A look at some disillusionment aspects of the application of Artificial Intelligence in Industry
In various articles we have discussed the development of Artificial Intelligence solutions most in demand in the market that we find ourselves implementing at our clients. However, not all projects succeed in bringing the benefit that client and consultant assumed when defining the work.
In this article, however, we will not focus on the purely numerical aspects, such as data quality or measurable performance of the models developed: in fact, even the best-trained algorithms sometimes struggle to get out of the context of the pilot and into production.
Machine Learning as a factor of culture
In many advanced analytics projects, as indicated by this report, the cultural factor can be a strong brake on the development of innovative solutions but, unfortunately, to some extent also disruptive.
The change must be accompanied with a path of growth and not a path of rupture, in order not to create in the man who suddenly finds himself working with a machine, a relationship of suspicion and distrust.
The impact of any form of Artificial Intelligence in human work is as timely a debate as ever: when bringing such a tool into a company, it is therefore important to prepare a ground where it is seen as a facilitator, not a substitute for human operations.
The goal to focus on should be supporting human activity, not hindering it: any intelligent support system designed in such a way as to disrupt or reduce the efficiency of the human operator is doomed to failure.
The role of smart tools has already been discussed in thearticle dedicated to the idea of Industry 5.0 launched by the European Commission inspired by Covid’s impacts on human work, starting with teleworking and city transformation.
In other words, the development of artificial intelligence solutions is not in itself a sufficient condition either to ensure the success of an organization or to make its adoption within the company automatic: in fact, a solid culture of data must be built, which must somehow be accepted and recognized even on planes other than the merely technological one.
Measuring the concrete model advantage of Machine Learning
As with any project, when it comes to applying Machine Learning systems in industrial processes, the investment is justified only when a benefit can be realized.
This benefit could be economic: direct in terms of resource savings, or indirect, in terms of, for example, time savings in the processes involved. Or it could lead to a qualitative improvement: for example in the customer’s buying experience or in simple terms of image.
However, some projects are started as PoC (Proof of Concept) or Pilot Project because of the low cost of a small-scale first trial of an Artificial Intelligence system, without having to concretize and quantify the benefits right away.
When the PoC achieves sufficient results to justify real deployment, the need for a significantly more impactful investment, partly because of architectural aspects, is met with the difficulty of calculating the time required to return on the investment, the so-called ROI (Return on Investment).
Moreover, even where one has been able to identify the problem that a Machine Learning model could solve and has quantified the threshold needed to make its implementation worthwhile, one happens to run up against the inability to achieve in the Pilot Project sufficient levels of precision or accuracy to justify the worthwhile investment.
For this reason, the fascinating world of machine learning brings with it a sad reality: so many advanced and very interesting projects will remain only a dream in the drawer.
How to prevent an AI technology-based project from failing? Regesta LAB is your ally in recommending arti the most suitable solutions and strategies for your business, tailored to your needs.
