Centric Consulting AI in manufacturing for Data-driven manufacturing

Insights

Today’s low-code and pro-code AI tools, integrations, and cloud-based platforms make it easier for manufacturers of all sizes to make practical, data-driven decisions and automate processes that solve actual business problems—but it all starts with your data and data infrastructure.

Who This Is For

CEOs, CIOs, and operations leaders at midsized manufacturing companies who know AI is the next step but want to know how to prepare their data infrastructure to make it work.

In Brief:

Generative AI is making data-driven manufacturing increasingly more accessible for midsized manufacturers, not only large companies.

Cloud-based platforms and user-friendly tools make AI in manufacturing more scalable and cost-effective.

Integrating real-time shop-floor data with business data helps manufacturers make efficient, data-driven decisions.

Building a data-driven culture is essential for harnessing AI’s benefits and driving growth in manufacturing.

AI in Manufacturing Is Not Only for Big Companies

Generative artificial intelligence (AI) is already transforming large manufacturing companies, but with the right data collection and management strategies, many midsized manufacturers are reaping AI’s benefits, too. In fact, according to research firm RMS’s Middle Marketing AI Survey, 91 percent of middle market companies now use generative AI in some capacity, compared to 77 percent in 2024. RMS calls these statistics “a clear sign that AI is becoming standard in business operations.” The reason is simple. Cloud-based platforms, low-code tools, and user-friendly interfaces make today’s AI use cases for manufacturers more accessible, scalable and cost-effective than ever. The challenge? Gathering real-time shop-floor data and integrating it with business data to generate the greatest efficiencies, most valuable insights, and smartest decisions. If your midsized manufacturing business can cross those hurdles, it can join the 82 percent of all organizations that want to deploy AI or are already using it to transform their businesses. In this blog post, we’ll use a hypothetical client, Door2Door Manufacturing, to illustrate the journey your data-driven strategy must take to begin successfully using AI in manufacturing. As we’ll see, the process of collecting, organizing, refining, and integrating data is challenging but achievable. At the end of our journey, you’ll have a framework for harnessing your data, building smarter systems, and creating a culture of data-driven manufacturing that’s primed for AI.

Meet Our Hypothetical Client

Door2Door Manufacturing serves middle-income homeowners who want to upgrade their home’s curb appeal with a semicustomized front door. Door2Door manufactures a range of base models, and customers can enhance them from a catalog of customization options at a significantly lower cost than purchasing high-end doors. When completed, the company delivers and installs each door at the customer’s home. The business began as a startup during the early dot-com boom, building on the home delivery craze. By 2019, Door2Door had grown into a $500 million company. They developed an information technology (IT) roadmap and moved their systems to the cloud. However, the 2020 pandemic prevented them from digitizing their operational efficiency and many of processes on the plant floor. As business improved following the pandemic, Door2Door returned to its IT roadmap, just in time for the rise of generative AI. Now experiencing the FOMO — fear of missing out — the company wanted to implement generative AI into its operations, but the CEO and CIO didn’t know where to start.

Data: The Raw Material for AI in Manufacturing

Though Door2Door is a forward-thinking company, generative AI is overwhelming. AI technology changes so quickly, and the company struggles to keep pace with demand. Fortunately, the path to AI integration doesn’t begin with complex algorithms or expensive platforms any more than an automobile assembly line starts with backup cameras and a Bluetooth audio system. No, both processes start with raw materials. For AI, the raw material is data. As Door2Door will learn, whether you’re currently using a basic spreadsheet or an advanced enterprise system to collect raw data, what truly matters is your data’s quality, consistency and relevance. That’s because accurate, well-structured data is the foundation for meaningful insights, operational improvements, and, ultimately, AI-driven transformation. By focusing on capturing the “right” data — that which reflects Door2Door’s manufacturing processes, performance, and pain points — the company can set the stage for smarter decisions and scalable innovation. So, where should Door2Door begin? Through our decades of working with clients, we have identified a few simple steps to start moving down the right path.

3 Steps to Prepare Your Data for AI in Manufacturing

Graphic image: 3 steps to prepare your data for AI in Manufacturing. 1. Define your improvement goals, 2. Identify the data you need, 3. Focus on data consistency

Step 1: Define Your Improvement Goals

Whether it’s reducing scrap, increasing throughput, or boosting machine uptime, you need to define your goals as a prerequisite before you create a data-driven strategy. With each goal in mind, you can select performance metrics that reflect current conditions. For example, Door2Door identifies “scrap reduction” as a goal. To mark progress toward their goal, the company adopts quality yield as one metric. They calculate it as: Number of Salable Doors Made ÷ Total Number of Doors Made A lower yield percentage indicates higher scrap, which gives the company a baseline to set a yield target and measure improvement toward it.

