Treasury Wine Estates' Innovative AI-based Marketing Approach

Big companies are already using AI to get a competitive advantage. Felicity Carter finds out about a new project from Treasury Wine Estates.

Reading time: 7m 30s

The Wine Event site
The Wine Event site

To its detractors, AI is going to upend life on Earth. To its fans, it’s going to upend marketing ― in a good way.

Justin Noland, the Senior Director, DTC Marketing & Ecommerce at Treasury Wine Estates in California, is definitely a fan ― especially after he recently co-created a website using AI.

Launched in March, is far from being a static, brochure website. It’s a deals machine, where overflow stock, end vintages and other wines are offered for sale. Everything sells out quickly, so the site needs to be dynamic and constantly refreshed.

"We want to adopt tools that can make our jobs better, rather than getting run over by them.”

Building such a website is normally a huge task, needing coders, designers, writers and photographers. And there was plenty of work the AI couldn’t do. But there was enough done by AI that the whole website functions as a sign of things to come.

“One thing I preach to my team is that we want to adopt tools that can make our jobs better, rather than getting run over by them,” says Justin Noland.

From digital healthcare to wine

Noland started his professional life on the agency side, before moving into healthcare.

“I got a really interesting sense of marketing across the board, but primarily focused on how we engage people in a digital way and build mechanisms that drive digital engagement and new audiences,” he says. “And quite frankly, if you can do it in health, you can do it in anything,” because there are so many trust and privacy issues.

Working in health had another effect on him, too. Where many people who read about diagnoses day after day become hypochondriacs, he says he went in the opposite direction. 

“When the number one reason for attrition in your membership is death, it’s draining. I took the time out and said, ‘what do I want to do? Where do I really want to go?’”

Noland says he comes from northern California and had grown up around wine. “The thing about wine is that most people associate wine with great memories and conversations they’ve had, with celebrations, restaurants and meals.”

That was very appealing, but so was something else: “It was ripe for digital disruption.”

Back to the wine industry

In 2018, Noland joined Wente Wine Estates and immersed himself in wine marketing;  Wente’s activities extend from tastings, clubs and e-commerce, to restaurants and even a golf course. “They gave me an incredible opportunity to support marketing across their entire DTC business, so I really got a sense of why people come to wine country.”

From there, he moved to working with friends and understanding venture capital, while exploring digital tactics. He knew that he wanted to stay in wine, but he also wanted to work for an innovative company. “That’s really what drew me to Treasury,” he says.”They value innovation.”


Treasury Wine Estates (TWE), headquartered in Melbourne, is one of the world's largest wine companies: they have a portfolio of more than 50 wineries, from classics like Beringer, all the way to the blockbuster 19 Crimes, with its groundbreaking augmented reality labels.

Using AI isn’t new for TWE ― in 2018 the company signed a contract with artificial intelligence software company Complexica, telling the market that they would be deploying simulator AI to answer “what if” questions related to everything from sales territory mapping to logistical journey plans. The recent emergence of AI products like Chat-GPT has opened even more possibilities.

“The idea really came from my colleague Shem Swerkes, the brains behind the DTC e-commerce and digital that we have,” says Noland. “He’s one of these first adopter kinds of guys.”


Robert Joseph thinks about thinking.

Reading time: 3m 20s

Simplifying marketing

Noland says that TWE’s DTC business constantly generates content, from web copy to emails and newsletters.

“You get into that creative block where it’s: ‘what am I going to write today? I’ve got 15 more emails that we need to create topics on!’” Now, he might ask Chat-GPT to supply 10 topics that a winery might talk to a consumer about during June, and it will quickly provide a list. “Not all of them are going to be very good, but it’s just like you asked your best friend about ‘10 things I could put in an email’, and that can help.”

"AI is particularly helpful for marketers who know their brands so well, they have trouble seeing them in a new light, or in new situations."

Once an idea or two has been chosen, Chat-GPT can expand on them; Noland says AI is particularly helpful for marketers who know their brands so well, they have trouble seeing them in a new light, or in new situations.

Plus it augments the team’s ability. “Even at Treasury, our direct consumer team runs relatively lean, and we like it that way. We like highly talented, motivated individuals that can be really thoughtful, but that creates bandwidth concerns. We need to do something, but do we really have the bandwidth?”

Project Real: a new website for deals 

Then there was the idea for the deals website. “There are a lot of wine shoppers out there that are looking for deals and we know we don’t want to do heavy promotions on our branded sites,” he says. “So we thought, let’s create this other website that maybe we can push some of these deals onto.”

