Thinking about adopting AI? First, find your Yoda

Manufacturers are racing to adopt artificial intelligence just as decades of human expertise near retirement. Once that knowledge walks out the door, no algorithm can bring it back.

Nick Leeder, a strategic advisor to ITI Group with 20 years’ experience helping companies turn digital ambition into tangible value, explains why capturing people’s knowledge should come first in any technology initiative.

I delivered a keynote at this year’s National Manufacturing Summit on the real barriers to adopting digital tools. When I asked the 200 manufacturers in the room about their digital maturity, 60% rated themselves as low. That should be the starting point for any conversation about AI in manufacturing.

That’s not to say the ambition isn’t there. Research shows two in three manufacturers want to move quickly with digital, including AI, with most planning to increase investment and proof of concepts over the next 12 months. But ambition and foundation are two different things.

Most manufacturers – and I’m talking about the SMEs that make up 99% of UK’s industry – don’t have the strategy, data infrastructure or process maturity to support an AI programme today. They’re too busy fighting energy costs, fluctuating demand and geopolitical uncertainty.

That isn’t a reason to wait indefinitely. But we have to be honest about where manufacturers actually are and how they move forward. Because that 99% – so often described as the lifeblood of manufacturing – is where AI will create the most value, not the large 1% who already have the scale, budget and infrastructure in place.

‘Let’s start using AI’ isn’t a strategy

I spoke recently with a board member of a UK PLC whose board had just mandated an AI strategy. When that was announced, he put his hand up and asked, “Hold on. What are we actually trying to solve?” That is the only question that matters. If you don’t know what you want to improve or reduce, what’s your basis for deciding where AI can genuinely help?

And AI absolutely can help. But only when three things are in place. First, a process that actually works. A flawed process plus AI only gives you faster, more expensive mistakes. Optimise the process first. Second, competent people making informed decisions within that process. Their judgement is what will shape and train the AI model. Third, data generated by that optimised process, validated and contextualised by those same competent people. Only then does AI have something meaningful to learn from.

This is why shoehorning AI into every use case is the wrong approach. If it doesn’t make work faster, clearer and easier, it won’t move the needle in terms of your performance.

Who understands your processes better than anyone else?

Every factory has that person. The maintenance engineer who’s been there 25 years. The one everyone calls when something goes wrong. They walk the line, hold a broomstick against a machine and know, by sound and feel and instinct, whether something is about to fail.

That person is your Yoda. And in most manufacturing businesses, they’re approaching retirement. When they leave, they take with them decades of tacit knowledge with them. Knowledge that currently lives nowhere except inside their head because no process or system or dropdown menu has been designed to capture it.

This is where AI in manufacturing gets really exciting. Capturing that tacit knowledge as part of the work itself, not as an additional step, using sensors that record what the machine is doing and what the operator does in response. The adjustment they make, the correction, the override and, most importantly, the reason why.

Raw machine data, combined with operator context, becomes the foundation of a model trained on real process expertise. This is what ‘human in the loop’ means in practice. Systems designed to learn from and support skilled operators, not replace them.

Get this right, and you create an advocate. Someone who once saw AI as a threat now sees it as a tool to make their job easier. And because Yoda is also the person the shop floor trusts, once they back something, others follow and you create momentum very quickly. We tend to forget that part. We focus on financial ROI from technology and overlook the cultural ROI that comes from driving positive change through its adoption.

Start with condition monitoring

If you’re looking for a practical first AI use case, predictive maintenance and condition monitoring is hard to beat. It’s not that technically sophisticated, the investment required is manageable, and the impact on OEE and unplanned downtime is measurable in a way that everyone understands, from the shopfloor to the boardroom.

A food and beverage manufacturer is using a condition monitoring system with vibration sensors on critical machines. The data generated is validated by a vibration analyst and then fed back into the system to improve model accuracy. It’s a practical example of skilled human judgement overseeing and validating machine intelligence in real time. It’s early days, but already the factory has seen a 30% reduction in unplanned downtime.

Beyond the financial impact, it has also made the maintenance team’s job tangibly better. Instead of constantly firefighting and coming in at the weekends because machine 3 has gone down again, condition monitoring has got them ahead of the problem. They’ve moved from reactive intervention to controlled prevention.

Find a problem worth solving and the person who already understands it

Manufacturing is and will remain a people game. So when leadership teams ask me what their AI strategy should look like, I always come back to the same thing: a problem worth solving. Then the question becomes whether you have the process, the people and data to solve it properly.

All three depend on Yoda. Find that person. Chances are, you already have them in your business. They understand your manufacturing process better than anyone else. They’re just not being used as the asset they are.

Visit ITI Group at Smart Manufacturing Week, NEC, 2nd and 3rd June, stand F144, to discover how we’re driving smarter, more efficient manufacturing solutions.