Step 2: Identify the Data You Need

With your metric defined, you must determine the specific data points needed to calculate and contextualize it. For Door2Door’s quality yield, leaders identify:

Total doors produced

Total salable doors produced

Reason for defect

Machine ID

Product type (standard, customized, deluxe customized)

Operator’s name

Shift or time period

These details provide valuable context, helping Door2Door uncover patterns and root causes over time.

Step 3: Focus on Data Consistency

Now that you know what data to collect to reach your goals, focus shifts to data consistency. Data consistency includes making sure you format data types correctly, you record data elements in the proper fields and in the correct formats, and that you handle key data elements such as dates and addresses consistently. Remember: Inconsistent data is AI’s enemy. Door2Door’s first concern is its manual processes, which don’t yield electronic data. Instead, they collect data using:

Simple forms at each machine that log the number of defective products

Manual records of total machine output, captured in writing at the end of each shift

Numbers hand-keyed from forms into centralized spreadsheets and databases

The good news: Door2Door does not have to begin collecting digital data immediately. Merely collecting data gives them a place to start. For now, they only need to focus on data consistency. They start by training employees to capture the necessary details using the proper forms and placing data in the correct fields. They document these practices for employees and make floor managers responsible for ensuring employees collect data properly. The company also trains its data analytics team on good data hygiene. Data analysts learn the principles of rooting out data ROT — Redundant, Obsolete, and Trivial data — while the CIO implements a data governance framework and regular data quality reviews. This work prepares Door2Door for the next step: replacing manual collection methods with inexpensive data collection tools, such as cameras and sensors. As more data becomes digital, the company begins taking small steps toward automated data collection.

Automated Data Collection: The Modern Way to Capture Your Data

Now that you are collecting operational data, assigning metrics to it, and ensuring consistency, you have established the groundwork for automating your data collection. Automated data collection can increase speed by feeding operational and production data directly into Door2Door’s IT system, but without the right equipment, automated collection introduces other challenges. For example, the sheer quantity of information gathered can overwhelm Door2Door’s IT infrastructure. Automated collection can also pose security risks, whether from within or outside the company. An operational technology (OT) network is part of the solution. An OT network connects to a communication platform, or edge device, that controls the flow of data between the OT and IT networks. The edge device also interfaces with all the company’s operational devices, securely passing data between the OT and IT networks, whether on-premises or in the cloud. Graphic Image: Integrating Operational and Business data through OT and IT networks to create actionable information. Door2Door evaluates Litmus Edge and finds it to be a good option. It is a single-system solution that interprets the real-time data coming from various sources and assimilates it. Litmus also features built-in connectors for more than 280 devices, and it can scale from one to hundreds of machines at multiple sites. Litmus is capable of being up and running in as little as three weeks, and you can also configure Litmus to send alerts for certain events. For example, it can alert managers if a machine goes down for a designated amount of time, or it can notify them if an employee reaches a certain production target. As Door2Door collects more data automatically, the company can begin harnessing it for real-time, AI-driven operational insights, especially as it moves toward the next goal on its IT road map: integrating operational data and business data analytics.

Manufacturing Data Integration With Enhanced, AI-Powered Decision-Making

To this point, Door2Door has primarily focused on preparing operational data for automated data collection and moving that data into the IT system for access. But to add greater context for data that will unlock deeper insights into operations, the company needs to integrate its operational and business data. For example, if floor managers receive an alert from Litmus Edge that a machine has been offline for 10 minutes, they can now access operational data that tells them exactly when the outage started, who was working at the time, and how it affected production. However, other manufacturing data, such as material suppliers and equipment maintenance records, may also be relevant, but the business IT system holds it. By integrating the operational and business data streams, Door2Door can quickly see that the last round of scheduled maintenance was postponed for production scheduling reasons. They now have a clue about where to look for problems. Data integration can also help with external factors that affect production. At Door2Door, raw materials swell slightly when the outdoor humidity is high. When the company’s operational data shows a dip in productivity, it can check the weather information that the business maintains. Mapping meteorological data to electronic device history records allows operators to adjust how the machines handle the raw materials when outside humidity spikes. To gain these insights, input the relevant data into a model that identifies their relationships. In Door2Door’s case, the elements would be downtime data (operational) and maintenance data (business), or production output data and meteorological data. The data model, residing on the IT network, then provides the necessary information for teams to perform data analytics and generate reports. The challenge is to not duplicate data in the company’s OT and IT systems, but to pull it together in a data lake instead. Once in the data lake, you can update data regularly and access it when needed. Not having the right data model would be like going back to those multiple spreadsheets and refreshing, reformatting, and cutting and pasting from them every time someone wanted to review information or create a report. So, the question becomes: How should Door2Door implement a data model?

