There are multiple reasons for organisations to fail when working with AI – and I will explain them later – but it all comes down to one development: the growing popularity of AI. Although AI originates from the ‘50s, it has become a real hype since the start of 2016. The hysteria involving AI is heightened due to some interesting aspects.
The rise of artificial intelligence
First of all, successful start-ups parade around with their up-and-coming progress regarding AI. Showing off like this gives the public the idea that AI is there for the taking. Think of the ‘Daily mix’ on your Spotify account or Apple’s Siri on your iPhone.
Another popular AI-application is predictive maintenance, allowing organisations to optimise their operations around maintenance processes. For instance, by analysing large amounts of historical data, the algorithm is capable of identifying key indicators that point out (very accurately) when your production line is likely to fail. Thanks to AI, organisations are able to perform maintenance in a more cost-efficient manner by optimising planning and resources.
Also, last year’s investments show some promising facts, confirming a growing interest in the potential of AI-applications. At the end of 2016 more than two billion dollars is invested in AI related projects. Another sign of its attractiveness is the growing number of issued patents involving AI.
Don’t be left behind!
It has become clear that AI is here to stay. However, these developments also feed the idea that artificial intelligence is the holy grail of supply chain management and other operational processes. Of course, this is not without reason. Not only startups have been using AI successfully; companies in all kinds of industries, like the energy, financial and medical sector benefit from AI as well.
With these and other accomplishments, organisations feel the need to jump in on the fast-moving AI train as soon as possible. CEOs are afraid they miss out on this opportunity and are in a hurry to invest in AI.
However, it is often forgotten that numerous organisations are struggling with implementing AI. Lack of motivation is not the reason for their failure. These organisations are often very eager to make use of AI, and this is part of the problem. Let's get into more detail.'
Reasons for failure
At M3, we see organisations facing similar issues when implementing AI to their internal processes and models – resulting in disappointing performance:
1. Data issues
The performance of AI applications depends upon the data it collects. For algorithms to reach their maximum potential, the data must be of high quality. For organisations implementing AI, this can be a problem. They either don’t have any (relevant) data or they have been collecting data for many years but without a strategy. Once inserted in the algorithm, the data appears to be of no use.
2. Working principles of the model unclear
Managers are so eager to start working with AI that they forget to take a thorough inward look. If applying AI in random places of your model, there is a fair chance that the results will be a disappointment. One of our clients used AI to improve the quality of an input parameter to improve the output of their process model. Unfortunately, the results were disappointing. After we had studied the results, we found that the parameter of entry was not related to the output of this model. Optimising this input parameter was pointless.
3. Simpler solutions exist
Implementing AI successfully is not a guarantee of positive results. When evaluating its outcome, it is possible that the cost regarding time and money do not outweigh the benefits. In some situations, other models are much more sufficient to solve the problem at hand.
Requirements for success with AI
In my opinion, there are three things required for reaching the full potential of AI:
1. Know your organisation, processes and models
Get a grasp of the internal processes in your organisation and make sure you detect the bottlenecks. Listen to the process owners, as they are the specialists who know the ins and outs of the process models that are vital to your operation. This information, allows you to define the parts where optimisation adds value. If processes already function at full capacity, AI won’t increase its results. By eliminating hold-ups and optimising slow processes, results will be worth the investment.
2. Define your goal
What is it exactly that you want to optimise? By how much and in which way? Also, think of whether you are only interested in the outcome of the model. If you want to track the AI’s decision-making process, AI-technology is not (yet) suitable.
3. Understand AI
For a successful implementation and usage of AI, it is crucial to comprehend at least the basics of the processes involved. For instance, the value of the results depends on the quality of the data inserted. Without some understanding of how AI works, it cannot reach its full potential.
Artificial intelligence: not an end in itself
Here at M3, we have studied many AI processes. In general, those companies that successfully use AI know their organisation and internal processes thoroughly, have clear objectives and understand the what and how of AI.
More importantly, successful organisations know that AI is not the answer to all problems. Artificial intelligence is a means to an end, not an end in itself.