AI and Wine Marketing – A Current Assessment

By Josh Kaiser

Exploring AI's Role in Wine Marketing

In today's world, artificial intelligence (AI) is no longer a distant concept that’s associated only with science fiction; it has seamlessly become a part of our everyday lives, whether we realize it or not. From voice-activated assistants like Alexa and Siri to personalized recommendations on Netflix and Spotify or Grammarly helping edit this blog for grammatical errors, AI is everywhere, doing its best to make our lives more convenient and tailored to our preferences while providing moments of unintentional comedy as we realize the limits of the technology. Marketing is one area that has seen a boom in the use of AI recently, with marketers looking to leverage the tool to streamline their processes and make their content more personalized. With its rise in popularity, I’ve been studying the many uses of AI in business during my time at Cal Lutheran University as well as my previous internship before Tablas Creek. When I arrived at Tablas for my summer marketing internship, Ian (our Director of Marketing) suggested we dive into AI’s potential uses at a winery like Tablas Creek to better understand its capabilities, figure out its shortcomings as well as strengths, and decide if we think AI might be a viable tool in wine marketing.

We thought this info might be useful or interesting to our readers as well.

Image generated by Gemini AI using prompt: "Please tell me what AI software looks like in an image."

State of AI Today

Artificial Intelligence is a field of software engineering concerned with building computers and machines that can reason, learn, and act in a way similar to how a human can, using its access to a vast information database. AI has seen a major rise in popularity over the last several years, and the many uses for AI software have fueled a tech boom over the past few years. From risk assessment in finance, to streaming service algorithms and online shopping recommendations, to more personal uses like designing diets and workout plans, there are thousands of AI programs for just about any need you can think of.

Uses for AI in Marketing

Based on my previous experience with AI systems, I was aware of how easily one can get overwhelmed by the number of options there are out there. To stay focused we took into account all the potential uses we were interested in, and narrowed it down to three main categories of features that seem to be in the broadest use:

• Generating written content such as social media captions, website information, emails, or blog posts. We found AI can generate copy of varying length for all these different channels quickly and easily. Whether you need a catchy Instagram caption or multiple email drafts, you simply instruct the AI software on what you want (for instance, an Instagram caption detailing different Rhone grape varieties) and poof! In seconds, the software scans its vast database and uses its knowledge about Instagram post captions and grapes from the Rhone Valley to generate text that is formatted like a caption and should fit the parameters set. That’s the plus side. There are two negatives, neither insignificant. First, there are still real issues with the reliability of the information generated, especially when working with more obscure topics like unusual grape varieties or topics where information in one part of the world may or may not have relevance in another, such as viticulture or weather phenomena. Quite often, the information generated by the AI was factually incorrect or interpreted in a way that wasn’t quite right. That means that every result has to be checked and cannot be used directly from the software. The second negative was that the language was often full of adjectives and adverbs: flowery and sometimes overdone. There’s a classic bit of writing advice that is “show, don’t tell”. The AI output always skewed toward the “tell” side.
• Generating photos. There are now programs that allow you to input any text and it will generate an image of what you input. While this was pretty impressive, the accuracy was certainly not perfect, and the systems often generate random additions to the photo on top of what you asked for. Think of the photos that circulate online where upon closer investigation people have extra fingers, or the lighting is inconsistent from element to element. One use we were hoping for was to find AI software that would allow us to edit text on an image while keeping the same font – think updating the vintage on a bottle in a lifestyle shot -- but we never got an image that we found convincing. AI still struggles with adding or editing text on images.
• We found that many marketers are using AI to assist in brainstorming ideas for all sorts of different content. A big part of marketing is trying to stay on top of trends and come up with fresh ideas to engage with your audience, and AI should theoretically be a great help in both of those. Instead of the marketing team scrolling through hours of social media posts and websites to try and discover new trends, AI can generate trending topics, hashtags, headlines, keywords, and even titles of trending music for things like Instagram Reels in mere seconds. This can be useful, although a critical aspect is ensuring the system has real time data. Some AI engines do not have up to date databases, for instance before the update in September of 2023 ChatGPT only had data up to 2021, two years behind. Meaning the “trending” ideas it generated would be what was trending in 2021, not current day. Real time data is becoming more and more common, so you probably don’t need to worry too much, but it’s definitely something to be aware of.