Discover Data Relationships to Improve AI in Manufacturing With a Data Model

As a cloud-based company, Door2Door has several data model options. One option the company reviews is Microsoft Fabric. This data platform facilitates cloud storage, Real-Time Intelligence, access management, and other features. By properly implementing a data model like Fabric on its IT network, Door2Door gains insights into not only the company’s operational and business data, but also relationships between various systems on the IT network, such as customer service and accounting. Because Fabric is also cloud-based, it doesn’t require any hardware, and Door2Door doesn’t need to load it onto many local systems. Like other data models, Fabric ties multiple systems together and comes with connectors for many of Door2Door’s systems. The company can even integrate new or custom IT systems into Fabric by mapping the desired manufacturing data elements into the data model. The rest of the company’s IT infrastructure will continue to function normally during integration. By establishing this data model, Door2Door has not only connected the information on all the company’s systems, and it can also better understand the relationships between different kinds of data. With the help of AI, operational and business employees alike may see meaningful correlations they might have never seen without these systems. The result is the power to more easily paint a complete picture of the business in real time, and to perform historical analysis as a foundation for improvement initiatives. Even more importantly, Door2Door is closer to applying AI in all aspects of its operations and scaling those efforts in the future.

Visualization Helps You See Data Analysis in a Whole New Way

Door2Door now looks back on its spreadsheet days and considers how far the company has come. Although some employees knew how to use conditional formatting in Excel to highlight specific data points or create graphs to illustrate data patterns, these employees were few and far between. Most still relied on columns and rows that concealed relationships, patterns and insights. But what’s the alternative? Thanks to the work they’ve been doing, they are now ready for advanced — yet easier-to-build — data visualizations. One solution is Microsoft’s Power BI. As Centric Consulting Senior Architect Jo Karnes says, “Power BI allows you to create dashboards and applications to get a view into the areas of your business that are pertinent for maintaining an understanding of the metrics most valuable to you and your executives.” Power BI presents data in relation to each other, which is similar to how AI analyzes data. When Door2Door employees see how their data fits with other data, they’ll be more prepared to embrace the advantages of AI in manufacturing. These graphical views are also more intuitive than spreadsheets. For example, Door2Door created a line chart showing productivity for each of the last 30 days, with different colored lines representing the first, second and third shifts. Unlike a traditional spreadsheet, this visualization showed — at a glance and in real time — how much one shift produced in relation to the other two. It also allowed employees to filter by data like product code or machine number, and then share views with the management team, even on their mobile devices. Another use case might be to categorize downtime to see how it varies by product, shift or external conditions, such as temperature or humidity. You can also use visualizations to drill into data to evaluate overall equipment effectiveness, such as scrap rates. For example, data may reveal that while one machine creates more scrap, another produces less — but more costly — scrap. For Door2Door, the company can now make a better decision. One of Power BI’s most powerful aspects is its ability to create a series of dashboards using data from Door2Door’s comprehensive Fabric model. These dashboards allow the company to display real-time views of various aspects of the business. In one place, leaders can review key performance indicators (KPIs) for multiple functions, from sales and procurement to production, safety, human resources, finance, and more. What once took several hours to compile now appears in real time and on any device, but only for authorized employees who are permitted to view it. Once Door2Door starts using Power BI to visualize business data, the company will find even more ways to better manage operations while becoming more efficient, effective and resilient. Door2Door can use data-driven decision-making to identify places to implement AI for the greatest results. The company’s next challenge is its people.

