Artificial Intelligence is making a strong entry into the manufacturing world. It is touted as a lever for increasing productivity, improving quality, anticipating failures, optimizing planning, reducing decision-making times, and making companies more responsive. Sensors, MES, ERP, IoT, dashboards, predictive algorithms, and automation systems promise a more connected, intelligent, and data-driven factory. But the promise of technology is realized only when the process underpinning it is stable, measurable, and free of waste.
Yet there is one point that is often underestimated: technology alone does not transform a production system. It can make it faster, more transparent, and more automated. But if processes are unstable, data is incomplete, responsibilities are unclear, and operational flows are rife with waste, digital transformation risks amplifying inefficiencies rather than eliminating them. This is where Lean once again becomes crucial: not as a theoretical framework, but as an operational method for distinguishing what creates value from what consumes resources without generating it.
The real question, therefore, is not: “What technology should we adopt?” The correct question is: "Which process do we want to improve?" Before introducing AI, Big Data, automation, or advanced dashboards, it is necessary to understand where value is generated, where efficiency is lost, and what forms of waste prevent the organization from functioning smoothly. Digitizing an inefficient process often means accelerating the problem, not solving it.
To learn more about the connection between methodology and continuous improvement, see also the content on Lean Thinking.
Every digital transformation should start with a simple principle: automation applied to an efficient process increases its efficiency; applied to an inefficient process, it increases its inefficiency.
This rule is particularly relevant in the age of AI. If a production flow is characterized by rework, delays, unnecessary movement, excess inventory, defects, overproduction, or outdated information, simply introducing technology isn’t enough. The risk is building a system that’s more modern in form but no more effective in substance.
Digitizing chaos means accelerating a loss. Automating a non-value-added activity simply makes it faster. Connecting systems that manage inconsistent data means increasing the speed at which incorrect information circulates throughout the organization. That is why the starting point is not the tool, but the process: digital technology must be a lever for simplification, not a multiplier of complexity.
For this reason, the correct path is "Lean First, Digital Second, Value Always": first, simplify, stabilize, and standardize the process; then digitize what truly makes sense to enhance. This approach is consistent with the principles of value-driven digital transformation, as also explored in depth by Bonfiglioli Consulting.
Based on our experience, transformation must be Lean & Digital and therefore must not start with technology, but with value: with the ability to analyze the process, distinguish what creates value from what generates waste, and build a system in which people, data, processes, and digital tools all work toward the same goal.
Lean teaches us to observe processes with a very practical eye: every activity must be evaluated in terms of the value it generates for the customer and for the organization. There are value-adding activities, non-value-adding but necessary activities, and non-value-adding activities that can be eliminated.
Traditional wastes are well known in the manufacturing world and are defined using Japanese terms. From an operational perspective, their meaning is not theoretical but practical: they indicate where the system wastes time, space, energy, and decision-making capacity.
For a systematic exploration of this topic, see also the Lean Thinking glossary.
Muda refers to the most obvious forms of waste: defects, rework, scrap, additional inspections, delays, and loss of reliability. Every defect is a visible cost, but also a sign of process instability. Overproduction is one of the most dangerous forms of waste, because it involves producing more than is needed, before it is needed, or in quantities that do not match actual demand. Overproduction generates inventory, takes up space, hides problems, and reduces flexibility. Waiting represents another significant form of waste: workers waiting for materials, machines idle due to a lack of components, production lines halted due to a lack of information, and decisions postponed because the correct data is not available. Every instance of waiting hides a loss of time, but also a loss of decision-making capacity.
Added to these are unnecessary transportation, excessive handling, oversized inventories, non-optimized processes, and skills that are not fully utilized. Every unnecessary movement, every redundant step, and every stock created to compensate for upstream instability is a symptom of a process that isn’t flowing smoothly. Lean analysis isn’t just about "seeing" waste, but about understanding where the dysfunction that causes it originates.
In Lean terminology, in addition to the classic forms of waste, there are also:
These forms of waste do not disappear with digital transformation. On the contrary, if they are not addressed first, they risk being replicated in information systems, dashboards, and algorithms. This is why standardization is not merely an organizational detail, but a prerequisite for any credible AI project.
With digitization, new forms of waste emerge—less visible than excess inventory or machine downtime, but just as impactful: digital waste.
