Is artificial intelligence the powerful lever for cost reduction and new revenue generation expected by retailers? Without a doubt. In any case, as noted by Pauline Gouache, Senior Consultant in the Strategy team of the Paris office of Equancy, in this forum for LSA, concrete applications are multiplying: personalized recommendation, digitalization of tasks, dynamic pricing or individualized , sales forecast ..., the prospects are very promising.
The number of startups working on artificial intelligence (AI) topics worldwide has doubled in one year, from 950 to more than 1,850! These young shoots arouse a lot of greed: they generated $ 17.5 billion in investment.
All sectors are questioning the potential of artificial intelligence for their business, especially retail players. Indeed, the colossal volume of customer data and product references, the strong issues related to logistics and inventory management, the existence of a giant competitor, Amazon, broken for years to intelligent algorithms are all reasons leading retailers to consider artificial intelligence as a powerful business lever, both in terms of cost reduction and new revenue generation.
For example, artificial intelligence is gradually taking over the entire value chain - from warehouses to stores - through the development of its prediction and image recognition capabilities. This last aspect of AI is born of deep learning and massive investments of Google for about 5 years.
So, to date, what are the concrete use cases of artificial intelligence in retail? Three categories stand out: those that are already unavoidable, those that emerge, and those that still pose questions.
The cases of essential applications
Customized product recommendation is one of the most common and mature use cases. Functionality already essential e-commerce platforms, it is gradually burst into stores via vendor interfaces. In addition, the personalized recommendation is no longer confined to digital interfaces: personalized paper catalogs are emerging. Other advances include the integration of new external data sources or context for a more relevant recommendation. New solutions, designed for the conversational and image-based business era, now incorporate the recommendation into consumer-brand conversation. This is the case of some startup, bot allowing a visual recommendation of products based on a picture provided by the user.
Automation and digitization of logistics, inventory and in-store tasks is another area in which artificial intelligence is advancing rapidly, often coupled with robots and productivity. First example: RightHand Robotics' RightPick robotic arm driven by machine learning to recognize products. He manages the "pick & place" tasks required for the preparation of e-commerce orders very effectively. In the areas of in-store merchandising inventory and optimization, we can mention Bossa Nova Robotics' US robots or the Qopius French visual scanning and shelf analytics VisionWits solution.
Emerging application cases
A potential application case concerns dynamic pricing and the production of real-time offers. If Amazon has been adjusting its prices in real time for a long time (1,736 price changes per minute already in 2013, according to Profitero), the novelty lies in the arrival of real-time pricing in stores, particularly thanks to the installation of electronic tags. For example a French specialist in electronic labeling, has joined forces with Market Hub, a real-time, demand-based pricing solution for tests at Marks & Spencer. In addition to adapting to market demand, algorithms now allow pricing to be adapted to the individual: this is the objective of Point 93, a solution that enables the point-of-sale consumer to indicate , via an app, the price he is willing to pay for the product. Their algorithm validates or not the proposed price, possibly offers another, allowing the brand not to miss the sale.
Another potential component is forecasting trends, sales and store traffic. It uses deep learning and image recognition to detect new trends in fashion and luxury from social networks, including Instagram. Otto, an international e-retailer, uses an algorithm originally developed by CERN in Geneva to predict its sales. This AI, which predicts 30-day sales with 90% accuracy, is now allowed to buy almost 200,000 items a month automatically without human intervention [2].
These examples illustrate that, little by little, artificial intelligence is no longer limited to analysis and adaptation tasks in real time, but to develop