The Data-Driven Mindset: Solve for the Brain Drain

Like many organizations, Door2Door has people in critical roles with institutional knowledge. The company prides itself on hiring experienced carpenters to work in its shops, and many seem to have developed a sixth sense for the business. They anticipate problems before they occur, or they make decisions based on their gut. Some even claim to know when humidity will rise, for example, because that’s when an old football injury acts up. But what happens when the company loses those people? Or when Door2Door teams them up with younger employees, only to find the veterans aren’t good at sharing their insights? You need to adopt a data-driven manufacturing mindset across your business. Door2Door hit the ground running by collecting accurate and timely data, integrating it with business data, and presenting it in a way that provides clear direction for action. The next step? Encouraging teams to lean into data-driven manufacturing. Teams should ask:

“Does the data support choice?” The team must share the data in a way that convinces others.

“Does the data refute my choice?” The team must confirm they’re using the right data, and that it’s accurate.

“Is my decision the right choice?” The team can’t assume that an approach that’s worked well in the past is the best of all possible options. Data-model-generated insights and AI can help evaluate alternatives.

These questions will help build more data confidence, but like any large change initiative — especially one that can disrupt the way many employees do their work — fully incorporating a data-driven mindset is best approached through a change management philosophy. While change management is its own discipline, at Centric, we have seen the following approach work well in the data space:

Start by letting your teams know that the shift to a data-driven mindset will ensure more consistency, reliability and resilience.
Help everyone understand WIIFM (What’s In It For Me). Point out the ways that using data can make each role easier.
Ensure that everyone has access to the data and systems they need — but not to the data they don’t need.
Train employees to access and use the data.
Celebrate examples of successfully applying data and the data-driven mindset and share them with your team.

The most resistance will likely come from people like those long-time employees, who have successfully relied on their intuition for years. To reach these employees, let them know that the data will, more than likely, validate their intuition and prove that they have been right all along. Or you might show them how becoming more data driven will give them time back: No more checking emails on vacation or spending parts of every day checking behind employees. Data helps other people fill in and maintain operations so seasoned workers can focus on more strategic initiatives. The bottom line: Data does not replace wisdom and intuition — it supports them.

Data-Driven Decision-Making for Leaders

Employees on the shop floor are not the only ones who may reject the data-driven mindset. The C-suite is susceptible, too. However, they play a critical role in building the mindset and setting the stage for AI in manufacturing. Centric CEO Larry English says, “Leaders need to show enthusiasm about equipping and educating teams on the potential uses of AI, helping their employees learn why their roles are important and how they can deliver more value with AI as a tool.” Below is our step-by-step process for helping leaders recognize the value of data-first thinking:

Start by asking them to consider the most important aspects of managing the business that cross their minds every day.
Next, ask them what indicators they use to gauge performance in those areas.
Encourage them to tie those indicators to data.
Invite them to rephrase their daily concerns.

For example, a general statement about “I worry about raw materials” might become, “Our overall equipment effectiveness (OEE) was only 60 percent today because our wood supply swelled in this high humidity. That slowed our machines and drove productivity down to 80 percent.” Now, the executive knows what happened, its impact, and how to improve the next day (adjust the machines). Once you address the immediate challenge, your leaders can begin thinking ahead to avoid similar problems in the future, such as sourcing less sensitive materials or — better yet — implementing AI for manufacturing to automatically adjust machine settings. This level of analysis is only possible because Door2Door implemented technologies to integrate its data. Operations will continue to improve because of this data-based decision-making, and the business is more resilient because it no longer depends on a handful of experienced employees or guesswork. Business research company AlphaSense sums up the power of generative AI in manufacturing by saying generative AI offers significant capabilities for manufacturing by processing large volumes of historical information from sources like inventory systems, market data, and customer input to generate useful insights and guidance. The company also notes that these capabilities enable manufacturers to enhance their production planning processes, discover ways to reduce costs, and make faster decisions that boost overall operational effectiveness. As your employees and leaders see these results, they will begin to trust their data and data-driven mindset more, which will open them up to accept the analysis, results and suggestions that AI can provide.