Digital waste occurs whenever an organization expends time, energy, and resources to manage information that is unreliable, redundant, incomplete, or difficult to find. It may seem like a minor problem compared to a machine shutdown, but in reality, it profoundly affects the quality of decisions.
There are numerous examples: duplicate data across multiple systems, parallel Excel files, information still collected on paper, repeated manual data entry, reports produced with great effort but rarely used, misaligned dashboards, metrics calculated differently from one department to another, and unintegrated IT and OT systems. These are not mere information flaws: they are organizational costs that slow down the decision-making process and make subsequent actions less reliable.
Every time a person spends time searching for data, verifying its reliability, comparing different versions of the same file, or correcting incorrectly entered information, the organization is expending energy on activities that do not generate value. Data redundancy, low-quality information, parallel Excel spreadsheets, paper-based data, and duplicate data entry are concrete examples of digital waste. Effective digitization reduces the number of steps—it does not multiply them.
Digital waste slows down processes, but above all, it slows down decision-making. On the factory floor, delayed decision-making often means taking action only after the problem has already had an impact. An anomaly not detected in time can lead to machine downtime. Quality data not interpreted correctly can result in scrap. An unreliable forecast can lead to excess inventory or service failures. An unaddressed alert can turn into an emergency. This is why data must be managed throughout the entire flow: collection, validation, integration, interpretation, and action.
Resources from NIST and the World Economic Forum are also useful on the topics of data quality and data governance.
Digitalization, therefore, must not be limited to "producing more data." It must help the organization produce better, more useful, more timely data that is more closely linked to action.
One of the most important principles of Lean & Digital Transformation is clear: you improve what you measure. But this statement is true only under one condition: what you measure must be accurate, meaningful, and linked to real processes.
Measuring poorly means improving in the wrong direction. Measuring too much means spreading your attention too thin. Measuring without taking action means creating bureaucracy. Measuring without quality data means generating misinformation.
For this reason, every AI and digital transformation journey must start with a fundamental question: Is our data truly usable for making better decisions?
Data must be native, complete, accurate, certified, and available when needed. It must be collected at the point where the event occurs, avoiding manual steps, transcriptions, and duplications. It must be accessible in departments, control rooms, operational meetings, escalation processes, and decision-making moments. This is not just a technical issue, but also an organizational one: without clear accountability for the data, no dashboard can drive real improvement.
Only then does data become information. And only when information is interpreted within a clear process can it become knowledge useful for improvement.
An indicator should not be a number to be passively observed. It should generate a question, a decision, or an action. If a KPI does not trigger a behavior, a sense of accountability, or an improvement, it risks becoming nothing more than information noise. A good indicator, therefore, does not merely measure the past: it guides future behavior.
In the industrial world, the shift from traditional data to Big Data is not just about quantity. It’s about how data is generated, collected, integrated, interpreted, and transformed into value.
Traditional data is often structured, collected at defined intervals, managed in separate systems, and used primarily for retrospective analysis. It is important data, but it tends to describe what has already happened.
Big Data, on the other hand, comes from a multitude of sources: machines, sensors, MES and ERP systems, quality control, maintenance, logistics, the supply chain, customers, suppliers, and digital platforms. It is more voluminous, faster-moving, more variable, and often more complex to interpret. The difference is not merely quantitative: it changes the organization’s ability to detect weak signals, correlations, and anomalies before they become problems.
The real difference, however, isn’t simply having more data. It’s being able to transform this mass of information into actionable knowledge. A factory can generate millions of data points a day and still be largely non-data-driven if that data isn’t governed, integrated, and linked to decision-making processes.
Being data-driven doesn’t mean having lots of dashboards. It means making better decisions because the information is reliable, available, understandable, and action-oriented. Data isn’t valuable because of its abundance, but because of its ability to improve a concrete decision.
The 8 V’s of Big Data help us understand what characteristics a truly useful information asset must have.
Volume is the amount of available data. Sensors, machines, ERP systems, MES systems, quality systems, and maintenance systems generate an ever-growing mass of information. But volume alone is not enough. The challenge is to find the relevant information within an ever-expanding data stream.
Value indicates how much a piece of data helps in making a decision, preventing an anomaly, reducing waste, improving a KPI, or anticipating a problem. If it does not lead to action, the data remains useless information.
