Why manufacturers shouldn’t grow their own coffee beans: Tom Clayton

With AI coding tools on every laptop, building your own industrial AI is tempting. Many companies are trying it with the best of intentions. In practice, they are exposing their businesses to unexpected costs, unknown liabilities and massive cyber risks, in a field moving faster than any in-house team can follow.

Some of the world’s biggest manufacturers have already learned this the hard way. At a recent London industry event, the head of digital transformation at a large consumer goods group described how their company had tried it all. It couldn’t keep up and the costs mounted. The conclusion was simple: stick to what you make and buy in the expertise. You might love a good coffee in the morning, but you don’t grow your own beans.

The mistakes are well understood. The money goes in and an ambitious project manager is appointed, but without deep knowledge of the thousands of processes in a factory, the pilot never ends. Then the person who built it moves on, leaving behind software nobody else understands.

The biggest risk is security. The people building these tools usually have wide access across the business, from production line controls to servers holding sensitive customer, supplier and staff data. Written by AI and not properly vetted, the code becomes a wide-open door. Is that really a risk worth running?

It’s said that every poet is a thief. The same is true of good consultants, who go from business to business solving problems and applying the solutions to the next client. Their customers get proven expertise, without the endless pain of discovery and invention.

IntelliAM AI was born out of consultancy in South Yorkshire and now serves many of the world's biggest food and drink producers. If a manufacturer needs to monitor a separator, we have almost certainly done it before. Our platform works with the sensors and equipment a factory already has, whoever made them, bringing data from across a site into one place. And because that data is organised in one consistent structure, we can take advantage of each new generation of AI model as it arrives, without rebuilding the foundations.

In our experience, the most successful manufacturers own the pilot and leave the machine learning to specialists. Before starting any project, boards should ask four questions: When should we adopt it and who needs to be involved? What does success look like before we scale? Who will maintain the code when its author moves on? Can we trust it?

We prototype tools quickly with AI, ring-fenced from the live plant, then our engineers build them properly, so every line of code is known and maintained. Manufacturers’ digital transformation teams play a vital role, safeguarding systems and protocols. But the factory keeps doing what it does best: serving your next pint, baking your daily bread and making the chocolate you can’t resist.

Tom Clayton is co-founder and CEO of IntelliAM AI