Artificial intelligence (AI) and machine learning are becoming ubiquitous in our daily lives. Modern AI systems are less like David in Spielberg’s film “A.I. Artificial Intelligence” or the HAL computer in Kubrik’s “2001 – A Space Odyssey” but help us to steer through our daily life more efficiently. From self-driving cars to scheduling appointments, AI-based systems are there to help. You probably encounter some AI system or another a few times every day, like photos being tagged in your social media account, schedules optimized in your calendar app – and the right notifications being sent at the right time. If you look at your phone, most apps you use incorporate AI and machine learning at some level.
How can retailers benefit from the potential AI can offer?
The challenges modern retailers face can be loosely grouped into two areas: Those aspects of the business that involve the customer – and those that don’t.
Improving customer experience (CX) is one of the big challenges any retailer faces. As the online and offline world become more entwined, e-commerce and omnichannel retailing become increasingly important. Customers increasingly use webpages and smartphone apps to look for products and interact with the retailer, other customers or their friends.
How can retailers improve their customer experience?
- A personalized “Personal Assistant”: When shopping for new clothes, sometimes we want to ask if a garment is available in a different size or color. Shopping online, we want to know about payment or return policies. Looking for an assistant in a store or trying to find the answer in an FAQ section can be frustrating. A new range of Personal Assistants is emerging (like those powered by IBM Watson) that bring the quality of the conversation with such a Personal Assistant or ChatBot to a new level. AI systems allow more natural conversations, asking questions in a way we’d also ask people – and the AI systems are able to reply in kind.
- The “virtual changing room” Buying clothes online is always a bit of a gamble. Not only can it be difficult to judge what the item of clothing will look like in reality (e.g. how the fabric, the pattern or the colors look like in reality), it is quite challenging to judge how the new dress or shirt might fit without trying them on. Specialists like com offer virtual changing rooms. Using a picture and a few measurements, their machine learning algorithms are able to visualize how the new shirt would look on you, giving you a much better idea of whether you would like to buy it – and if the size and color are right for you.
- Personalized Fashion Recommendations: Amazon took the virtual changing room a step further with the introduction of the Echo Look. In addition to a microphone array and speaker, this device also includes a depth-sensing camera that takes pictures and videos of you wearing an outfit from your wardrobe. Over time, Amazon gets a full view of all the clothes in your closet and can then advise if a particular combination looks good or recommend new products.
Engaging with the customer and being able to offer the right choice at the right time, addressing any needs that might arise is crucial to any retailer’s success. However, operational excellence is equally important. Closing the deal but not able to fulfil it is a sure path to losing customers.
- Stock management is the foundation to optimizing any retailer’s operational business. Even in the day and age of electronic inventory, it remains a challenging task. Depending on the business and the value of each product, equipping each item with RFID tags allows precise tracking of not only how many items are in stock but also where they are located. However, in many situations (e.g. a supermarket), introducing such systems are not always practical. Simbe Robotics recently developed the robot “Tally” that drives around a store and records inventory level on shelves. As this can be done multiple times per day with customers present in the shop, inventory can be tracked throughout the day, freeing customer service agents to help customers in the shop. Accurate inventory data is a key ingredient to successfully implementing our next item.
- Automated replenishment: Optimizing replenishment helps retailers in several ways: Keeping inventory levels at optimal levels ensures that customers can always (at least up to a specified service level) buy the products they want while minimizing bound capital, transport and logistics and, in case of perishable goods, write-offs due to product spoilage. Modern AI-based systems can include accurate data from inventory management systems, data from historic sales patterns, as well as external influential factors such as weather, local events and holidays, etc., to accurately predict future demand, which is then used to optimize the inventory levels, including possible constraints from lot size, delivery cycles and lifetime of a specific product. The optimization is done according to one or more specified key performance indicators (KPIs) such as profit, stock-out rate, waste or bound capital. Users may also choose any weighted combination of these KPIs or even “soft” KPIs such as customer satisfaction as their main optimization target. However, in the case of such soft KPIs, careful definition and the ability to accurately measure them are key. Defining a strategy via a number of KPIs, which should be optimized, then allows the AI-based replenishment system to make optimal ordering decisions for each individual product at each location or point-of-sale.
- Dynamic Pricing: What is the best price of a given product? Setting the “best” price is one of the core tasks for any retailer. However, what might seem feasible when considering a single or just a few items becomes an unsurmountable task once tens of thousands of products across hundreds of stores have to be considered. Furthermore, prices may be allowed to vary according to current or desired stock levels, promotions, competitor information, etc. Instead of setting prices directly, modern AI-based pricing systems are set by a pricing strategy for individual articles, products groups or other higher-level entities. For example, a business may want to optimize for profit in one area, enhance revenue in another, enter a new market in a third, etc. Even combined strategies are possible, which optimize to 80% profit and 20% revenue, for example. Once a pricing strategy is defined, the AI-based system can measure the price elasticity from the sales patterns observed in the market and optimize the prices of individual articles at specific points-of-sales automatically.
- Automation of Business Processes: Today, AI-based replenishment and pricing allow retailers to automate a significant part of their operational decisions. Experience shows that up to 99% of all operational decisions in these areas can be automated, which allows human experts to focus on the crucial decisions that require intervention or policy changes. Rather than reducing the staff count, experts are now given the time to think about the best course of actions in highly irregular circumstances and are no longer “drowned” by the sheer amount of day-to-day work. Amazon recently demonstrated with Amazon Go that not only “back office” tasks can be automated — at least in principle — but operational steps that involve the customer directly. The experimental store does not have any check-outs, instead, AI-based systems monitor customers in store and keep track of what they buy or put back on the shelf. Although the single current store demonstrates the potential, the technology is still not quite ready for such a setup. According to a report by The Verge, the system can only handle 20 customers at the time, which is not sufficient for any but the smallest of convenience stores.