Welcome back to 3 Minutes Inside Operations. In this episode, Richard Holland looks at AI in manufacturing and at what it changes about the speed at which an operation can improve.
AI in manufacturing is the use of machine learning and analytics on production, machine and quality data to spot problems earlier, find their causes faster and predict failures before they reach the shop floor. The priorities it serves are the ones manufacturers have always had: productivity, quality and cost. What AI changes is how quickly a business can improve them, and Richard’s argument is that the more useful question for leadership teams is whether their operation is improving fast enough to stay ahead of its competitors.
Where AI in Manufacturing Is Being Used Today
Most manufacturers are still applying AI a long way from the production line. Make UK’s 2026 report, AI, skills and the future of the UK manufacturing sector, found that 83% of companies use AI in business support functions such as HR, marketing, finance and administration, while only 11% use it in production and 6% in quality control. Just 2% of the manufacturers in the survey have AI widely embedded in their operations.
That gap matters because production and quality are where problems cost the most, and where the speed of the response decides how much they cost. The AI use cases in manufacturing that change performance are the ones that spot a drifting process before it produces another hundred defective parts, because that protects margin, capacity and the customer at the same time.
How Problems Hide on the Production Line
Think about what happens when a problem starts to develop on a production line. A machine begins to deteriorate, and although it has not produced a defect yet, variation is creeping up. Eventually parts drift out of specification, and the issue may not be picked up until a later process. Because the defect rate is low at first, it never reaches the daily top five issues raised through the hourly and daily management system, so production carries on, operators make adjustments, and it can be weeks before anyone recognises the pattern and starts to look for the cause.
When the investigation does begin, the team has to collect the data, find the process where the defect started, look for correlations, list the potential causes, establish the root cause and then implement a solution. In some businesses that cycle takes weeks or months, and every week of it adds scrap, rework and lost capacity. For Richard, the length of that cycle is the biggest opportunity AI offers.
Where AI Speeds Up Root Cause Analysis
The problem-solving cycle has six stages: identifying the problem, analysing it, generating potential solutions, selecting and planning the solution, implementing it and evaluating the result. AI is strongest at the stages that depend on sifting large volumes of data, which are identifying the problem, analysing it and evaluating whether the fix has worked. People remain essential for the stages in between, where experience, judgement and ownership decide what happens on the shop floor.
Shortening the data-heavy stages is what compresses the whole cycle. If AI can detect drift and anomalies as they appear, find the patterns and likely causes in the data, and then monitor thousands of KPIs to confirm that the fix is holding, your team spends far less time finding and analysing problems and far more time solving them. Root cause analysis still needs to be a daily habit, and structured methods such as 5 why analysis still matter, because AI points the team to where to look while the team has to prove why the problem happened.
Why Connected Data Comes Before AI
Many manufacturers already hold enormous amounts of operational data, coming from their machines, quality systems, production planning and ERP. The challenge is that much of it is still disconnected, so nobody can easily see how a machine setting on one shift relates to a quality escape two days later.
Connecting those sources, for example through a manufacturing execution system (MES), lets a business start to understand the relationships between machine settings, production conditions, product quality and performance. That connected data is what AI needs to identify patterns, detect abnormalities and, in some cases, predict a problem before it becomes obvious on the shop floor. Make UK’s report makes the same point in its guidance, advising manufacturers to “ensure pilot areas have clean, secure, accessible data from sensors, machines, MES/ERP systems.”
The practical prize is easy to picture: recognising the conditions that lead to a quality defect before another hundred defective parts are made, or seeing a deterioration in machine performance before it causes an unexpected stoppage, which is where predictive maintenance fits.
AI Supports Problem Solving, People Still Own It
Richard is clear that AI does not replace good operational management or effective problem solving. You still need people who understand the process, challenge the information, establish the root cause and make the right decisions. AI can accelerate the analysis and monitor thousands of performance indicators, but your people have to turn that insight into action.
This is also where many AI programmes in the manufacturing industry stall. In the Make UK survey, over half of manufacturers identified skills and capability gaps as their main constraint on using AI effectively, and 50% said their staff do not have time to train, so technology that arrives without the routines and skills to act on it tends to stay in pilot. Our article on AI adoption in manufacturing sets out a crawl, walk, run approach built on the data you already have.
Four Questions to Ask Before You Build AI Into Your Manufacturing Strategy
- How many variables does your operation have, and how are you monitoring them? Machine parameters, material batches, environmental conditions and operator settings all influence quality, and if most of them are only checked when something goes wrong, problems will keep surfacing late.
- How long does it take to identify and permanently resolve an operational problem? Measure the time from the first sign of a problem to a fix that holds, because that figure is the clearest baseline for judging whether AI is making you faster.
- Do your systems provide connected, reliable data? If machine, quality, planning and ERP data sit in separate systems or spreadsheets, start there, because AI built on fragmented data produces fragmented answers.
- How will you use that information to help your people make better decisions, faster? Decide who receives each insight, at which meeting, and what they are expected to do with it, because an alert that nobody owns quickly becomes noise.
Improving Faster Than Your Competitors
The lasting advantage comes from improving your operation faster than your competitors, and AI is one of the strongest ways to do that when it is built on connected data and run by people who know the process. With only 2% of UK manufacturers in Make UK’s survey reporting AI widely embedded in their operations, the businesses that apply it where problems arise, on the line and in quality, have a real head start.
If you want to understand where AI and connected data could shorten your problem-solving cycle, our Digital Solutions team works side by side with your people to build the data foundations and daily routines that make it stick. Start a conversation with TBM.
