Artificial intelligence (AI) is wildly popular, and it seems as if it is omnipresent applied. It is the foundation for your social media timeline, and with almost every interaction online follows an AI implementation providing you with more relevant suggestions based on your interaction. But useful applications that go beyond simple recommendations are hard to find. Why is that? Organizations find it scary to rely on AI.
The state of the art of AI
The power of artificial intelligence (often used as a synonym to machine- and deep learning) lies in the combination of the application of mathematics, data, and technology. It is one of the many technologies utilized by the data scientist. Thanks to the availability of large quantities of data and computing power, it is a technique that quickly performs better and more consistent than humans when it comes to specific, repetitive tasks. AI is thereby growing even more rapidly into an application that organizations can and should use. But there is so much more to achieve with AI than currently is the case. I believe that the next step for that should be to gradually expand the application: from predictive to eventually autonomous decision-making.
Most applications of artificial intelligence are currently limited to the role of predictor. These include the application of AI in Spotify and Netflix, to suggest the next artist or film. These predictive models, also known as ‘recommenders’, are based on the content that you and similar users have consumed in the past. The recommendations forthcoming from these predictive models are often highly relevant. So, we receive many requests to create similar predictive models. It is definitely one of the things data science is all about: making predictions that border on certainty by applying machine learning to large amounts of data. But aside from making predictions, AI presents many other opportunities that have a positive impact on businesses. Condition is, however, to make people trust the algorithms to such extent that they allow them to make decisions autonomously.
As artificial intelligence approaches the realm of making autonomous decisions, we have noticed that organizations begin to become more apprehensive. Suddenly, computers are partially responsible for success or failure. Take an insurance company chatbot, for example, which is used to support the approval or denial of damage claims. This is a form of artificial intelligence that carries responsibility for customer contacts and is capable of assessing the claim according to policy conditions, all without human interference. But one thing that I have noticed is that the system is still mainly used to collect all data to subsequently make recommendations to an employee. The process is of course much more efficient and less time-consuming than processing claims manually. But at this point, there are still humans involved to check the recommendation and make the final decision on whether or not to pay out the claim. The fact that algorithm still don’t make the final decision is not a technical limitation, but rather a question of trust. Trust in the autonomy of an algorithm must, as is the case with a new human colleague, naturally first grow. This is all too understandable: considering the debacle of the Tay chatbot, which suddenly began posting racist statements online. So what would happen if we allow AI to act autonomously?
Let us zoom in on an application of artificial intelligence that lets mathematical models make business-critical decisions autonomously: the dynamic pricing algorithms. These algorithms determine the best price for a wide range of products and services without human interference. Both theory and practice indicate that this process results in the maximization of revenue and savings over the long term (the overhead costs for manual calculations and administration are dramatically reduced). Take the American company Walmart, for example. The company uses dynamic pricing algorithms that adjust prices almost 50,000 times per month. The model already realized 30% more online sales in the first year after it was implemented. But the period before implementation was undoubtedly nail-biting, as it takes a lot of courage to leave something that critical completely to a computer.
Dare to let AI make decisions
Artificial intelligence is a fantastic technique, and enables you to make better data-driven decisions. In practice, AI applications are mainly predictive and rarely permitted to make decisions autonomously, let alone business-critical ones. Predictive applications of AI are still valuable, but the time has come to place greater confidence in AI, in order to realize even greater efficiency and higher profits. The technique is available (and advanced); so it is simply a matter of overcoming our fears. Dare to trust in the algorithms when making decisions, and let AI become a fundamental element of your operations.
Are you wondering how artificial intelligence or other data science techniques can help your organization make better decisions and improve your efficiency, predictability, effectiveness, and sustainability? Let’s get in touch by contacting us here or follow our LinkedIn to stay up-to-date.