Getting the best out of AI in Retail

AI in action

AI can help to provide a more interactive customer experience, encourage repeat business and help stock availability and speed to market. Here are a few examples:

• Smart mirrors allow customers to try on clothes without the need to undress by overlapping clothes in a clever way that allows shoppers to ‘try on’ multiple items at once. Intel’s “MemoryMirror” even allows a customer to review what they have previously tried on via their smart phone or tablet. Ralph Lauren, Kohl and Rebecca Minkoff are experimenting with this technology.
• Stock management systems are less sexy than in-store gadgets, but fundamental to getting the right stock into the right store, or delivered to the customer, at the right time. Data analytic tools that predict customer behaviour can assist retailers by better anticipating the needs and behaviour of their customers. This in turn leads to more accurate inventory ordering, stock management and carrier scheduling.
• IBM’s Watson is the AI behind 1-800-Flowers.com launch of a gift concierge system that acts as a personal shopper. It asks the shopper a series of questions to identify the perfect gift. The more you use it, the smarter it gets.
• The North Face is providing customers with more accurate product recommendations by using a search tool that is also powered by IBM’s Watson. The technology focuses on natural language use to produce more accurate results.

Sounds great, but what is the catch?

Upfront capex/opex: implementing new software and/or hardware, particularly across multiple stores or factories, is not necessarily cheap or straightforward. Consider your 3-5 year aims and the potential shelf life of the technology you are considering. Can you accurately measure the ROI of your investment?

Understand your products and your market: with so many AI options, it can be easy to be overwhelmed. As with any technology, it is important to invest in the most suitable technology and to calculate your ROI in advance. For example, if you sell lower priced or essential items, it is perhaps less likely that investing in better gift recommendation technology will result in a material sales increase. Once you know what you want and have identified a suitable provider, it is important to undertake a pilot or evaluation. Ideally, seek references from other business that have used the services.

Key legal considerations

In addition to choice and cost, there are (of course!) some important legal issues to be addressed.

Tech contracts - the devil really is in the detail: navigating your way around new technology contracts can be challenging. Many AI solutions are SaaS solutions. Items to watch out for typically include: service availability (and understanding maintenance/support windows); server location issues; security issues; data charges (e.g. ensuring you have enough data usage headroom within any pricing structure); charging “add-ons” for extras; termination rights; data ownership and data portability/migration support on exit.

Data security and other data processing obligations: AI learns by processing large data sets and it is hungry: it requires feeding to keep evolving. But the more ‘personal data’ you collect, particularly ‘sensitive personal data’, the more likely you are to become a target for hackers, face greater sanctions and regulatory fines in the event of a security breach and failure to comply with privacy laws and/or consumer complaints. This will be a heightened risk once the General Data Protection Regulation (GDPR) becomes law in 2018.

Collecting new data or using ‘dark data’ (i.e. data collected but not previously utilised) may be a great goal, but do you have a system for interpreting and understanding the data? In the case of older data, does your processing of that data comply with data protection requirements such as having legal grounds for the processing? Is your purpose for processing different to the purpose for which the data was collected? How old is the data and should you have already deleted it? There may be solutions open to you such as anonymising the data before using it, though in a world of improved connectivity and intelligent learning, it can be difficult to truly anonymise data. Netflix and Amazon learned this the hard way when they released, what they thought was, anonymous data to the public, only to discover that it was possible to identify individuals within the data sets released.

The law on data protection will become more onerous and wide reaching in 2018. The GDPR permits the use of collected data for research (with appropriate safeguards) which may be useful for certain types of AI learning. The GDPR also specifically requires that appropriate safeguards are put in place for profiling, and AI is likely to be caught. Good data practices will allow you to maximise data opportunities but without damaging your reputation or annoying your customers! It may even enhance the value of your business as large and ‘useable’ data sets become increasingly important.

There are, as with any new initiative, issues to be considered and challenges to be addressed but the good news is that, according to digital visionary Kevin Kelly, no one is an AI expert yet, which means no one in the retail sector is too late to join the party.