Artificial intelligence and machine learning are now a firmly established part of our private lives. But these new technologies not only act as digital aids in our everyday lives. They can also be leveraged by retail businesses, in particular.
Artificial intelligence (AI) has become part of our lives. Siri and Alexa are being adopted as personal assistants. Facebook recognizes faces in photos. And autonomous driving is a hot topic. The potential benefits of AI are enormous - both in a private and a work context. Furthermore, the dramatic evolution of IT in the past few years, and its continued and constant evolution going forward, present a major challenge to businesses. It is not easy to maintain a competitive edge by staying constantly up to date and abreast of developments. And enterprises who miss the boat will find it extremely difficult to catch up again.
AI and machine learning (ML) are playing a growing role in the world of work, as companies strive to become intelligent enterprises in which automated algorithms take decisions and manage processes autonomously. Against this background, it is vital to identify and exploit the potential benefits of AI and ML to master each industry’s unique challenges. How can pioneering technologies support everyday work and help enterprises to advance? Where can AI be deployed and what added value can it deliver? What are its capabilities today, and what are they likely to be tomorrow?
Moving from keys to touch: The way we operate machines and technical equipment is changing rapidly. Smartphones are a prime example - today’s cellphones are almost completely touch-operated, and the number keys are no longer needed. The next step, which is fast approaching, will be intelligent voice control. Siri, Alexa and co. show us how it works. How can this trend be harnessed in e-commerce, for example? Is it possible to shop online by voice alone?
But there is a major challenge: Spoken language varies greatly. First, there are different dialects and forms of communication - not to mention facial expressions and gestures, which are an essential part of spoken language. And second, there are a great many ways to express something in a given language . There is never just one way of saying it. For example, a variety of phrases can be used to make a purchase: "Please buy", "Conclude the order", "Place binding order" or "Please send me ...".
This means the system must be able to recognize what the purchaser is saying and what they mean. So it requires appropriate training. Before a product is placed in the shopping cart, there needs to be a clear, unambiguous interaction between the purchaser and Amazon’s Alexa, as well as communication about the product, such as the desired brand, size and color in the case of a pair of jeans.
One solution is to specify predefined phrases which the speaker has to know and use to control the system, such as "Place in shopping cart", "Complete shopping cart" or "Order goods". For example, Amazon’s voice-control system, Alexa, can be connected to the SAP Commerce Cloud (previously called SAP Hybris Commerce). But there are still some hurdles to overcome before voice-controlled ordering - for which KPS has developed a demo version - is fully fledged. In fact, this depends to a large extent on the willingness of brick-and-mortar retailers to disclose their end-to-end order processes and buyer behaviors to their competitors and to major platform providers, or alternatively, to develop and set up their own platforms.
All businesses depend on sales reports, broken down by month, product or region, to stay informed. The main challenge is to obtain this information quickly and in an easy-to-read form, as it can be very time-consuming to retrieve the right reports, even when a professional reporting tool is deployed. Report retrieval can be simplified by means of chatbots that act as talking assistants. The chatbots can respond to employees’ queries via voice and text, helping them to find the right report with the parameters they require. This way, the employee no longer needs to type in a query, but simply asks the chatbot: "Find me the sales report for product X in region Y for September 2018.” The system then finds the corresponding report and outputs the numbers by voice or text. Supplementary data can be requested in the form of downloadable Excel charts. The charts also provide the employee with relevant in-depth information, such as an overview of sales over time.
This may sound simple enough, but it is in fact an extremely complex task. With the help of AI and ML, the salesbot developed by KPS learns to interpret the user’s words and commands. The system is always configured for a specific language, i.e. for the words and phrases used in that language. The intelligent system picks out the keywords from the user’s voice input , and can also learn to recognize speech patterns. For example, different pronunciations of words are recorded and correctly assigned. Speech-based interaction with the system makes it quick and easy for employees to use. In addition, it minimizes the time and effort involved in searching. To guarantee data protection, employees must be authenticated via browser before they can use the system to access reports.
The next step in speech recognition will be to identify a person by their voice alone. ML-based identification through image recognition is already possible, and this capability must now be implemented in voice-controlled systems. This is a significant challenge, bearing in mind that patterns of speech are far more variable than visual patterns.
AI can not only play a role in voice control and image recognition, but it can also supply information on products that offers valuable insights - for planning, for example. Digital capture of product information gives businesses an accurate picture of product lifecycles and product quantity-time curves
associated with the warehouse or the store. ML can be used to provide and integrate further information not related to specific products, including environmental factors such as weather data, or local event or appointment calendars. For this purpose, the algorithms are supplied with actual data. The algorithms learn by checking the data is correct, and become better at responding to a given situation over time. The entire process is fully automated and not dependent on predefined rules. For example, a supermarket needs to offer more meat and charcoal when good weather is expected and large numbers of people will be having barbecues. If a large-scale party is scheduled for the coming weekend in the neighborhood, an increased supply of drinks will probably be required. AI enables businesses to plan their orders and resources with greater precision.