AI for Retail: Benefits and Use Cases

AI for retail is a concept that can be easily understood superficially. After all, artificial intelligence is pretty much an absolute necessity nowadays, whether its simple inventory management or complex customer service automation.

However, many companies and business establishments have not yet leveraged this powerful AI technology tool. In fact, fully realizing its potential in commerce is still a cultural and traditional struggle. Corporate structures are usually still locked within the standard operative know-how of managerial institutions of the late 20th century.

In this article, we look at the importance of AI for retail. We will briefly discuss how its prominence came to be, and hopefully deliver several cases that will allow readers to both understand its utmost importance, and provide a somewhat better idea of how retail businesses can benefit from AI.

Why AI is Important for Retail

Before we begin addressing the question of what exactly AI does for business around the world, we need to elaborate first on the conceptually simple, but consequently complex importance of AI for retail.

Anyone who has read or watched the news about current software systems used in business applications already has an idea of its significance. After all, the consumer tech market is currently built from the foundations of commerce, accounting, and finance. And what do commerce, accounting, and finance entail in business? That’s right, customers and consumers — the “end-user” for every product and service provided.

Thus, the basic importance of AI for retail is optimizing end-user data. What is considered end-user data? Anything that has to do with the target customer of your retail business. End-user data directly answers questions such as:

• What is the general trend of your purchased products?
• What are the specific choices of your individual consumers?
• How satisfied do your customers feel when using your services?
• How exactly do they use your services and/or products?
• What type of interaction do they have when speaking with representatives — whether human or automated?

Of course, other elements such as logistics, resources, and manpower are important variables in implementing AI for retail. But end-user data is the most important because it will directly determine the staying power of your business institution.

Evolution of AI in Retail

The history of AI in general, while usually associated with scientific research achievements, is actually also intricately intertwined with business and commerce.

After ENIAC and UNIVAC — two of the first computers developed — there was a push to develop computers and software in the 1970s that could be mass-produced. The aim was to introduce affordable versions of previously very expensive hardware to regular business employees in order to streamline routine computational elements of their daily jobs. However, applications were very limited due to how largely insufficient hardware specifications were at that time.

Several decades later during the 1990s, two very important trends in AI development revolutionized retail applications. First was the focus on AI that can actively perform a wide array of human decisions on its own, thanks to leaps in computer hardware improvement. This was in contrast to earlier AI systems that would simply relay data to its users, or have a very rudimentary human-like decision-making capacity. And second, the first practical application of the once theoretical neural network.

These combined concepts allowed AI to finally take the form and shape with which we are familiar today. Though not as advanced, the objectives and tasks given to retail AI at that time were fundamentally similar, such as automated planning and scheduling, inventory sorting, product distribution, customer service relay, among many others.

As the first decades of the 21st century came and went, the ever-improving internet finally became an integral part of most data used to train retail AI. Consumer information from around the world became much easier to access and was often updated live. End-user data are now largely stored in cloud-based databases, allowing for adjustment and adaptation at every possible additional step in business operations.

Today, there is a massive push towards the integration of natural language to any retail AI system. This is not only to allow intuitive interaction with customers using such automated systems but also for use by managerial staff.

For example, Google Duplex is an AI system introduced in 2018. It was developed to use natural conversations to conduct “real-world tasks”, with a convincingly human-sounding customer service representative. The uncanny delivery of its responses, as well as its method of understanding the user, made it possible to seamlessly integrate it in a way that could potentially go unnoticed by a consumer.

Use Cases for AI in Retail

There have been a lot of notable applications of AI in retail throughout the years. But as for use cases that are focused on end-user data, we have:

1. Taco Bell Develops the Taco Bot

Taco Bell has earned the reputation of being the very first fast-food chain to allow customers to order food via its natural language AI, Taco Bot. Unlike current voice recognition systems however, this service is accessed through the communication platform Slack, where you directly type in orders from the chat menu.

The challenge was to develop an automated ordering system that is internet-based, does not require VoIP, and does not need long forms to be submitted before confirming an order. To this end, its developers at Deutsch set out to build a basic chatbot AI, but with the partial flexibility of Siri or Google Assistant when it comes to response and learning. That way, it could still converse in a basic manner and relay suggestions, while still being able to take orders with a wide writing format recognition margin.

The implementation was practically an overnight success. Not only did it achieve all its design objectives with flying colors, but customers also appreciated how the Taco Bot offers seemingly witty quips. Not the Turing test-breaking kind of course, but entertaining enough to be amusing.

2. Zipline Drone Delivery Zooms Through Logistics

Many drone delivery systems have been conceptualized and tested over the last decade. However, only Zipline currently has the most efficient, fastest and perhaps the most operation critical drone delivery system currently available on the planet.

Instead of simply giving AI the task of automatically processing orders, Zipline tackled the logistical problem of creating a reliable fleet of drones that could deliver medical supplies cheaply. First, is the development of an interface that would automatically determine destinations and drop points simply by pointing and clicking on a map. Second, is the navigation computer of the drone itself, which is connected to a database at its central headquarters.

Operations start from the reception of the order via an online pharmacy. The items are then packaged as quickly as possible, and the drone is launched. After dropping the package via parachute, the drone finally makes its way straight back to base.

As of 2019, Rwanda’s Zipline medical delivery system has made more than 13,000 deliveries, flying more than 1 million km total. It is by far the most successful drone delivery service as of late, and other investors are now looking into the system to consider increased funding.

3. Uniqlo Scans Brains to Know Your Choices via UMood Kiosks

Perhaps one of the most sophisticated, largely sensor-based applications with AI for retail, Japan-based Uniqlo employs UMood kiosks in its shops to provide product suggestions based on your mood, or “how you feel” about each item available at the store.

As a company with businesses centered on choice, streamlining user options is Uniqlo’s primary objective with this concept. This is mainly achieved by measuring five different variables — interest, like, concentration, stress, and drowsiness — via a special neurotransmitter headset. The system’s algorithm observes and analyses these variables by having the user watch a series of videos. The system then chooses a set of suggested clothing from the category selected.

Because the UMood kiosks were implemented more as an experiment than an actual system developed to improve customer service, there is not much to say about its commercial efficiency. However, it did presumably prompt the company to delve further into its potential, as evidenced by future projects of similar nature such as the Uniqlo IQ, a “digital concierge” service utilizing machine learning that shares personalized style recommendations from the Uniqlo’s collection.

“A lot of the value that we’re getting from machine learning is actually happening, you know, kind of beneath the surface, and it is things like improved search results, improved product recommendations for customers, improved forecasting for inventory management, and literally hundreds of other things.”

Amazon CEO Jeff Bezos spoke these words at a 10X interview in 2017. This was part of his response when asked about Amazon’s approach to AI. He stressed that even though there are many “superficial and obvious” applications on AI for retail, the underlying mechanisms of algorithms that tweak, adjust, and optimize are the ones that do the real job.

In other words, should you ever decide to integrate AI systems on your commercial ventures, it is highly recommended to do so at the most basic, fundamental level: starting of course, with your primary end-user base.

Open Call for Writers

Do you like writing about tech topics like this one? Then join the Kambria Content Challenge and share your insight and expertise with our growing developer community. You could receive over $200 for the best submission. For complete details about our Content Challenge, click here.

Originally published at Kambria.