Few terms are met with as much confusion and angst as artificial intelligence or AI. What is AI? Is it real? Will I lose my job to a robot or an intelligent algorithm? I hear these questions a lot these days.
Generally speaking, AI is a computer model that mimics human characteristics of reasoning, discovery, and learning from past experiences. There are many types or variations of AI. Machine Learning is built on the concept of providing data to self-learning algorithms to allow them to “learn”. Deep Leaning takes the Machine Learning concept and narrows it focus to develop neural networks that mimic human thought. Cognitive computing (think IBM Watson) is another subset of AI that is built on the concept of natural human interactions. Whereas an AI algorithm may make a decision, in a cognitive model, the decision will be shared, naturally, with a human. Think of Watson’s voice from its appearance on Jeopardy.
Contrary to many beliefs, AI is not a new phenomenon. Dedicated research into AI began in the late 1950’s and has gone through various starts and stops before finally gaining a firm hold in the technology landscape in the early 2000’s.
Today, AI represents an $8-10B industry with new applications appearing on a regular basis. This growth has been driven by a few factors. The first is the availability of computing power. AI algorithms require massive computational cycles. Early efforts failed as the processors simply couldn’t handle the workload. Today, armed with the most powerful computer processors in history and the availability of compute cycles in a variety of cloud environments, engineers and scientists are seldom limited by compute power.
The second major factor is the availability of data. Typically AI “learns” through trial and error. Data that is both broad and deep (many different examples of various data scenarios) is required for algorithms to self-optimize. Today disk storage prices are at historic lows and input/output speeds are faster than ever. This new infrastructure is leveraged by technologies capable of efficiently organizing and managing massive quantities of data of varying types. This “big data” ecosystem is critical to the success of AI efforts.
The third major factor is significant growth in the sophistication and availability of machine learning algorithms at the core of any AI solution. Advances in programming languages have made it quicker and easier for scientists and engineers to create and test new, sophisticated models. This is evidenced by the availability of AI functions on all of the major cloud platforms including Amazon AWS, Microsoft Azure, IBM (Watson), and Google.
As AI continues to mature, practical application is found in all industries. However, few industries are as ready as retail to reap the benefits of AI. Retail data sets up nicely for a variety of AI use cases. Some of these use cases include:
- Product Assortment
- Product Recommendations
- Customer marketing and promotions
- Inventory allocation and replenishment
- Physical store layout
- Fraud Detection
Let’s touch on each one of these in a bit more detail.
Product Assortment
Retailers begin and end with the customer in mind. To that point, the products a retailer makes available for sale are designed and/or selected to target a specific set of customers. Merchants have to possess a unique ability to understand the wants and needs of the customer and ensure their product line captivates the customer’s interest. Today’s merchants rely on advanced analytics to help them understand customer behaviors, tastes, and trends. As the customer is changing more rapidly than ever, it is a daunting task for merchants to anticipate and build a winning product assortment each season.
Artificial Intelligence is being used to quickly analyze shopping, trend, and behavioral data to provide guidance to merchants on the appropriate products for their assortments. In addition, since AI can deal with data at a very fine level of granularity, it provides the capability to truly localize assortment choices for specific markets.
Product Recommendations
We have all been to an online retailer website, selected a product, and then watched as the retailer displayed a number of related products. These recommendations are driven from previous purchases and analytics that determine most common product associations.
In the future, AI will take this a step further. By combining past purchase behavior with other data points including promotions accepted and declined, previous web searches, products removed from cart, etc. the AI algorithms can infer customer intent. This provides the basis for building much more engaging, and personalized, product recommendations. In fact, this creates the foundation for building digital assistants to help customers in the online world in a similar fashion that store associates help them in the physical world.
Customer marketing and promotions
Are you tired of receiving generic coupons and BOGO offers from retailers? Wouldn’t it be better if the retailer really understood your interests and provided offers that were of interest to you?
One of the most common use cases referenced for AI in retail is the creation of very accurate, personalized, and timely offers to customers. Machine learning algorithms can sift through tons of customer purchase history, web searches, social media posts, etc. to understand the products most interesting to a specific customer. Some online retailers are already using these AI-generated personalized offers to great success. Further, pairing this technology with Internet of Things (IoT) and a mobile app, allows the retailer to personalized real-time promotions to customers at the point of purchase inside a physical store.
Inventory allocation and replenishment
Retailers have struggled for years with inventory management. Having the right product at the right place at the right time is fundamental to retail success. However, there are many moving parts in the replenishment process that have to be totally in sync in order to get the right amount of product to the right location to satisfy customer demand.
AI algorithms are being developed today that can optimize replenishment decisions and significantly improve retailer’s inventory productivity. The AI algorithms go beyond the basics of considering POS Data, promotions, and tuning parameters (min/max, weeks of supply, etc.). AI can take the basic information, add demographics about a location, information about products, weather forecasts, holiday shopping patterns, shipping patterns, etc. to build highly accurate and finely grained replenishment decisions.
Physical store layout
We have all visited stores where we just can’t seem to find the product we want. We look for signage or a nearby associate to help direct us to the products we seek. Once we have our product in-hand we suddenly think of a related item. Hence, we go through the same process again.
Intelligent store layout begins with understanding the product in the store, the natural product adjacencies, and how customers routinely shop. Using video analytics and AI, it is possible to evaluate thousands of scenarios to arrive at the optimal placement of fixtures and products to simplify the shopping process for customers.
Fraud Detection
I recall talking to our Loss Prevention team years ago about the use of analytics to help detect fraud and send automated alerts to the appropriate LP team members. They were so excited to get some tools to help them detect potential fraud issues in our stores as much of their casework to that point in time had been highly manual.
Today, AI algorithms can be constructed to detect potential theft issues in stores and online based on patterns in sales and credit card data. In addition, leveraging video analytics and AI, physical store theft patterns can be identified and appropriate store personnel alerted before product leaves the store. This particular use case isn’t full mature yet, but it is garnering a lot of attention.
We have answered the question, “What is AI” and have described real world use cases in the retail industry. Will you lose you job? While you may not lose your job, if your job is built on fact-based decision-making, it is likely to change. As retail moves from the world of batch processing and reports to real-time analytics and decisions, automation will play a key role. AI will play a key, enabling role that allows retailers to automate transactional decision-making and improve sales and productivity.
2018 will be a pivotal year for AI in retail. Retailers are already searching for solutions. Many start-ups and established technology companies are delivering their first AI products in 2018. I look for a number of success stories to emerge throughout the year as adoption increases. If you are a retailer and you aren’t envisioning a future with AI, you need to get on board. The train is leaving the station and your competitors are likely already in their seats.