Today’s manufacturing industry is being revolutionised by Industry 4.0. Essential to its success is the use of Artificial Intelligence (AI) to boost efficiency, quality and innovation. So why are only 36% of UK manufacturers currently utilising AI in their operations? Nicholas Lea-Trengrouse, Head of Business Intelligence at Columbus UK, argues that one of the biggest barriers holding back AI adoption is the lack of data quality.
AI is not a plug-and-play add-on. It needs feeding and nurturing with good data. Here, Nicholas explains the data journey for AI readiness, from raw data cleaning and hygiene to the end goal of actionable AI-enabled business insights.
AI-ready data is crucial to successful AI implementation. Gartner predicts that by the end of 2026, organisations will abandon 60% of AI projects unsupported by AI-ready data. In manufacturing, data comes from a wide range of sources – sensors, Industrial IoT devices, machines and control systems. But raw data alone isn’t enough. To generate real-time, actionable insights with AI, that data must first be cleaned, contextualised, structured and processed.
Without context, data can mislead AI – causing missed signals or false alarms. To avoid this, manufacturers need a robust data platform with strong governance and quality control. Building this foundation is essential for feeding AI accurate, reliable data. Here’s a roadmap to help manufacturers make it happen.
Step 1: Data cleaning – making data simple for AI to interpret
The multiple sources of data in the manufacturing industry can cause headaches for AI when trying to interpret and understand the data. Ensuring data is accurate, consistent and complete is crucial.
Contextual metadata such as machine ID, timestamps, digital product passports and batch numbers can help manufacturers fix errors, handle missing values, validate sensor outputs, remove duplicates and flag anomalies before data is fed into AI platforms.
Step 2: Management & Security – ensuring your data stays your data
For three years running, manufacturing has been the most cyberattacked industry and with data coming from multiple sources, the threat of cyberattacks is vast. Manufacturers need to ensure sensitive data is securely managed by utilising role-based access control and encryption.
First, clear ownership of data needs to be established with access rights and compliance rules, then a data catalogue can be developed so stakeholders know where data is, when it’s available and how to access it.
Step 3: Break down data silos – unify workers and systems
One of the biggest barriers to effective data use in manufacturing is the siloing of Operational Technology (OT) and Information Technology (IT). To overcome this, data must be brought into a central platform. But consolidation alone isn’t enough – standardised definitions are needed to align data from the top floor to the factory floor.
Using a unified namespace or knowledge base helps connect equipment, processes and sensor streams, reducing confusion and enabling consistency across the business.
With these foundations in place, manufacturers can transform raw industrial data into a structured digital twin. The next step? Feeding that data into advanced analytics and machine learning.
Step 4: Time for a systems upgrade – turning numbers and figures into real-time actionable insights
Once raw manufacturing data has been cleaned and contextualised, the next challenge facing manufacturers is where data can be stored and processed. Many manufacturers still operate with outdated legacy databases and nightly batch processes which are unable to operate at the speed of Industry 4.0. In fact, over one-quarter of UK organisations have stated legacy technology as a key barrier to AI growth.
Enter, modern data architectures. These systems are flexible, scalable and capable of handling large datasets and enable high-stake manufacturing environments such as supply chains, production lines and maintenance schedules to operate with real-time insights.
Step 5: Don’t get caught in a data swamp
In enabling real-time data insights, data can no longer be stored in data lakes, a traditional centralised system that can store large quantities of raw data in its native format. These are great for encapsulating sensor readings or machine logs for deeper analysis but without governance or structure they can quickly turn into ‘data swamps’. To move to the next level manufacturers, need to adopt a data lakehouse.
A data architect that combines the scale and flexibility of data lakes with the governance and schema management of a data warehouse allows for all areas of the business to work from one unified platform. This means everyone from data scientists who are interested in the raw, unstructured data to business analysts who want well-structured data tables, can work and collaborate using the same system.
But that’s not all. Through enabling machine learning, business intelligence and predictive analysis, data lakehouses can store data cheaply while enforcing structure to foster collaboration and speed up analysis.
Step 6: Every second counts on the factory floor
Given the fast-paced nature of the manufacturing industry, data loses value if it arrives late. Take a factory setting for example. If a crucial machine overheats or malfunctions, it needs to be reported and flagged instantly for maintenance to be actioned.
Real-time streaming technologies process sensor data the moment it’s generated, enabling immediate action when issues arise. But the benefits go further – automated fault detection can spot anomalies in machine temperature or vibration, while live dashboards give operators instant visibility into throughput and quality.
The result? Faster response times, fewer disruptions and smarter process adjustments reduce downtime, minimise waste and boost efficiency.
Step 7: The hybrid pathway to smart factories – from edge to the cloud
In today’s modern manufacturing industry, many data architectures process all data following an edge-to-cloud model. Edge computing devices in the factory handle the here-and-now tasks such as local inference for anomaly detection or filtering sensor noise and cloud computing devices store large-scale analytics, historical data analysis and advanced AI model training. This hybrid model gives manufacturers low latency at the edge and the ability to tap into the vast quantities of data in the cloud systems.
The predictive maintenance aspect of the manufacturing digital transformation journey will benefit greatly from this approach as the edge devices do the real-time monitoring and the cloud utilises aggregated data from multiple locations to refine AI models. This is crucial for manufacturers as a recent McKinsey report stated that predictive maintenance can reduce maintenance costs by 10 to 40% and downtime by 50% and increase asset lifetime by 20 to 40%.
The data is ready and waiting, manufacturers must now look after it
There is no denying that manufacturers are data rich but those that will benefit greatly from it and the powers of Industry 4.0 and AI will be those that look after it. Turning raw unstructured industrial data into structured AI-ready data is no easy feat, but manufacturers that successfully do so will be able to ensure AI delivers accurate, real-time insights. This will help to boost efficiency, quality and profitability across all aspects of the manufacturing environment.
