Prime Day is upon us again. The fifth installment of Amazon’s annual sales extravaganza nearly rivals that of Black Friday and the Yuletide shopping season, with last year’s sales totaling up to $4.19 billion, nearly a two-fold increase from the previous year.

Statistic: Global Amazon Prime Day sales from 2015 to 2018 (in billion U.S. dollars) | Statista
Amazon Prime Day sales as tracked by Statista

As eager Prime members crack their knuckles and Amazon workers get ready for their day, there’s another major player behind the scenes: artificial intelligence. Here’s how AI plays an important role in helping everyone from merchants and fulfillment centers to shoppers during this annual exercise in consumer spending.

How AI helps Amazon

Among voice assistants, Alexa has become a household name like Siri, but it’s just one of Amazon’s most high profile AI products. In fact, the massive company thrives on cutting-edge AI applications that are unknown to the average consumer.

First, those recommended products you see at the bottom? Those don’t just come from fellow shoppers — they’re actually a complex mix of recommendations from other shoppers and machine learning. Special algorithms predict not just related items, but products that are most likely to interest the user based on their search history and customer background. In fact, it’s estimated that as much as 35% of Amazon’s total sales come from product recommendations.

Then there’s Amazon Go, the brand’s push into brick & mortar selling with a twist. Instead of using a traditional cashier for checkout, shoppers just can just grab their needed items and go. Each store uses computer vision to track shelf inventory, deep learning for customer purchases, and automated billing so shoppers are billed on their Amazon account.

Back on the virtual storefront, fake reviews are an ongoing concern. According to Fakespot, which tracks the reliability of online reviews, the number of deceptive ratings rose 34% year-on-year. To combat the rising trend, Amazon uses AI designed for detecting review fraud. The system has been in place since 2015 and uses a mix of human moderators and machine learning to identify the hallmarks of “review gaming”. It then automatically flags suspect reviews and even coordinates with social media sites to report and shut down the merchants.


How AI drives logistics

At present, Amazon has 110 warehouses, 50 delivery stations, and 45 sorting centers across the country. Manning these locations are 125,000 full-time workers, augmented by an army of AI-infused robot helpers.

The most distinctive of these are the 100,000 bright orange “Roombas” that can carry entire shelves. When a human worker needs a specific product, the squat foot-tall robot brings along the entire shelf, using QR codes printed on the floor to guide themselves to the requesting worker’s station.


Then there are large industrial robots that stack pallets, move huge containers and store product inventories in a “human-free” zone. When a human worker such as a maintenance tech needs to access the area, they are outfitted in a RF vest to clear robots from their path.

However, when it comes to actually packing individual products into a cardboard box or shelf, human hands are still needed. According to Amazon’s director of robotics fulfillment, fully automated packing is still at least a decade away.

Another emerging application is the Prime Air delivery drone. The electric quadcopters are fully autonomous and can carry packages up to 2.2kg for 15 miles within 30 minutes. The drones are meant to augment or even beat Amazon’s famous one-day delivery, are are slated to start operating “within months” based on the company’s re:Mars conference last June.

The latest version of the Prime Air drone unveiled in Las Vegas last month


How AI helps shoppers

Before Alexa, Prime used to be the exclusive domain of monitors and phone screens. But thanks to millions of Alexa-enabled gadgets and appliances, shopping can now be done by voice.

In fact, to encourage voice-driven sales, the retail giant launched Voice Shopping last year, where exclusive discounts were only available via Alexa. For 2019, Amazon upped the voice ante by offering a three-hour shopping window even before Prime Day officially kicked in. To get early access, Prime shoppers with an Alexa-enabled device simply had to ask “Alexa, where are my deals?”

While this made shopping incredibly easy for consumers, the whole setup relied on an intricate marriage of natural language processing and machine learning to accurately log purchases and provide a seamless experience.

And finally, there are AI-driven apps solely designed to predict Prime Day deals. One such app, Pricepulse, uses AI-generated mathematical models to predict future prices. The app then recommends the best time to purchase based on historical Amazon data, and advises users whether to buy now or wait.

Leave a Reply