and behavior in real time. Sensor data from numerous sources—along with historical data—is combined with machine learning and advanced analytics to create digital models and spatial graphs that constantly match the status, position, and working condition of their physical counterparts. These exact digital simulations enable a company to conduct extensive analysis and optimization experiments without disrupting day-to-day operations. It’s a virtual process that can deliver real-world benefits.1 Figure 1 shows examples of digital twins in manufacturing settings.

Figure 1: Facility and Production Digital twins must be enabled to replicate real-time behavior, run analytics, and optimize manufacturing processes and throughput.

In today’s highly competitive world, companies may rush to manufacture goods on a large scale to leverage economies of scale, while still trying to offer customers the ability to customize the products to meet their individual needs and preferences. AI could help organizations minimize manufacturing costs and offer a wider variety of desirable products to customers. AI can be leveraged in real-time to help achieve the following 2:

  1. Predictive and Prescriptive Maintenance: AI predictive models may be able to predict potential equipment failure points, allowing maintenance to be performed before the actual failure occurs. This could minimize equipment downtime, improve overall equipment efficiency (OEE), and increase productivity.
  2. Safer Operating Conditions: AI-enabled robots could replace humans in hazardous working conditions, potentially reducing the risk of casualties in the workplace. AI predictive models could also alert of potential failures in advance, helping to avoid any mishaps and improve working conditions for workers.
  3. Better Quality Control: AI algorithms may allow organizations to monitor production processes, conduct continuous quality inspections, and detect deviations from desired outputs, which could allow for the identification and correction of defects in real-time. AI algorithms could increase consistency and improve quality assurance.
  4. Supply Chain Optimization: AI algorithms could be leveraged to monitor the use of raw materials from procurement to the delivery of the final product, possibly allowing organizations to optimize material flow, reduce material waste, and minimize inventory costs in real-time.
  5. Material Procurement Optimization: AI algorithms could forecast the cost of raw materials based on historical data analysis and current market trends, potentially helping organizations cope with fluctuating material prices, procure raw materials at optimal prices, and build inventory based on macroeconomic forecasts.
  6. Production Planning Optimization: AI could optimize production processes through real-time process monitoring, identifying potential bottlenecks in advance, and optimizing routings and resource allocation. This may increase productivity, reduce production costs, and improve delivery timelines.
  7. Process Optimization through Augmented (Virtual) Reality: AI algorithms could simulate different manufacturing scenarios, introducing unknown events and predicting the outcomes of production processes and equipment behavior, thereby helping to improve the production processes. Furthermore, augmented reality could help workers access real-time product data and assembly instructions, possibly reducing defects.
  8. Optimal Workforce Utilization: AI algorithms could optimize workforce/staffing, manufacturing shifts schedules, and workforce training programs, potentially improving employee satisfaction, and reducing labor costs.
  9. Improved Product Development: AI-enabled sensors could monitor production yield in real-time and provide closed-loop feedback to the R&D team to help improve product design for increased production yield and minimum production defects, likely reducing the cost of manufacturing.
  10. Enabled Proximity Search: AI could translate natural language processing (NLP) into search parameters to search for appropriate assembly parts in proximity, possibly reducing time to assemble complex products.
  11. Dynamic Production Line Testing: AI algorithms could perform unit and system integration testing of software-hardware integration as the product progresses through the assembly line. This may help organizations detect and rectify defects early in the production, rather than waiting for end-of-line (EOL) testing.
Figure 2: Best in class manufacturers are utilizing AI based analytics to drive benefits across asset efficiency, quality, cost, and safety.

Figure 3: Organizations must prioritize the AI use cases (mentioned above) to meet their business requirements and plan for implementation to maximize AI benefits in manufacturing.

Take a Practical Approach

Too many AI initiatives in manufacturing are either overly tactical and technical (too narrowly focused, and often highlighting technical capabilities that are exciting but not very useful), or overly strategic and ambitious (too difficult and expensive to implement, requiring data and advanced capabilities that don’t currently exist). To succeed with AI, manufacturing companies should have strategies and roadmaps based on a practical understanding of what parts of the business are best suited for AI.

One early and ongoing focus area for AI in manufacturing is making machine maintenance more predictive and less reactive. Another key focus area that is getting a lot of traction these days is using AI to improve interactions with customers and field workers. Also, some manufacturing companies are starting to explore the use of AI to help them handle extreme weather and other hard-to-predict events. By harnessing the power of AI vision and other advanced AI technologies, companies can monitor and analyze vast amounts of information— including data from field sensors, drone video, and weather radar—with a level of timeliness, accuracy, and thoroughness that humans alone simply cannot achieve.

A key focus area for AI is making machine maintenance more predictive.


Expanding on the idea of machines helping humans be more efficient and effective, AI’s single biggest impact in manufacturing could be helping companies address the future workforce gap.

The Biden administration’s multi trillion-dollar commitment to infrastructure is expected to dramatically increase business activity throughout manufacturing, but could also create a significant shortage of workers and expertise. AI can help address this gap by augmenting the work done by humans—doing much of the preparatory analysis and heavy lifting so human workers can focus on activities that require skills and expertise that are uniquely human.   M

1.     Source: The Energy, Resources & Industrials AI Dossier | By Deloitte AI Institute
2.     Source: Deloitte Analysis

This article contains general information only, does not constitute professional advice or services, and should not be used as a basis for any decision or action that may affect your business. The authors shall not be responsible for any loss sustained by any person who relies on this article.

About the authors:

Stavros Stefanis
is Principal, Product Engineering & Development at Deloitte. Stefanis is a leader in Deloitte’s Product Engineering & Development market offering with a focus on hardware and software development transformation using digitally integrated model-based capabilities.


Mohit Kapoor
is a Manager at Deloitte Consulting LLP. Kapoor has 18 years of experience in Product Strategy & Lifecycle Management with a focus on defining end-to-end global product development processes, system design and implementation, PLM integration with ERP, strategic planning and execution, and enterprise digital transformation.

View More