Veracity refers to the reliability of the data. Inaccurate data can create misinformation and lead to wrong decisions. This is why we need data that is certified, verified, and consistent across different sources.
Visualization should not be limited to simply displaying numbers. A good dashboard must help people understand what is happening, where to take action, and what the priorities are.
Variety—the diversity of sources and formats—pertains to the origin of the data: machines, operators, management systems, sensors, customers, suppliers, and digital platforms. Integrating diverse sources is essential for building a comprehensive understanding of the process.
Velocity refers to how quickly data changes. Some data requires real-time decisions, while other data calls for deeper analysis. The speed of the data must align with the decision-making process it is intended to support.
Viscosity describes how difficult it is to extract and process data. If data is slow to process or locked in separate systems, the organization loses responsiveness and improvement slows down.
Virality refers to the speed with which an alert, anomaly, or deviation reaches the right people at the right time through defined routines and channels. This dimension is particularly important in manufacturing contexts, because the speed at which information spreads directly affects the timeliness of intervention.
These 8 V’s show that the issue is not simply "having data," but building an information system capable of generating better, faster, and more reliable decisions. The real challenge is not possessing large amounts of data, but transforming it into value.
To generate value, data must be managed methodically. It’s not enough to simply collect it, store it, or display it on a dashboard. A structured process is needed to transform raw data into reliable information, information into knowledge, and knowledge into operational decisions.
What problem do we want to solve? What performance metric do we want to improve? What decision do we want to support? Do we want to reduce machine downtime, improve OTD, decrease scrap, optimize inventory, increase line efficiency, or make forecasts more reliable? Without a clear objective, there’s a risk of collecting a lot of data but not knowing how to use it. Data must always stem from a business question or a process need.
Once the objective has been defined, you need to identify what data is needed, where it comes from, who generates it, how often it needs to be updated, and what level of reliability is required. Data collection should take place as close as possible to where the event occurs. The more manual steps involved, the greater the risk of errors, delays, and duplication.
Data quality is not a given. Before it can be used, the information must be cleaned, normalized, validated, and made consistent. This phase allows you to eliminate duplicates, missing values, inconsistent formats, entry errors, and discrepancies between different systems. It is one of the most important steps, because dirty data leads to weak analysis and poor decisions.
Exploration allows you to identify correlations, anomalies, deviations, trends, and recurring patterns. This is where the organization begins to transform numbers into insights. For example, you can observe relationships between micro-stops and production shifts, between defects and material batches, between delivery delays and product families, and between energy consumption and operating conditions. This analysis allows you to move from perception to objective knowledge of the process.
Once the process behavior is understood, it is possible to build analytical, predictive, or prescriptive models. This is where advanced tools such as AI, machine learning, simulations, and optimization algorithms come into play. Modeling allows us to anticipate scenarios, assess risks, predict failures, support scheduling decisions, identify root causes, and suggest corrective actions. But the model is only valuable if it remains connected to the actual process and if people know how to interpret its outputs.
The final step is to make the data usable during decision-making. Dashboards, control rooms, alerts, automated reports, digital workflows, and escalation systems must help people quickly understand what is happening, where to intervene, and with what priority. Visualization should not be decorative but operational. A good information system doesn’t simply display data—it guides action.
This 6-step process allows you to move beyond the mere availability of information and build a truly data-driven culture. This is where AI and Big Data can generate value: not as a technological exercise, but as part of an integrated decision-making process.
AI reaches its full potential when integrated into a mature Lean system. Lean reduces waste, variability, and complexity. Digital technology makes data visible in real time. AI helps identify patterns, anticipate scenarios, and support predictive decisions.
Together, these elements enable a shift from reactive management to preventive, proactive, and predictive management. It is at this intersection that digital technology ceases to be a mere technological investment and becomes a measurable competitive advantage.
In maintenance, this means anticipating failures and reducing unplanned downtime. In quality, it means identifying deviations before they lead to defects. In planning, it means simulating scenarios and improving scheduling and forecasting. In internal logistics, it means optimizing flows, routes, and inventory levels. In performance monitoring, it means speeding up root cause analyses.
But it all starts with one foundation: clear processes, reliable data, and the elimination of waste.
