There’s little question what the most talked-about technology of the past 18 months has been. ChatGPT, the next-generation AI ‘chatbot’ from OpenAI, represents another significant milestone in artificial intelligence mastering ‘human’ cognitive behaviors. In this case, arguably the most definitively human cognitive behavior of all. Language.

What ChatGPT can do is mighty impressive. Ask the ‘bot’ a question, and it will provide pretty much any text-based answer you want. It will write that essay or report (or blog) you’ve been putting off. It will answer questions directly, rather than leaving you to trawl through the top search results yourself like a search engine does. It will write a resume and a covering letter for your next job application. You can even use it to write code for an app.

Like we say, pretty impressive. And as happens with these things, everyone is falling over themselves to talk about all the wonderful potential use cases in industry X, Y and Z.

But what about retail? Sure, you can see how such a tool might help ecommerce stores speed up writing product descriptions or make the chatbot support AIs on their website provide better answers. But can this technology really be a game-changer in retail?

Large Language Models

To find out what’s most exciting about ChatGPT from a retail perspective, you have to look under the hood. ChatGPT and its ability to write like a human is really just one application of an altogether more impressive piece of tech.

ChatGPT is an example of what is known as generative AI. Generative in the sense that it can ‘generate’ or create something independently. Generative AI works using a particular type of AI algorithm called a large language model, or LLM. ChatGPT operates on the GPT family of LLMs developed by OpenAI, a Silicon Valley-based AI research outfit that Elon Musk was involved in setting up back in the day. It’s by no means the only LLM in town. Google, Microsoft and Meta all have their own competing models.

LLMs work by analysing language. Very, very large quantities of language. It’s been estimated that OpenAI’s latest GPT-4 model was ‘trained’ on something in the region of 1.76 trillion ‘parameters’. A parameter is a variable that an AI algorithm learns the pattern of. To consume 1.76 trillion language variables, you’re talking about an AI that has digested more or less the entire world wide web.

This is what’s really exciting about generative AI. The gargantuan feats of data analysis that LLMs are accomplishing. And not only that, but the types of data they are using. In data science, language is classed as ‘unstructured’ data – that is, it doesn’t have the kind of structured mathematical values that conventional data analysis relies on. And yet LLMs are working out the patterns of the vast cacophony of human language, and pulling out meaningful intelligence from it. At incredible scales.

A new level of customer intelligence

You know how they say understanding your customers is the key to success in retail? Think about how much better your customer insights would be if you could analyse every customer review, every social media post, every live chat conversation, every email, every phone call even. Conversations with customers represent an enormous untapped data resource for most retail businesses. LLMs offer the chance to hit this rich seam.

Imagine an analytics tool where you could, say, feed in details of a proposed new product and get an in-depth analysis of likely customer interest and responses, without ever having to conduct market research. Or even use such insights as part of the product development process.

The same principle could be applied right across your business. On-going sentiment analysis from customer interactions could be used to inform merchandising, promotions, branding, even the content you have on your digital channels or in-store media. Everything optimized in real time, based on what real customers are saying.

Then there’s the potential of making all of these additional insights available directly to the customer in order to improve their experience. One day, consumers could be walking around stores with personal AI shopping assistants, getting personalized recommendations and detailed responses to questions of a kind you usually associate with premium retail establishments that offer clienteling services.

So in summary, the bad news for retailers is that ChatGPT is not some magical genii that will reveal the secrets of retail success and all the untold wealth that goes with it. Although you’re free to ask. But what is truly exciting is the potential for LLM technology to be applied to the field of customer relationship management, and to take customer intelligence to a whole new level. Watch this space.