A deals website is normally a labour-intensive project, as sold-out products must disappear instantly, while the latest offers have to be added, with pictures, tasting notes and content. Emails and newsletters then need to announce them. It’s a lot of work. In February this year, TWE began creating the new site, which they dubbed Project Real.

The only way to get something usable from AI is to feed in very clear, very logical, very precise, step-by-step instructions ― like creating a recipe for a beginner cook.

Noland and Swerkes began developing the prompts. The site was built on Shopify, but the custom coding came from AI. was used for bottle shots and background images. Then came the content, partially generated by Chat-GPT and Bard. Compliance and tax tools came from Bloom.

Training the AI

One problem that many users have run into is that Chat-GPT is basically a predictive text system, meaning that it produces streams of words that are patterned after what it’s been trained on. If it doesn’t have the information it needs, it will ‘hallucinate’, or simply make things up. People who have relied on its integrity have made spectacular and often very public blunders.

Noland said that TWE was lucky because it already had plenty of material to train the AI on, from product pages to existing marketing materials. He adds that he was surprised at how creative the result could be. “We all get to a point where we use the same adjectives over and over again, but it can break out all of these things in a different way. It actually feels like a fresh take.”

“We all get to a point where we use the same adjectives over and over again, but AI can break out all of these things in a different way."

Which is great, but what does it mean for the coders, designers and writers inside TWE? Noland says there are plenty of things the AI systems can’t do yet.

“Everybody still sits in the same place as they were sitting before,” he says. “I just think they have new tools available to them to help create bandwidth.”

Project Real, now the, launched at the end of March.

Getting people to use it

It’s one thing to create a website, but unless people know it’s there, they won’t use it. And when there was no email list of people dying to start bargain buying. Noland says TWE used the same marketing techniques it does for all its websites: paid Facebook and Instagram promotions, and search engine marketing. “We start off with a light approach to digital marketing to get people using the sites, to see what problems the site might have, and where the user flow could be optimised.”

ChatGpt robot learning (Photo: Azar/, generated with AI)
ChatGpt robot learning (Photo: Azar/, generated with AI)

Learning the art of prompting

The way to unlock AI is through prompting, which takes time to learn. Noland’s advice is to play around with Chat-GPT in ordinary life. “Go into ChatGPT and say: here are all the things that my family really likes to eat.” Then tell it to list all the ingredients you need to buy in a single shopping expedition, and create all the recipes for the entire week. “Because you’re the expert in your own life, you can see where it has flaws.”

Once someone has used it in their personal life, it becomes much easier to see where it might fit in their business life. Noland gives the example of a consumer survey, where the business might have received thousands of answers. “Those are the kinds of things you can feed into something like ChatGPT and say, ‘what do my consumers care about? What do they really want out of our emails?’”

It’s just a tool for getting things done, not a replacement for human creativity.

The future of AI

Noland doesn’t want to predict where AI is going―”I’m not a prognosticator!”―but he says to remember that it’s just a tool for getting things done, not a replacement for human creativity. “One thing I preach to my team is that we want to adopt tools that can make our jobs better, rather than getting run over by them.”

He’s genial and interesting to talk to, and his enthusiasm for AI is palpable, extending to a touching faith that it will improve customer service. “I think that AI is going to get increasingly better at being able to answer more complex questions to the point where it will shrink the number of customer service requests.”

Noland thinks the days of the customer service doom loop, where the customer is shunted to a chatbot that can’t help, and then directed to a call centre that doesn’t answer, are coming to a close. “The volume of overall queries that need an actual human to solve will go down, so that when somebody does need an actual person, it’s much faster to get to that person.”

A new hope for customer service

Optimistic, indeed, considering that customer service has been steadily degraded ever since the invention of the telephone tree and the outsourced call centre, but here’s hoping.

Regardless, Noland believes that consumers now expect top-notch service online, and that wine brands must cater to that. He also thinks it’s not enough for wineries to tell stories about themselves any more. Instead, they have to understand customers and their needs: “understanding who those customers are, what they care about, what they’re looking for and how we play a part in their story”.

He adds that working with AI has an unexpected benefit ― learning the art of prompting has made him better about giving instructions, including to his children. “I’m thinking about how I give them directions on stuff. How do I think about it in a different way, so that I’m actually making sure that they’re understanding?”

The results? Yes, they understand more clearly. They don’t always respond the way he’d like, but that’s because they’re not AI. In that way, at least, it’s clear that some things never change.


Most wineries make every mistake when they sell their wine. With a new generation of consumers now quickly moving on to spirits and cocktails, the world of wine better solve the problem soon, or it will be left far behind. Paul Wagner wrote a book about it.

Reading time: 4m 50s



Latest Articles