AI does not replace the method. It enhances it. It does not eliminate the need for standards, roles, responsibilities, and routines. On the contrary, it requires an organization that is even more disciplined in managing processes and information.
An algorithm can suggest a forecast, but the organization must know how to interpret it, validate it, and turn it into a decision. A system can generate an alert, but there must be a routine in place to handle it. A dashboard can highlight a deviation, but trained personnel are needed to interpret it and take action.
Without this connection between data and action, digital transformation remains incomplete.
One of the most common mistakes in digital transformation is confusing visualization with improvement. A dashboard can make data more accessible, but it does not automatically guarantee better decisions.
For a dashboard to generate value, it must be integrated into a management routine. It must be clear who uses it, when it is consulted, which metrics are prioritized, which thresholds trigger an escalation, what actions must be taken, and how the effectiveness of decisions is verified.
The control room, daily meetings, problem-solving sessions, escalation processes, and periodic performance reviews are where data becomes action. That is where the organization interprets deviations, identifies causes, assigns responsibility, and initiates improvement. A dashboard, on its own, does not improve anything: it only improves performance when integrated into a rigorous and consistent management practice.
Unused data is a waste. An unaddressed alert is a waste. A report that is produced but not discussed is a waste. A beautiful dashboard that isn’t linked to decisions is digital waste.
True transformation doesn’t happen when a company "sees" more data, but when it makes better decisions based on that data.
Every AI project should start with value. What problem do we want to solve? What waste do we want to eliminate? What process do we want to make more stable? What decision do we want to improve? What data do we really need?
Only after answering these questions does it make sense to choose the technology. Otherwise, the risk is starting with the tool and then looking for a problem to apply it to. A well-designed project doesn’t start with the software, but with the need for improvement.
Lean & Digital Transformation is not about adopting software or a platform. It is a journey of redesigning processes, the organization, skills, and governance. Technology is a lever, not the starting point.
This approach helps avoid three very common mistakes. The first is digitizing unstable processes. The second is building information systems full of data but lacking in meaning. The third is adopting advanced technologies without creating the organizational conditions to actually use them.
Digital transformation works when it enhances a process that is already understood, simplified, and governed. AI generates value when it works with reliable data. Big Data becomes a competitive advantage when it supports better decisions. And Lean remains the method that allows us to distinguish what creates value from what produces noise.
Because applying technology to an inefficient process does not eliminate waste—it merely speeds it up. Before digitizing, the process must be simplified, stabilized, and standardized according to the "Lean First, Digital Second, Value Always" principle.
These are inefficiencies related to data management: duplicate information, parallel Excel spreadsheets, data entered manually multiple times, rarely used reports, or misaligned dashboards. They consume time and energy without generating value, just like traditional Lean wastes.
They are the eight characteristics that a data asset must possess to generate value: Volume, Value, Veracity, Visualization, Variety, Velocity, Viscosity, and Virality. Together, they help determine whether the data is truly useful for supporting better decisions.
Through a six-step process: define the objective, collect the data, prepare it, explore it, model it, and finally present it in a way that concretely guides action—for example, through dashboards, alerts, or management routines.
AI enhances a mature Lean system: it helps identify patterns, anticipate failures, optimize planning, and speed up root cause analyses. But it only works when applied to stable processes and reliable data—it does not replace them.
In an increasingly complex industrial world, the real difference will not be made by companies that collect the most data, but by those that know how to transform it into knowledge, decisions, and continuous improvement.
Lean waste and digital waste are two sides of the same problem: organizational energy wasted on activities that do not generate value. The former manifests itself in physical, production, and operational flows. The latter manifests itself in information flows, systems, data, and decision-making processes.
For this reason, digital transformation must begin with a holistic approach: lean processes, high-quality data, clear governance, widespread expertise, and technologies aligned with business objectives. It is a matter of method even before it is a matter of tools.
AI can become an extraordinary accelerator, but only if it finds fertile ground: stable processes, transparent workflows, reliable data, clearly defined responsibilities, and effective decision-making routines. From this perspective, value arises from the alignment of people, processes, and data—not merely from the availability of technology.
Because you improve what you measure. But you truly improve only when you measure well—using reliable, meaningful data that’s linked to value.
ai translated