AI has become a pretty big deal in the retail space, with a wide range of different adoption patterns across the industry. Predictably, the biggest adopters of AI technologies in this space have been very large multinational organisations, largely in the FMCG and fashion retail space, though these technologies have been trialled or adopted in nearly every retail vertical.
With retail and marketing going hand in hand, one of the key use cases is to gather and effectively use marketing data, especially in online retail. Retargeting, improving conversion rates and website personalisation are all popular use cases.
In the in-store environment, AI is being used in a variety of ways, largely in two broad categories; Active and Passive.
In the Active AI field, technologies such as Augmented reality ‘magic mirrors’ are being trialled extensively in fashion retail in order to allow customers to virtually ‘try on’ different items without having to gather all of them together. AI engines behind this can act as recommendation engines, suggesting the items most likely to look great, and add on items to complement the chosen products, increasing the overall customer spend. In high end retail, especially outlet or mall retail, these experiences can be combined with a seamless ‘add to bag’ and payment workflow that allow customers to have their chosen items picked, packed and either waiting for them to collect, or moved to a central location. This can also increase net spend throughout a visit as customers who do not carry around their purchases are likely to spend more.
In the passive space, AI can be used to streamline the overall customer experience, or reduce risk and shrinkage. The combination of AI and video is particularly compelling, with cameras able to identify all manner of activities and infer decisions from this. We have also seen trials of Video analytics in the self-checkout hardware, to automatically detect age and not to challenge for restricted products if the customer is identified as clearly over the challenge threshold. In addition to this, cameras can be combined with barcode data to learn a product set, and then when in inference mode, detect attempts at fraud, either from barcode switching or falsely identifying an item within the system. Notifications can then be silently issued to store staff on their workstations or carryable/wearable devices.
At the very high end of this movement, we are seeing the biggest names in the space making acquisitions in AI following successful pilots. McDonalds recently paid $300m to acquire Dynamic Yield, an Israeli AI startup, in order to build AI platforms that dynamically personalise menu content based on a number of factors. Intelligent triggers are being used in a wide range of markets; BMW for example worked with Driftrock to drive dynamic adverts triggered by weather and placed on social, pushing their xDrive four wheel drive models into locations where the weather was particularly inclement at the time. While this is more rules based than full dynamic AI, the momentum of this intelligent targeting to reduce cost and increase clickthrough is increasing dramatically in the marketing space.
But while there is a lot of potential in using AI for retail, AI is part of a much longer journey. Though the final destination is always a moving target, AI has a number of downstream dependencies that organisations must get right before pursuing the fun stuff. AIs need to be trained, and training them requires the right platform, the right models, and critically, the right data. Without a solid data platform, the utility of any AI will be compromised. Without a holistic approach to data architecture in technology, people and process, AI technologies can only be deployed on tactical, point use cases, rather than across the whole environment.
While not every organisation will follow the example of Amazon and Alibaba in their Go and HEMA concepts, in an environment of unprecedented competition for consumer spending, AI technologies will continue to become more pervasive throughout the retail environment.