Microsoft and Google Cloud recently made significant announcements (see here and here) on new generative AI (Gen AI) capabilities for retailers ahead of the annual industry event NRF. The moves come as no surprise; according to a Google Cloud survey, 81% of retail decision makers feel a sense of urgency to adopt Gen AI and 72% are ready to deploy it this year.

What could the retail industry look like with Gen AI in 2024? How could the new solutions from Microsoft and Google Cloud help? And where may retailers face challenges?

Why retail makes a good use case for Gen AI

Retailers operate in a competitive environment with significant operational complexity and continuously evolving consumer preferences. They also collect a lot of data, from customer to natural language data, which is used alongside domain-specific syntax and nomenclature to describe products. This makes retailers great candidates for AI and other automation techniques. However, the large volume of unstructured text data also makes it difficult to automate and personalize many processes. There are so many combinations of inputs and outputs that traditional AI approaches and rules-based systems can be unwieldy.

Gen AI’s strength lies in matching user needs with information, particularly unstructured text data. The combination of large language models (LLMs) and vector databases (vDBs) is creating new opportunities for automation and personalization. The reason is simple: LLMs are better at understanding what users want. Combined with a vDB, the solutions are superior at understanding existing information and providing a synthesized response that fulfills a user’s request, whether it’s a retail employee or a customer. Gen AI helps match users with information, a process that usually causes friction in consumer purchase journeys and in many enterprise processes.

The bottom line is this: The retail industry could be one of the biggest beneficiaries of Gen AI technology, and so it comes as no surprise that Google and Microsoft are leading the transition with their in-depth technology and industry expertise.

Where Google and Microsoft come in

Google is naturally providing an LLM-backed product search solution for retailers, which should significantly improve matching requests to product information. It also has new tools to augment customer service interactions and update product catalogs. These are the types of everyday solutions that will help match employees with information faster and more accurately.

The company’s new consumer-facing chatbot for retailers’ websites and apps is particularly interesting (Google refers to these as virtual agents). Chatbots and product searches can help drive sales by connecting shoppers with purchase evaluation information. If these don’t work reliably, it is easy to lose that moment of engagement.

The same is true of Microsoft’s new Copilot template for retailers, which uses the power of LLMs from OpenAI to match shoppers with the right product information. Through our own user experience testing with more than 15,000 consumers across Gen AI chatbots, Applause found that users are delighted when the information is accurate, fast, and enables a more natural conversation to take place. Microsoft also has internal use cases for retail employees, from marketers to front-line retail associates.

An early adopter of Microsoft’s retail Gen AI solutions is Walmart. The retailer recently announced a new Gen AI search feature on iOS that allows shoppers to search for specific use cases, such as a football watch party. This capability uses Microsoft’s Azure OpenAI Service combined with Walmart’s own data.

The transition is not without challenges

There is no free lunch. Gen AI has proven highly beneficial for a number of use cases, but it also brings several challenges. Among the most significant is the inability to precisely control the output returned to a consumer or employee interaction. Issues range from hallucinations – making up facts – to potentially delivering responses using inappropriate language, or divulging data that should not be shared publicly. This is where the promise of automation can directly undermine a customer relationship.

There are several technical approaches being employed to mitigate these risks. However, the models and retail as a category cover such a broad range of domains that large-scale user testing has risen in importance. This is because humans have turned out to be better than automation for identifying many issues. User testing is employed to identify risks that retailers can develop mitigation strategies for and to validate that these mitigation techniques work properly.

The upside of Gen AI is widely recognized. What we are seeing now is a focus on making it work in everyday scenarios that adds significant value for users. That includes a heightened focus on the user experience combined with new testing approaches to mitigate risk and optimize performance.