AI in retail – a game changer?

The topic of artificial intelligence is currently dominating specialist and popular media. The latest wave of euphoria is fuelled by the development of the new chatbot ChatGPT. So it’s time to take a look at how AI is being used in retail.

The current fuss over ChatGPT distracts from the fact that AI is not a completely new technology. People who are already familiar with the topic also know that it is always wrong to talk about AI as a uniform entity, when it is actually a combination of different technologies. And some of them have long been in use in the retail sector.

Automated stores and self-service

For some time now, REWE Group has been opening branches where people can shop without having to queue at a checkout. They don’t even have to scan the products. This is made possible using technology developed by an Israeli startup. The approach works solely through the use of artificial intelligence.

In this case, with optical pattern recognition. Machines can’t see in the same way that humans do. They can’t read like us, either. This means that algorithms must be used to evaluate the data from cameras and sensors in a way that creates digital information.

So the challenge in these supermarkets is for the system to reliably recognize and track a person as they make their way through the shop. It must also recognize the goods correctly without needing to read a barcode. An example of the power of artificial intelligence in everyday life.

The logistics sector already uses AI, too

As we all know, the coronavirus pandemic gave digital commerce a massive boost. More and more people are shopping online. And more and more parcels are on the roads. For delivery services, AI helps to optimize routes in real time. It draws on the latest data and takes into account traffic congestion information, which is then fed into a system for forecasting.

In general, predictive analytics (i.e. computer-based forecasting) plays a more important role in retail and was already in use before ChatGPT dominated the news.

AI supports retailers with dynamic pricing for online marketplaces and provides forecasts of future order quantities and sales. This is because pattern recognition can identify trends much faster than we humans can, even in vast amounts of data, making it easier to identify future top sellers.

Machine learning and machine analysis help (online) retailers to evaluate shopping baskets, regardless of how many shopping baskets are involved. As self-learning systems, they submit additional offers to customers, auto-compensate and become increasingly precise in their predictions and therefore more accurate in subsequent offers.

ChatGPT: impressive, but don't overrate it

“Write a blog post on the topic...” If this input results in a longer text that sounds convincing and is elegantly written, we humans are understandably impressed. All the more so when such a system is as easily accessible as ChatGPT.

In fact, the performance behind this bot is impressive. Nevertheless, when using it, we must not forget that ChatGPT does not “think” and is not “creative” like we humans are. At its core, the GPT-3 language model is still something that is referred to in professional circles as a “stochastic parrot”. Because “parroting” is a true description of what it really does.

The system imitates humans based on statistical probabilities. It accesses countless data and information, summarizes it all and then assembles the content. In other words, the answers that ChatGPT provides are based on thoughts that someone has already had and which have also already been published somewhere.

This is in no way meant to diminish the achievements of the developers. Systems such as OpenAI can already make everyday work easier for knowledge workers − for instance, when they reliably summarize page-long texts. And without doubt, the experience gained with ChatGPT will be helpful for the development of even better chatbots – improving communication with customers via voice or chat and making them seem more natural.

But we are still a long way from a global revolution.