The legal risks of using artificial intelligence in brick-and-mortar retail

Register now

The increasing dominance of online shopping and the recent collapse of some of the world’s most notable retail chains has left many wondering—what can retailers do to stem the tide? One answer has been to deploy artificial intelligence/machine learning solutions to enhance efficiencies and improve the in-store consumer experience.

As of 2018, retailer investment in AI/ML is on track to exceed $3.4 billion and is projected to overtake the current investment leader—banking. The scale of this investment is unsurprising given that 83% if retail executives believe it will have a critical impact on their future.

Some of the most notable ways AI/ML has been employed are through in-store personalization and backend inventory management. On a very basic level, AI/ML enables computers to make autonomous decisions. Solutions such as advanced gesture recognition and augmented reality (AR) take this capability and apply it to personalizing a consumer’s in-store experience.

For example, AI/ML can be employed through strategically placed cameras that monitor consumer physical gestures and in-store traffic, recognize consumer habits and preferences, and provide valuable insights in to product popularity so that stores can quickly adapt layouts and inventory requirements.

AR devices, such as virtual mirrors, can similarly enhance the shopping experience by quickly scanning a person’s body type, cross reference it with it consumer preferences, and make continuous recommendations including what accessories best compliment the product. This reduces restocking costs and the guess work in predicting trends, while at the same time increasing secondary sale potential.

On the backend, AI/ML deployed to predict supply and demand can analyze store data, market trends, social media and past performance to make predictive inventory recommendations or purchasing decisions ensuring that a store’s inventory remains ahead of trends. In addition, AI/ML enabled robot deployment can increase supply chain efficiency by restocking shelves, prioritizing shipments from distribution centers, resetting floor plans, and increasing product exposure based on the data derived from sales and consumer facing AI/ML integrations.

Despite the clear benefits of AI/ML deployment, an AI/ML solution is only as good as its data. There is a direct correlation between the increase in datasets and the improvement of AI/ML models. However, there are three main legal issues where retailers need observe caution—data collection, bias and ownership.

The collection of data is subject to a variety of privacy laws (e.g., the GDPR, FTC Act, etc.), which set forth certain disclosure and consent requirements that govern a retailer’s ability to collect and use that data.

One developing area of consumer privacy is biometrics. Currently, three states (Texas, Illinois, and Washington) have passed laws that require, among other things, express consent for the use of biometric data for commercial purposes. These laws have broadly defined biometric data to include fingerprints, facial geometry, retina or iris scans, voiceprints, hand geometry and other unique biological patterns, all of which are used by AR and advanced gesture recognition technology.

Unlike online retailers, which can place access to certain products or features behind a consent wall, brick-and-mortar retailers in these states have the difficult task of obtaining consent from every shopper entering a store in which it will use biometric identifiers for a commercial purpose, such as tracking the shopper and charging the shopper for the items the shopper takes.

Another significant issue is that of AI/ML bias. For example, a model developed and based on data collected from consumers in a Boise, Idaho store may not be as effective when implemented in Miami, Florida. Due to the origin of the dataset, there could be an increased risk that the model’s ability to identify minority groups not previously represented in the dataset is limited.

While unintentional, it has the effect of potentially denying services to a person based upon their race. Because of the world’s vastly different demographics, the creation of datasets requires a careful consideration of the source of the data and retailers should continue to keep the FTC’s guidance on big data in mind when determining the suitability of a dataset.

Lastly, data ownership is a continuing concern as many retailers do not have the resources to develop AI/ML in-house. However, retailers may have competitive concerns regarding confidentiality of datasets, ownership of the model, and the ability to use insights/results from the model/engine.

Because AI/ML models continually improve when introduced to more data, a retailer’s data can be valuable to a vendor. Thus, retailers should carefully consider the type of model, the impact to the retailer if competitors have access to the same model, and whether it is comfortable with granting a vendor a right to use that data for other projects/services for other retailers.

Ecommerce competition and the potential for increased costs are driving brick-and-mortar retailers to AI/ML adoption. If fully realized, retailers can create fully adaptable ecosystems to stay ahead of the trends and drive increased store traffic. However, there are real legal concerns that must be considered before retailers continue to venture into the world of AI/ML.

For reprint and licensing requests for this article, click here.