Concerns with AI

As with any new technology there are a few concerns that anyone experimenting with AI should be aware of.

• Probably the biggest issue with AI as of now is the information that the programs generate is not yet reliable enough to trust. It’s still necessary to carefully read and make edits to all the content an AI engine puts out. This is exacerbated in an age where misinformation on the Internet is all too common.
• Many systems also do not have data that is current and up to date. While this might not matter for rephrasing text or brainstorming ideas, it can be an important drawback when you’re looking to share information on new research or topics on which there is relatively little data available online.
• The content generated can sound robotic and not flow well, or just have a tone that feels inappropriate to your business. There are now things that can be done to train the AI in a tone or voice to try and avoid this (by giving the software content from a website, social captions, blog posts, etc. you can “teach” the AI your tone and how you write), but it is difficult to fully teach the system a voice, and there will often still be times that it does not come out the way you envisioned.
• Photo generation and editing is still lagging. Sure, AI can turn a text prompt into a photo quickly, but the accuracy was lacking. When generating photos there would always be a few details off, or a few unwanted additions. We hoped to find a program that easily edits text on images, like changing the vintage or the percentages of the grapes in on a wine label in a photograph. This could save needing to retaking photos each bottling. However, we found AI struggles greatly when trying to edit small details of text on images, and in general the photo sector of AI has a long way to go (the picture below illustrates one example of an attempt to change the vintage, as well as blend percentages on an Esprit de Tablas. The system tried to match the font and the size, but as you can see it was not very successful). In fact, strangely-rendered text is one of the surest tells on an AI-generated image at the moment. That said, now that apps like Photoshop and Illustrator are adding AI capabilities, I have no doubt the photo sector of AI will advance rapidly.

Bottleshot adjustment attempted using Canva AI

Pricing

In AI, as in much of life, you get what you pay for. Although the free versions of AI software like Chat GPT get a lot of the press, we found that they are often quite limited. These tools can handle simple tasks such as rewording sentences or generating synonyms quickly. There are also paid platforms such as JasperAI (starting at $40/month per account) or HypotenuseAI (starting at $56/month) which offer more advanced features including generating images or creating automated email workflows. In general, the paid platforms offer:

• A free version or free trial with minimal capabilities and limited trial periods aimed at enticing users to upgrade
• A basic tier, ranging from $10 to $50 per month, suitable for individuals or small teams interested in a few basic tools, with additional capabilities
• A premium tier around $60 to $80 per month, offering more accounts, more features, and increased personalization.
• Some platforms also provide more flexible “business” plans, that advertise “custom” or “flexible” pricing. These are targeted towards larger teams who are ready to heavily invest in AI, and likely offer more robust and personalized capabilities. But these were expensive enough that we didn’t try them out.

One challenge (or opportunity, depending on your perspective) is that at this stage of AI’s life there are a dizzying range of engines to choose from. Do enough digging that you find a software that fits what your needs may be. Then plan to put in the time to train that software to your liking. And expect that in order to access the latest technology, data, and optimal user interface, you’ll likely have to go beyond a free engine or the free introductory version of the platform you choose.

Key Takeaways

AI's ability to streamline tasks and personalize experiences is undeniable, however, as our exploration has shown, there are also notable challenges and limitations, particularly regarding the accuracy and reliability of AI-generated content.

Despite these hurdles, the potential of AI is vast, and its continuous evolution promises even greater capabilities in the future. For now, it's important to approach AI with a balanced perspective, leveraging its strengths while remaining vigilant about its shortcomings. Ultimately, AI is a tool that, when used thoughtfully and judiciously, is already capable of enhancing marketing efforts and providing value to our, or your, audience. But thankfully (as a marketing professional) it seems like it’s going to be a while before it can step in and do our job for us, particularly at a place like Tablas Creek where we value the human touch and authenticity of real experience. If we can use AI to enhance bringing that experience to you, we will. But the journey with AI is just beginning, and we look forward to navigating it together!

Artificial Intern image created by DeepAI using prompt: "Make the person in the image look like a robot"