AI in Manufacturing: The Value Proposition

As humans, we have an insatiable curiosity to know “Why?” — and we usually look for answers using methods and sources that seem logical. We discount data analysis and sources that don’t seem relevant because including them wouldn’t “make sense.” One of AI’s greatest advantages is that it doesn’t need to rationalize what factors make sense. AI can sift through all the data and create a model that highlights correlations and potential causations that we don’t even consider. At a recent quarterly review, Door2Door noticed a productivity drop every Friday. Many leaders dismiss the drop by thinking that more people might be off to have a long weekend. Or maybe people are simply distracted by weekend plans. They decide there’s not much they can do about human nature, and the company puts a premium on work-life balance. Still, they are striving to be more data-driven, so they pull up the production data to support their theories: the number of scheduled versus unscheduled absences, the number of work hours in the day, even the type of products that employees manufacture on Fridays. Previously, after reading a survey from RMS that called Microsoft Copilot “the second most-popular generative AI platform” with a 43 percent adoption rate among participants, Door2Door leaders selected Copilot as its sanctioned AI tool. Eager to test it, they asked their new tool to review the company’s operational and business data to explain the drop. Copilot delivered a surprising result. It discovered that every Friday, maintenance cleans the primary air compressor filters. On those days, the back-up compressor runs at a lower capacity. Fridays are also the days the operators use compressed air for routine quality assurance (QA) testing, further taxing the backup compressor’s capacity. After the executives slapped their foreheads and said, “Why didn’t we think of that?” they realized the value AI delivered. Air filter maintenance and QA testing are only two among hundreds of tasks employees perform every day at the plant. Each task generates thousands of data points. Executives and employees would have needed weeks of observation to pinpoint the air compressors. Now, they have a solution: Stagger maintenance and QA testing to decrease the impact on the compressor. The through-line for these examples is AI’s ability to see things that humans cannot. AI doesn’t replace people on the shop floor, but it helps Door2Door make better decisions, faster. To continue the company’s mission of building a data-driven mindset, the Door2Door executive drafts a value proposition for using generative AI in manufacturing: “Generative AI helps us use data to make smarter decisions and automate certain processes to improve efficiency, reduce downtime, and make everyone’s work life better. It is our engine for building resilience and future-proofing our company.”

Data Paves the Way for Successfully Using AI in Manufacturing

We’ve demonstrated that the benefits of generative AI are within the reach of all manufacturing companies, large and small. While the journey may seem overwhelming at first glance, the technology is surprisingly easy to implement at a lower cost than you might think. Graphic Image: the data-to-decisions journey in manufacturing. Data collection, data integration, data model, data insights and visualization, data-driven mindset. ai readiness and resilience To recap, below is Centric’s process that will pave the way to using AI in your manufacturing facility:

Collect data manually to help you understand the operation better and begin focusing on data consistency.
Transition to an automated data collection system to improve speed and accuracy.
Tie the data into a model that includes other relevant business information, adding context.
Visualize the data in a way that clearly illustrates results and when action is needed.
Work with your team to adopt a data-driven mindset that emphasizes the importance of incorporating data into all decisions.
Use AI to provide data-driven insights that guide actions.

By starting your journey to using AI for smart manufacturing now, you’ll set your company up for sustainable and scalable performance enhancements that improve customer and employee satisfaction while driving higher revenue and profits.

Frequently Asked Questions

How much data do I need to start using AI? Consistency and structure is more important than volume. The first priority is making sure data is collected consistently and stored in the right fields—even if you’re collecting the data manually. That solid foundation will set the stage for automated collection and AI integration. What's the difference between operational data and business data in manufacturing, and why does it matter? Operational data comes from the shop floor — sensors, machines, production logs. Business data lives in IT systems — maintenance records, supplier information, and financials. Separately, each tells a partial story. Together, they reveal relationships you'd never spot otherwise, such as a machine outage that traces back to postponed maintenance. What AI tools are most practical for midsized manufacturers just getting started? Practical entry points such as Microsoft Fabric for data modeling, Power BI for visualization, and Microsoft Copilot for AI-assisted analysis are good starting points, because they don't require custom infrastructure or large upfront investment. However, a well-structured data model will outperform any AI tool that can only access inconsistent or incomplete data. How do you get employees to trust and use data-driven decisions? Start by showing them that AI validates what experienced workers already know, it doesn’t replace it. Also, help them understand what's in it for them: fewer fires to chase, less time spent checking behind others, and more focus on higher-value work. Finally, share early wins visibly, and make sure the right people have access to the right data from day one. A well-crafted vision and strategy serve as your safeguard against failure and unlock AI’s full promise. But first, you need to know if your company is ready for AI. Are you ready to move forward with a manufacturing data strategy for your organization but aren’t sure where to start? Our Data and Analytics and Manufacturing experts bring a tried-and-true approach for executing strategies into practical, pragmatic and actionable plans. Talk to an expert.