I’ve been talking to a great deal of business and industry groups about artificial intelligence (AI) recently. Despite their initial skepticism that AI is the latest technology buzzword, it has been uplifting to hear how many companies are now contemplating their own deployments.
However, many are holding back from committing fully due to concerns over the complexity involved. And I believe that we as a technology industry are at fault here, for having focused too much on the technology and not enough on its application.
This became clear to me at one of the events I recently participated in: a customer advisory council in the UK. This brought together a group of companies representing a broad spectrum of industries – all of whom are approaching IT in an innovative way. My role was to provide an overview of where I think artificial intelligence is today – as well as where I think it is going – and as you can imagine, this prompted a great deal of interesting discussion and debate.
The first thing that really stood out for me is that every one of the companies there is looking into AI, but interestingly, none have yet embarked on a major deployment. My biggest take-away is that as an industry, we’ve focused a little too much on picking AI apart and defining all its different elements – from neural networks, to machine learning and deep learning – when actually all businesses want to see is actual use cases, to understand how it might benefit them.
We need to move the dialog away from the technology and focus more on the multiple ways in which it can be applied.
Another insight I gained was that AI-related case studies are convincing, even across industry boundaries. I described the work we’ve been doing with Siemens Gamesa, using AI to take away the rather monotonous process of running in-depth quality control checks for their fiberglass wind turbine blades – helping to accelerate and improve the quality control process.
This AI story resonated well with every organization I spoke to – and after all, what manufacturer wouldn’t want to improve the process of detecting and eliminating defects in its finished goods?
Another result of our industry’s focus on the technology not its benefits is a general perception that it is highly complex and expensive to run a dreaded six-month proof of concept, and that you need teams of PhD data scientists to implement. But that’s really not the case.
You don’t need to run a proof of concept for six to 12 months – in fact, you can get a clear idea of whether a business can achieve what it wants with AI in a month flat: that’s how long it typically takes to confirm whether they have enough of the right kind of training data available for a machine learning system, or whether they have an expert whose knowledge can be tapped into.
I experienced similar feedback when I gave a keynote at the Global Industry club at the Hanover Messe fair in Germany. The meeting wasn’t exclusively about AI, however every participant at my session was already energized by the range of cool demos that they had seen on the show floor – the event had a strong showing of exhibits relating to automation, robotics, machine learning and AI. This helped drive a very lively debate about the possibilities presented by AI technology.
It’s not your product, it’s your process that should define whether to implement AI
One frequent misconception about AI is that only technology companies can benefit from it. For example, when I start to talk about AI in the automotive industry, most people immediately think I mean autonomous cars. Actually, there’s a huge amount of potential for AI in terms of optimizing manufacturing processes.
And that’s the key. The transformative potential of AI generally lies in the manufacturing process, and not the product at all – it’s all about improving how you make it. If you consider one of the least technical products imaginable, such as chocolate chip cookies, there are manufacturers who are creating significant value by streamlining their production and reducing waste with incredibly automated processes driven by AI.
Earlier this year I voiced a concern that, as AI expands into new fields, then as a society we need to carefully monitor how we use it. I was recently quoted in a British Parliamentary report that emphasized the strategic importance of civil society in shaping the conversation on how AI must benefit individuals and society as a whole. However, from many conversations, I have the impression that most proofs of concept are running without government support or financing – as it just takes too long.
But it’s a case of striking the balance. We need to both allow competition and also protect society from potential risk. My recommendations were to remain mindful of how we use the technology, and its potential consequences. Also, we need a legal system that can keep pace with the speed of technical change. That said, most of the industrial applications of AI are very specific – there’s not going to be a societal effect triggered by a business implementing predictive maintenance.
For many businesses, there’s actually a great deal of low-hanging fruit in terms of potential AI deployments. Most manufacturers have been collecting data for decades and have more than enough to feed a machine learning application. But the key to a successful AI deployment is being very clear about the issues you want to address.
For this you need an expert with segment specific expertise. A generic AI expert isn’t going to be able to get the job done. You need someone who understands the intricacies of their complex supply chains and production processes to identify how to enhance them with AI and related technologies.
That’s why our digital co-creation approach works so well in these situations. We collaborate closely with our customers to define the issues they want to address and focus on getting a proof of concept rolled out quickly. Also, we have a growing global network of Digital Transformation Centers that are designed specifically to help customers at the ideation stage, in clearly defining projects and arriving at conclusions more quickly.
Most industries have already come to the conclusion that they must implement AI to remain competitive. The question is not if but when. My recommendation is to explore the potential uses of AI immediately, as almost all industries have processes that can benefit.
The other danger they face is being overtaken by the cookie manufacturers of this world, who are quietly turning their manufacturing processes into showcases for automation and AI. At the same time, as an IT industry, we need to do a better job of helping our customers to join the dots between use cases and explain how they can derive significant benefit from AI.