In the post-pandemic, post-Brexit world, businesses of all sorts face a range of new challenges – and many will be wondering if AI-based automation could help them win through. From adding more self-service capabilities for hotel guests through modernising e-commerce fulfilment to replacing missing workers in farming, the opportunities are many, but so are the pitfalls.
Given all this, some research that we carried out last year on attitudes to AI – and in particular its subset, machine learning (ML) – is looking even more relevant now than it was then. It gives a picture not just of where AI could add value, but of key routes to get there and of hurdles that must be overcome along the way.
As well as asking how our respondents perceived AI and ML, and hearing a lot of weariness with the noise and hype, we asked how well their organisations understood “the AI imperative”. We found that a significant proportion were already prioritising AI investments and projects, and that in manufacturing industry – which was a particular focus for the study – the proportion was higher still.
Estimating costs, returns and timescales is a challenge for many
One of the big things that came out in the report though, as we asked about levels of activity, diversity of applications, and so on, was the many uncertainties and hurdles that get in the way of broader AI usage. Estimating costs, returns and timescales featured high on the list of challenges, as did finding or acquiring the necessary skills. But many also said they considered it hard to scope the necessary features and functions, and to specify the components needed to build a robust platform.
So how can you get over these hurdles? Where can you find some certainties – or at least, fewer uncertainties? Fortunately, our survey respondents – and especially those with interests in manufacturing – had some useful thought on which options and routes forward they found most appealing.
The answer for many wasn’t services, it was pre-integrated systems and reference architectures. In other words, it was AI technology platforms and appliances, either purpose-designed to support specific use-cases or adaptable to a range of use-cases.
What’s the appeal of purpose-designed AI appliances and platforms?
Significantly, our subset of manufacturing respondents found these even more appealing than the mainstream did. Why would the industrial sector see more appeal than the average in approaches such as reference architectures and appliances?
Part of the answer may be that on-prem infrastructure is the norm in much of manufacturing industry. But there is also the engineering and industrial mindset: if you have a requirement or challenge, and a tool or proven design exists to deal with that, then assuming that you can get the funding and demonstrate ROI, you buy or build that tool or design and you put it to work.
However, when the tool is as complex and has so many dependencies as AI, putting it to work is easier said than done. Indeed, it needs to be a multidisciplinary team effort – and sure enough, this was the last hurdle that our survey revealed. If you’d like to learn more, you can download the full report for free here.