Artificial intelligence will take over parts of the economy, but not the parts of it that require scepticism, creativity and wisdom

A question: if we decide we want to use artificial intelligence, which part of the intelligence are we handing over to a machine and which part do we retain for the humans?

I’ve been thinking about this lately because it’s now been more than two years since we started using Artificial Intelligence (AI) in our operations at Harrison Manufacturing Company, and my assessment of this futuristic technology is that it works best in highly specific uses, but its broader claims are just marketing. Its weakness is that human beings can be lazy, and they’ll use any application to plug their poor efforts; but at the same time, it’s human scepticism and experience which must be used to control AI.

Humans Are The Problem And The Solution

AI is a tool, and the success of all tools is in the hands of the user.

Let’s start with what is good about AI. Imagine a complex manufacturing process such as ours, which produces many variations of greases, lubricants and additives. Every variation of our product lines has to be made to precise recipes and ratios, using exact measurements of temperature and pressure, and we have to produce these disparate product lines to a certain benchmark of volume-per-day, or we will not be commercially viable.

The applications of AI to this maze of variable processes, are obvious. By using AI in discrete applications, we leverage the ability to not only increase volumes and speed but also improve accuracy.

Producing more volume, speed and accuracy is a Holy Grail for manufacturers and so when the Enterprise version of GPT was released for businesses under the Copilot banner, we started using it.

AI In Manufacturing

We were not going to immediately use a totally new system in our production lines, so we started with AI in our back office. The AI apps are very impressive in areas such as financial management, supply chain and administrative tasks such as receivables.

The applications of AI were less obvious in production, given the myriad processes. However, two lessons emerged from a longish and so-far successful embedding of AI in our production lines.

First, AI does not escape the ‘garbage in, garbage out’ rule of computing. As with any application, from a simple spreadsheet to the most complex financial analysis algorithms, if you feed bad data into an AI app, you’ll get bad results out of it. While that sounds simple, what it means in manufacturing – where software-driven processes operate adjacent and subsequent to one another – is that the entire value chain of production has to be rethought. The value of the AI application can only be realised if it has the quality of data at a certain rate. Remember, AI is really just advanced machine-learning – if one of the machines is advanced and fast, and the other is legacy and slow, you can develop bottlenecks where none were obvious before.

The AI apps in a production line also have to be trained and verified. And you can’t always do this work with another machine; you need the human quality of curiosity and scepticism. There’s no substitute for an experienced person looking at a pressure meter and saying ‘I don’t like that’, or inspecting the first run of a batch and thinking there’s something wrong with it. People can perhaps not outthink an AI application, but they’re capable of being wiser and more sceptical and those are priceless qualities in manufacturing.

Enhancing Employee Productivity

This brings into play the second lesson of AI in our production line: the most effective use of AI does not get rid of employees – it enhances the employees’ productivity.

The best results we see are where experienced technical people and managers have established the most productive role for AI, and trained and prompted the app to produce specific analysis and results in really niche elements of production. 

Conversely, the worst results are where AI allows people to switch off and become too relaxed about something that should involve stress. Manufacturing is a stressful business because getting a small piece of the puzzle wrong can ruin the entire production run. When people start to switch off because they think an AI program will catch the mistakes, then you’d better be sure that AI gets it right all the time, which we know from our peers and ourselves is not the reality.

I’m really happy to be leading a business in the AI era. The opportunities are immense, for speed, efficiency, quality control and even for R&D and product design – areas where we can build an AI design program from Harrison Manufacturing data that we feed into the app. Given that we have our own commercial R&D business, this is a particularly exciting development, given how fast AI can run multiple equations and development pathways when we’re experimenting with innovation.

My Thoughts

Technology is always at its strongest when we use it as a tool, not as a crutch. And just because AI represents the fastest processing of information ever enjoyed by humans, that doesn’t mean people stop being wise and discerning and bringing judgement to a problem.

AI is here to stay. But in our business it’ll be humans using AI, not the other way around.