fbpx

Artificial Intelligence – 3 Dos and 3 Don’ts for Manufacturers

The National AI Strategy

Last month, the Government published their National AI Strategy, laying out their 10-year vision of Artificial Intelligence in the UK. Curiously, two key statements caught my eye:

1)     The UK is already a global Artificial Intelligence superpower, with access to talent, a culture of innovation and a progressive regulatory/business environment.

2)     AI has the ability to increase resilience, productivity, growth and innovation across public and private sectors.

As someone who works with manufacturing businesses every day, I know that increased productivity and business resilience are both highly desirable. Many of the people we work with dedicate their entire careers to the pursuit of these two goals.

upwards trending graph on paper

All of this left me wondering, if AI can fulfil these key business needs, and the UK is already a global AI superpower, then how come we rarely, if ever, see AI providing value on manufacturing shop floors?

More importantly, what should be the AI strategy for manufacturing to achieve these benefits?

To answer this, I decided to distil my thoughts into 3 dos and 3 don’ts for manufacturers wanting to successfully harness the capabilities of AI.

Tech Push or Business Pull?

Technologists, myself included, are often guilty of implementing tech for tech sake, but to businesses who need to defend margins, win work and keep their operations running, this is not a luxury they can afford.

Every action they take has to add value and benefit the business. With technology, it’s no different.

Simply put, it has to be a business pull, not a technology push.

man pulling gray rope

This brings us to the first (and most important) do/don’t of artificial intelligence for manufacturers:

DO: Implement projects that provide quick ROI and measurable business value.

DON’T: Choose far-off, hard to measure or trivial projects that are more fun than functional.

Aligning a technology project to your current business needs is one of the most effective ways to extract short-term value. Here’s why:

  1. It will be easier to get buy-in from other staff members because they will have first-hand experience of the problem you’re aiming to solve. E.g. we don’t know our machining capacity so we find it hard to schedule and meet lead times.
  2. Similarly, You’ll know very quickly if the project is working or not because you’ll have first-hand evidence of the project’s impact on the problem you’re tackling. E.g. we have been able to measure our machining capacity, giving us better scheduling and reduced lead times.
  3. Because you can quickly see the results, you can minimise your risk by dipping your toe in with a solution and evaluating its performance over a few weeks or months. If it works, great! Time to roll out. Otherwise, you can cut your losses without much time or money invested.
  4. Because of the low risk and quick evaluation, you can afford to move quickly, avoiding analysis paralysis, and taking decisive action to drive your business forward.

The Minimum Viable Project

There is a joke in the world of start-ups that any product can be prototyped with an Excel Spreadsheet, chewing gum, a sheet of paper and a mobile phone.

person holding pencil near laptop computer

It’s a bit of a stretch, but at its heart is an important point – you want to determine if a project has value very quickly, without sinking time, effort or money into it. Therefore, you should aim for a Minimum Viable Project that allows you to do a low-risk evaluation of the general idea before jumping in with both feet.

This brings us onto the second do/don’t of artificial intelligence for manufacturers:

DO: Create a quick, simple and easy-to-implement solution to prove the general concept  

DON’T: Use artificially intelligence (unless you’ve reached a point where it adds value)

It sounds pretty counter-intuitive, but you’re always better off starting off with a rough-and-ready implementation that allows you to learn, iterate and improve within short timescales, rather than trying to dive straight in with an all-singing-all-dancing AI-powered solution that takes years to develop.

Don’t believe me? In which case, let’s look to Google – they have published their 43 rules of implementing machine learning, a dominant subset of artificial intelligence. Here is rule number 1:

“Don’t be afraid to launch a product without machine learning” (source)

It seems that even the big players in machine learning value building a simple solution without the bells and whistles to start with, followed by iterative improvements, adding the AI smarts further down the line when they’re required.

Do What You’re Doing, But Better

If you’ve seen any coverage in the news about artificial intelligence, the tag line was almost certainly about how it will revolutionise our world, change the way we do business, and bring in new ways of working that are beyond our imagination.

Whilst I admire the ambition and agree with the opportunities, I feel this messaging is unhelpful. It makes it seem like anything related to artificial intelligence has to be a massive leap, not an incremental gain.

silhouette photo of man jumping on body of water during golden hour

Let’s face it, massive leaps are risky, scary and hard to predict – all of which are barriers to adoption. Incremental gains, on the other hand, are much easier to manage, have more defined value in the short term and are altogether more palatable.

When looking at the fundamental technologies of AI, they simply enable us to perform the same computational work we currently do, but with greater speed, greater intelligence, or deeper complexity.

There are very few things that AI can do that you couldn’t do yourself with many sheets of paper and a pen, it’s just that AI will do it millions of times faster with far more accuracy, and without needing a coffee break in the middle.

With this in mind, we want to avoid a big leap of faith, and instead opt for incremental progress in the direction that we’re already travelling. With that, we come to our third and final do/don’t of artificial intelligence for manufacturers:

DO: Use artificial intelligence for incremental improvement of your existing processes

DON’T: Try and revolutionise your business, processes or people with a moonshot AI project

In manufacturing, we strive for continuous improvement, not abrupt changes. So it only makes sense to aim for the same when it comes to artificial intelligence.

gray vehicle being fixed inside factory using robot machines

The reason this is so important is that technology only really exists to support people, industries, and ways of doing business. All three of these gradually change over time, allowing gentle adaption, learning and eventually success through the right blend of novelty and familiarity.

To Summarise

I believe AI can provide genuine value to manufacturing businesses if the dos and don’ts we’ve seen in this article are kept in mind. I don’t believe AI should be as scary, or as revolutionary as we are told it is.

Much like a horse, a steam engine, or a car can get us from A to B with various levels of speed, I believe AI is just a new tool to support manufacturers in doing what they already do, but better

Leave a Comment

Your email address will not be published. Required fields are marked *