How ChatGPT and generative AI will revolutionise grocery

ChatGPT, please write an article on ChatGPT and its impact on the grocery sector in the style of The Grocer.

The wheel turns as it builds its answer.

“Artificial intelligence has been making waves in the grocery sector, and the latest innovation to hit the scene is ChatGPT. This advanced language model, developed by OpenAI, is designed to generate human-like responses to a wide range of queries and prompts. With its ability to process large amounts of data, understand natural language, and provide personalised responses, ChatGPT is poised to have a major impact on the grocery sector.”

It’s not bad, you’d have to say. Grammatically sound. Factually accurate. Clear and concise. But a little dry. Can you try again, but a little more ‘exciting’ this time?

The wheel turns. “Hold on tight because the future of grocery shopping just got a whole lot smarter with the revolutionary new technology called ChatGPT! With its cutting-edge artificial intelligence and human-like response abilities, this advanced language model is turning the grocery industry on its head.”

OK, too much. For now, at least, this writer’s job is safe. But what about elsewhere across the industry? While ChatGPT isn’t yet “turning the grocery industry on its head”, it is definitely turning heads. Several major food companies from Coca-Cola to Shopify have already announced trials or partnerships, while many more are testing it as part of wider trials for this new wave of AI models.

“We’re not too far from a world where everything from ready meals to new crisp flavours are discovered through AI”

Diarmuid Gill, chief technology officer at Criteo

ChatGPT is an example of generative AI, a type of artificial intelligence that generates new and original content, using techniques like neural networks and deep learning. Other types of AI tend to focus on classifying or labelling data. Given a bunch of images of baked beans, it would find patterns and use them to find more images of what it reckoned were also baked beans.

Generative AI is concerned with creating new data based on simple prompts that are similar to existing data. This means it can make a completely fresh and unique image of beans in response to a natural language prompt like ‘a photo of some baked beans in a red bowl in the style of a Picasso painting’.

The outputs might be 2D images like ChatGPT’s sister model Dall-E 2, conversational text like ChatGPT itself, or video, computer code, game backgrounds, even music. Or indeed a ‘multimodal’ combination of all of them.

The underlying elements of the technology have only come together in recent years. And the launch of ChatGPT – its most famous form – has been a viral smash, gaining 100 million users within two months of launch, making it the fastest-growing consumer application in history, according to a UBS study. By comparison, TikTok took nine months to hit the same landmark while Instagram took two-and-a-half years.

OpenAI’s success has boosted attention on several other providers, and pressured the likes of Google to rush out its competitor Google Bard.

Grocery players are already piling in for a piece of the action. Some are superficially tapping into the hype: Patrón Tequila in February launched an AI art tool – the Patrón Dream Margarita generator – for users to “create the margarita of their dreams”.

But others are going much deeper. “Coca-Cola’s vision for the adoption of OpenAI’s technology is the most ambitious we have seen of any consumer products company,” said Zack Kass, head of go-to-market at OpenAI. In February, the drinks giant announced it would be the first company to partner with a global services alliance between OpenAI and consultancy Bain & Company.

“There’s so much to potentially go after,” said John Murphy, president and chief financial officer of The Coca-Cola Company.

Ambition and potential are great, but what about specifics? What exactly could grocers and manufacturers use such generative AI for? Who better to ask than ChatGPT itself. “Marketing campaigns, customer service, and product recommendations,” it replies.

Let’s take these one at a time. First up, marketing. According to Murphy at Coca-Cola, there are already “a couple of really interesting cases with the marketing team to enhance the work we’re already doing with our new marketing model, and to be able to marry the ability to deliver creative content at speed and to do it with exponential efficiency”.

What exactly this would look like remains unclear. But in theory, using multimodal models would allow for engagement such as a customer getting served a unique social media ad with AI-generated imagery and voiceover designed specifically for them based on their individual likes and purchase history.

That’s admittedly at the ambitious end, but even conservative predictions see generative AI like ChatCPT soon drafting display ads, social media posts and TV adverts in extra quick time.

“There is space for these in food and drink marketing,” says Brian Barthelt, MD, global martech at TMS. “Anyone who’s had to hand-create a storyboard knows that the details are not the key element. So working with an AI image generator that occasionally gives people 13 fingers is less of an issue. It speeds up a process and frees time for designers to work on other things.

But, he adds, “it’s about efficiency, not about ditching creative jobs for machines”.

Human conversation

ChatGPT said its second use was customer service. And that’s not a prediction either, with Carrefour, Shopify and Instacart already using it to provide a personal shopping assistant service.

