Last month, Johnnie Walker’s parent company Diageo rolled out artificial intelligence (AI) whisky selector ‘What’s Your Whisky’, which analyses drinkers’ flavour preferences to pair them with their perfect single malt Scotch. Here, we take a look at the ground-breaking technology, and consider the ways in which algorithms could revolutionise the drinks industry in years to come…
Think back to a bottle of alcohol you bought without ever having tried it. What compelled you to pick that one, rather than another? Perhaps it was the look of the label, or the price. Maybe a friend recommended it, or you spotted it on this very blog. Or, if you’ve just road-tested Diageo’s new AI whisky selector, it might be because an algorithm told you to.
Named ‘What’s Your Whisky’, the selector uses FlavorPrint taste profiling technology to match your individual tastes to one of 18 featured single malts, explains Benjamin Lickfett, head of technology & innovation at Diageo. It asks eleven questions to understand your preferences – e.g. ‘how often do you eat bananas? How do you feel about chillies?’ – and then analyses your responses.
“To do this, we use an algorithmic machine learning analysis of 500 different flavour points based on data from the food science and expert sensory science sectors,” he continues. “Once individual flavour preferences have been mapped, the app uses AI to continuously learn what drives consumer preferences.”
Elsewhere, AI isn’t just matching you with your optimum booze pairing. It’s creating it. In November, Circumstance Distillery created the world’s first AI gin, called Monker’s Garkel, in collaboration with tech companies Rewrite Digital and Tiny Giant. They designed a ‘recurrent neural network’ named Ginette, explain Liam Hirt, Circumstance co-founder.
“She was trained to compose gin recipes using an enormous data set of botanical and recipes,” Hirt says. “We chose her best two recipes for further traditional development at Circumstance Distillery. One recipe emerged as a favourite, although it was very close. Ginette also came up with the name for the gin. A separate neural network was used to create the label and the wording on the back of the bottle.”
Circumstance isn’t the only producer to harness the power of AI to make great-tasting spirits. In May last year, Swedish distillery Mackmyra teamed up with Microsoft and Fourkind to create a whisky informed by Mackmyra’s existing recipes, sales data and customer preferences. In January 2017, Virgin’s travel arm partnered with super-computer Watson to analyse the social media posts of 15 million holidaymakers, match them to 5,000-plus flavour descriptions and reviews, and create a one-off rum recipe at Barbados’ Foursquare Distillery.
Is there a danger our industry’s tastemakers could soon be overthrown by AI distillers? Not quite. “AI technology is in its infancy, and is not ready to take over from a skilled distiller like those at our distillery,” reckons Hirt. “Where I see AI making a difference in the near future is as a creative muse used during product development. At Circumstance Distillery we do a lot of product development and contract distillation for customers. AI in its current form can be a useful tool at the brainstorming stage to contribute ideas that might be quite different and take development in an unexpected and novel direction.”
In what ways, then, could AI potentially revolutionise the industry as we know it today? For now, the answer lies in behind the scenes operations. French drinks company Pernod Ricard, which owns Jameson whiskey and Beefeater gin, has been “developing a series of successful pilots and then projects at scale for quite a large array of applications” for a few years now, explains global media and content hub leader Thibaut Portal.
This could be something as simple as identifying trending venues using data from Google Maps, Google Venues traffic, Trip Advisor and social media channels, he explains; information that helps the company map and structure its approach to the on-trade. Automated algorithms help the company optimise its social media campaigns, too – by defining and predicting best days and hours of the week to interact with consumers as well as personalising messages and communications.
“We have applied AI mainly so far and at scale for our marketing and sales department activities, as data are massive and easy to collect,” says Portal. “AI technology definitely enables us to react faster and prepare for more informed decisions, leveraging and computing data available internally or sourced externally in a flash. It provides solid analysis capabilities and unlocks new business opportunities: from product launch to market share increases.”
While it’s still early days for Diageo’s customer-facing whisky selector – which launched across nine European countries in six languages – Lickfett says the team is excited about the potential of this untapped tech. “Once we’ve received the initial results, we’ll be looking to optimise how we integrate the AI experience in bars, supermarkets, online and beyond,” he says. “As with any new technology application, it is key to put the consumer at the centre of the experience, ensuring real value is added and to avoid creating technology for technology’s sake.”
He makes a point. With that in mind, are there any challenges the industry might need to overcome to integrate AI technology successfully? The most obvious one, Hirt says, is knowledge. “Circumstance Distillery is very tech-focused, with successful projects such as issuing ‘whisky tokens’ in the form of our own cryptocurrency,” he says. “Most small businesses in the drink sector are not as tech-focused as we are.”
It’s a sentiment backed by Portal. “AI technology has developed so fast with so many suppliers that confusion is already there,” he explains. “It requires expertise, knowledge and capacity to select the right project.” With a little knowledge, however, the sky’s the limit. “There are so many offers on the market, available and easy to access for all,” he says. “We are entering a democratisation phase, as well as a learning curve for all to build.”