You know what my first point of contact with AI was? It was in 1991, when I was 7 years old and completely enthusiastic about K.I.T.T. – not (only) about the car, butespecially about the concept of an “intelligent” companion, who is always there, helping and supporting where mankind reaches its limits. Today, almost 30 years later, I work with a (real) AI system every day – and my enthusiasm from back then is greater than ever! Why AI technologies are so inspiring, I will outline in this blog post. And if you’re thinking now “This is way too technical for me.” – staywith it anyway! Because it is above all economically profitable!
Clearly defining the construct of artificial intelligence is incredibly difficult. If only because the term intelligence alone can hardly be clearly defined. That’s why AI is currently used almost inflationarily as a buzzword – in principle (almost) always when it comes to machine learning. However, there are some very decisive technological nuances that make the difference, especially in economic terms.
Machine Learning is the generic term for various processes that can be used within an artificial system. Machine learning in itself means “only” that an artificial system is capable of analysing data and uncovering patterns or regularities within the data. Based on this, the system can also evaluate unknowndata. So far, so good. But this kind of machine learning is, in my opinion, far from being an AI. For me, artificial intelligence starts where an artificial system is able to adapt to changes in its environment and to develop itself accordingly. This is where reinforcement learning comes into play.
Reinforcement Learning – a subdomain of machine learning – is a very powerful algorithmic procedure and describes the concept of self-enhancing learning. This means that this type of learning is always linked to a very concrete goal, which is approached step by step. Every single learning step is checked and evaluated: If it serves to achieve the goal, it is rewarded. If it does not, it is punished. Do you already see where this leads? A system that develops itself in this way learns to “think” economically. In doing so, it continuously adapts to the given environmental conditions in such a way that it achieves the specified goal. By theway: If you are now wondering whether an AI is more likely to reward itself with an ice cream or rather with chocolate, just ask me. ;-)
What does this have to do with Dynamic Pricing? Especially in dynamic price optimization, I think it is very important to rely on an AI that develops itself further through reinforcement learning. Because for you, this means that you actively control the algorithm from the outset. You specify the economic goal (such as sales, turnover, profit or margin) and the procedure then learns a strategy for your individual business with which the KPIs you have specified are best achieved. Because your target is also the reward value for the AI, the process will constantly evolve to ensure that your target KPIs are met at all times – even when environmental conditions change (inventory, demand, competition, etc.).
So if you have not yet worked with AI, because these approaches seemed like a black box that you cannot master, the complete opposite is the case. You set the global strategy and the concrete goals – the AI just does all the work for which your time is too precious (and/or has long since ceased to be enough)!
You wonder how this works in daily practice? We can look at it together immediately!
