AI in retail: Technologies and applications
With retailers constantly adopting new strategies to stay ahead, many companies are implementing AI technology into their workflows. For example, we can now see AI in retail stores through retail robots and AI shopping systems. It’s also playing a behind-the-scenes role in retail analytics and warehouse automation. Through a blend of data analysis, machine learning and quality training data, AI systems are helping retailers field customer enquiries, offer better recommendations and improve inventory management.
In this article, we look at a few popular examples of AI in retail together with real-life use cases of the technology at work and tips for implementation. We also look at developing technologies for AI shopping and retail automation that could transform the industry as we know it.
AI shopping: automated checkouts
Automated checkouts enhance the shopping experience by removing the constraints of the traditional checkout experience. Take the Amazon Go convenience store as an example. Customers use a mobile app to enter the store, they then grab what they need and leave. There are no lines and no check out, because the item scanning and payment process are completely automated.
Another notable example of retail automation is clothing giant Uniqlo, whose automated checkouts don’t require staffing. Customers simply place their basket in the scanner, pay for their items, and go. This simplifies the shopping process for customers while freeing up staff to attend to customers on the shop floor.
How it works: Automated checkouts utilize computer vision technology to scan and recognize items either in store or at the checkout area. These items are confirmed against an inventory database, with payment automated through an app, or built into the checkout process. Preparing for an automated checkout system requires a database of store inventory annotated for scanning purposes. The exact annotation of these items depends on where the cameras in your store are located.
Retail robotics is growing in popularity, and the most interesting implementation so far is store assistants. The robot assistant Pepper, for example, can now be found in Softbank stores across Japan. Pepper answers basic customer queries while engaging visitors in novel and fun conversation.
Similarly, home improvement retailer Lowes utilizes the Lowebot, which answers questions in multiple languages, and guides customers to what they’re looking for. Through the use of computer vision, the Lowebot can also monitor inventory to give staff feedback on how products are selling across the store.
How it works: Similar to virtual assistants, retail robots utilize ASR technology to understand customer queries. Their speech recognition systems digitize live speech into machine-readable form, analyze that speech for meaning, and respond based on previous input and pre-programmed algorithms. In order to respond to a wide variety of questions, these robots require large amounts of audio training data.
Search recommender systems
When it comes to AI in eCommerce, the range of available products and services has increased exponentially. As a result, many customers are now faced with information overload. For this reason, search recommender systems act as a guide; they’re the path a customer walks to finding what they want. Youtube, Netflix, and Spotify all utilize recommender systems, but the technology plays an important role in retail, too.
Amazon’s search recommender system is often seen as the gold standard, as their recommendation engine reportedly generates 35% of their revenue. However, retailers and eCommerce sites like Best Buy, Ebay and Etsy also utilize recommender systems to help drive sales. These recommendations take a variety of forms, including:
- On-site recommendations
- Recommending related items
- Showing what similar shoppers have purchased
- Letting users know if a newer version of a product is available
- Offering recommendations based on past purchases
How it works: On a basic level, search recommender systems work from a database of customer types. The system compares users by defined features, and ranks products among groups of users. Products are ranked based on factors such as popularity, price, etc.
When building an automated recommender system, there are three points to consider in terms of data for your system: the first is ensuring quality data collection, the second is ensuring quality data annotation, and the third is validating your search engine for accuracy.
The future of AI in retail
The field of retail AI is full of experimentation, as companies look to improve service and profits by implementing new technology. Here are some examples of automated artificial intelligence we’re likely to see more of in the next decade:
Chatbots: Interactions with chatbots happen through the messaging apps customers are already familiar with. Chatbots can act as shopping assistants for customers in need of specific items, answer queries, and recommend new products or trending items to regular customers.
Visual search: See a t-shirt you like? What if you could take a picture of it, and that picture brought you to a shop where you could buy it? That’s visual search in a nutshell. Visual search uses machine learning algorithms trained on image datasets. Whenever a user takes a photo and uploads it to the visual search engine, the system searches the library for any items that match that photograph along with relevant details like brand name, price, etc.
Voice Search: With the proliferation of virtual assistants, voice search is seeing steady growth. Some sources report that as much as 20% of mobile searches are done this way. Voice search is convenient for people who are on the go, and is a good investment for retailers with an active customer base.
Where to start
As the retail industry develops, automation is quickly becoming a necessity. With this in mind, knowing where and how to invest in technological development is hugely important. However, the key to successful AI implementation starts with the following three steps:
- Define your goals: What problems will your AI automation solve?
- Define your data: What data is necessary? Do you already have it, or will it need to be collected?
- Define your path forward: How will you label the data to ensure accurate training for your machine learning model? Who will build the model? Does it need to work in multiple languages?
If you don’t have the answers to these questions, don’t worry, TELUS International can help. Our team can help analyze your needs to develop an AI-based solution, and our AI Community of 1,000,000+ professionals can help you collect and label your data. Ready to transform your business? Contact us today.