This also allows initiatives that discuss product benefits, suggest recipes and promote sustainability credentials in a conversational way, via an avatar picked for that individual customer.

“Imagine shopping with a digital assistant in your supermarket trolley which knows what you want to cook so it can tell you what you need to buy, always bearing in mind your previous preferences and dietary requirements,” says Caroline Day, AI expert at Haseltine Lake Kempner. “AI could be used to guide you through a supermarket, making suggestions in a conversational way which seem almost human.”

ChatGPT’s final possibility was product recommendations, which, given the exhaustive customer data supermarkets hold in their loyalty schemes, could soon enable suggestions at an individual level.

“We will face a moment where AI will integrate social analysis, behavioural analytics, and reasonably identify consumer problems, generate multiple idea combinations against these, refine them based on pre-existing feasibility criteria and probably live test,” says Lee Powney, senior partner at consultancy Vivaldi Group.

But there are possibilities too that ChatGPT didn’t come up with, like new product development. “We’re not too far from a world where everything from ready meals to new crisp flavours are discovered through AI,” says Diarmuid Gill, chief technology officer at online advertising firm Criteo.

Since it launched a few weeks ago, several brewers have turned to ChatGPT to formulate new beers from available hops. Banded Oak Brewing in Denver, Colorado, asked the AI to predict the next craft beer trend. “Wellness Beer” apparently. It even came up with a recipe and a name: The Bots Made Me Do It.

“It might not be perfect, but I think it’s going to be pretty good actually,” says head brewer Chris Kirk.

Diageo has also experimented with using ChatGPT to come up with “unique, flavoursome cocktails”, while drinks app DUSK tapped the AI to come up with ‘the best cocktail in the world’ called Heavenly Sipper.

Given ChatGPT has aced the infamously difficult Court of Master Sommeliers exams, it might well know its stuff.

“AI will play a pivotal role in ideating and creating innovative food and beverage products,” says Eli Wood, design director at Designit. “It can predict the most successful product ideas and flavour combinations by analysing trillions of data points, while ensuring that these products also consider factors such as cultural preferences, nutritional content, and environmental factors.”

It could even “help to alter logistics, inventory and supply chain decisions quicker than most humans” adds Barthelt. “That might not be passing the Turing test, but it can entirely change the way a business runs and increase profits,” he says.

Elon Musk pushes back

Despite its huge potential – or perhaps because of it – there are growing calls for progress to slow, or halt while there’s still opportunity for society to get a handle on generative AI’s power. Last month, Elon Musk and Apple founder Steve Wozniak were among hundreds of prominent tech experts calling for an “immediate pause for at least six months” working on systems more powerful than GPT-4.

The World Economic Forum predicts generative AI’s impact will be “nothing short of seismic”, yet it is happening without required checks, the letter’s authors say.

“Unfortunately, this level of planning and management is not happening, even though recent months have seen AI labs locked in an out-of-control race to develop and deploy ever more powerful digital minds that no one – not even their creators – can understand, predict, or reliably control,” they state.

Concerns are emerging too about how generative AI engines handle data, with doubts over the safety of uploading terrabytes of customer data. “The issue at hand is that a significant proportion of the population lacks a clear understanding of how these AI models function, which can result in the inadvertent submission of private information,” says Richard Forrest, legal director at law firm Hayes Connor.

“What’s more, the interfaces themselves may not necessarily be GDPR-compliant. If company or client data becomes compromised due to its usage, current laws are blurred in terms of which party may be liable.”

Last week Italy moved to ban and investigate OpenAI “with immediate effect” over GDPR fears and in March, it emerged ChatGPT users could access the chat titles of other users without their permission. Something the company said it felt “awful” about.

“It is the onus of businesses to take action to ensure regulations are drawn up within their business, and to educate employees on how AI chatbots integrate and retrieve data,” Forrest adds.

And society could quickly turn on the technology presently best known for interesting experiments and funny viral images. Especially if Goldman Sachs is right when it says the technology could “substitute up to one-fourth of current work” with two-thirds of jobs in Europe and the US exposed to a degree of automation.

But it’s probably too late to put the genie back in the bottle. As Wood says: “We know this is just the beginning of the AI revolution.”

A final word from ChatGPT itself: “While there is certainly a lot of potential for these models to be used in a wide range of applications, it’s important to approach them with a critical eye and consider their limitations as well.” It may be a little dry, but it remains undoubtedly wise.