Generate sales & marketing assets to promote your brand with AI
In April I gave a presentation at the KBI conference on A.I. This promised to be a
…demonstration of generative AI (artificial intelligence) tools, including ChatGPT (text generation), DALLE-2, and Midjourney (images), to highlight the ways kombucha brands can use these new technologies—at no cost–in sales and marketing campaigns.
Generative AI has greater potential than superficial first impressions suggest. It can provide insights into customer needs, preferences, and behaviors and create targeted content that drives sales. Additionally, AI-powered systems can help brands create compelling content for their campaigns by analyzing customer reviews, feedback, and industry trends.
We will demonstrate how ChatGPT can build headlines, keywords, page outlines, FAQs, and blog topics for a website. We will also look at ways to generate images and offer suggestions and guidelines from industry experts.
The full version of this presentation with audio recording and all the other KKON23 presentations is now available for purchase from KBI (who own the rights). Use the code SUMMER100 before July 31, 2023 for a $100 discount!
Here’s a sample of the material from my presentation.
Generative AI has attracted a lot of press since ChatGPT was introduced in November 2022. While “AI” is all over the news, it’s nothing new. If you’ve used spell check in the past twenty or more years you’ve used an AI program. What *is* new is *generative* AI. Programs like ChatGPT and Midjourney.
Undoubtedly it will have an impact on many professionals and white collar workers. However, if you ferment kombucha for a living your job is safe! There’s no chance an AI program will take over the steeping, fermenting, bottling and distribution of your beverage. But, as small businesses, you can leverage generative AI to help create compelling content for advertising and marketing campaigns by analyzing customer reviews, feedback, and industry trends. Think of it as a new smart assistant, available for free.
The poll of the 20 or so people in the room showed only a small number had used ChatGPT at that point (six months after its release).
The initial model was trained with information from the internet including news articles, books, websites and more. But, to make the responses more human-like and conversational, human AI trainers also added data, such as providing conversation examples from both a user and an AI perspective.You can geek out on the technology behind this form of AI here. Just be aware of the limitations, as ChatGPT itself shares (in response to the prompt on the slide):
ChatGPT, like any other AI language model, has several limitations. Here are some of the key limitations:
- Lack of real-world understanding: ChatGPT lacks true understanding of the world. It doesn’t possess personal experiences, emotions, or common sense reasoning. It can generate plausible-sounding responses without having a deep understanding of the topic.
- Sensitivity to input phrasing: ChatGPT is sensitive to how the input is phrased. Slight rephrasing of a question can lead to different responses. This means users may need to experiment with their queries to get the desired results.
- Over-reliance on training data: ChatGPT is trained on a large corpus of text, but this data may contain biases, inaccuracies, or incomplete information. As a result, the model can inadvertently generate incorrect or biased responses.
- Tendency to be verbose or excessively detailed: ChatGPT has a tendency to be excessively verbose, providing more information than necessary. It may also make up details or provide a level of precision that is not justified.
- Difficulty with ambiguous queries: When presented with ambiguous queries or requests for clarification, ChatGPT may guess the user’s intent or provide a general response. It can sometimes struggle to seek clarifications when the query is unclear.
- Lack of multi-turn context: While ChatGPT can handle short conversations, it may not maintain a consistent understanding across multiple turns. It may not remember information mentioned earlier in the conversation, leading to disjointed or repetitive responses.
- Inability to provide real-time information: ChatGPT’s knowledge is based on data available up until September 2021. It cannot provide real-time information or access information from the web as it does not have browsing capabilities.
- Potential for inappropriate or biased responses: ChatGPT learns from the data it is trained on, which may contain biased or offensive content. Consequently, there is a risk of generating inappropriate, biased, or offensive responses.
It’s important to exercise critical thinking when interacting with AI models like ChatGPT and verify the information provided by consulting reliable sources.
Tools like ChatGPT (such as the newly released Claude 2 – available in the USA and UK since early July, and not around when the presentation was given) can assist with a number of sales and marketing related tasks, including:
•Automate repetitive tasks: Producing outputs on a consistent basis with a certain format and quality. Example use cases: producing ad copy, creating product descriptions, writing computer programs to extract phone numbers from text.
•Accelerate writing: Writing down the first draft, or even the final version of a piece of text. Example use cases: composing emails, writing blog posts, providing customer chat responses.
•Brainstorm ideas: Generating a skeleton of a bigger piece to be worked on, instead of working off a blank canvas. Example use cases: generating article outlines, finding business ideas, writing story plots.
•Augment a skill: Augmenting the skill of a writer who might not have sufficient proficiency. Example use cases: writing poems, writing fiction stories, formulating product pitches.
•Condense information: Getting a summarized version of a document that strips it to its essence. Example use cases: summarizing reports, articles, and podcast transcripts.
•Simplify the complex: Rewriting a piece of text into a simpler, more accessible way. Example use cases: simplifying technical explanations, understanding complex text, extracting key concepts from a passage.
•Expand perspectives: Adding variety to the voice and idea beyond just the person writing. Example use cases: generating opinions in essays, constructing arguments in debating, adding variety in speech scripts.
•Improve what’s available: Turning a piece of text into a better version. Example use cases: making a passage more coherent, rewriting podcast transcripts.
This was demonstrated in the KKON presentation by ChatGPT’s response to the prompt on the slide:
- Enhanced vitality: The living cells and energy in our kombucha can provide a natural energy boost, helping you feel more vibrant and alert throughout the day.
- Improved digestion: The presence of living cells in our kombucha can support a healthy gut microbiome, aiding in digestion and potentially reducing digestive discomfort.
- Nutritional value: Our locally brewed, organic kombucha is packed with nutrients, providing a rich source of vitamins, minerals, and antioxidants that can contribute to overall well-being.
- Refreshing and invigorating: The effervescence and lively nature of our kombucha can offer a refreshing and invigorating experience, providing a sense of revitalization and enjoyment.
- Natural mood booster: The consumption of kombucha can contribute to a positive mood due to its potential effects on gut health and overall well-being, promoting a sense of happiness and contentment.
- Stress relief: Enjoying a glass of our kombucha can serve as a calming ritual, helping you unwind and alleviate stress after a long day.
- Conversational starter: Offering our unique, locally brewed kombucha can spark conversations and create a shared experience among friends, family, or colleagues, enhancing social interactions.
- Health-conscious choice: By choosing our organic kombucha, you demonstrate a commitment to your health and the environment, potentially inspiring others to make similar health-conscious choices.
- Community support: By opting for locally brewed kombucha, you contribute to supporting local businesses and the community, fostering a sense of connection and empowerment.
Note: The benefits listed are general and may vary depending on individual experiences and circumstances.
Prompting is at the heart of working with LLMs. The prompt provides a context for the text that we want the model to generate. The prompts we create can be anything from simple instructions to more complex pieces of text, and they are used to encourage the model to produce a specific type of output.
Coming up with a good prompt is both a science and an art. On the one hand, we know the broad patterns that enable us to construct a prompt that will generate the output that we want. But on the other hand, there is so much room for creativity and imagination in coming up with nuances in prompts that can get the best out of a model.
Again, the demonstration took the prompt (asking about prompts!) on the slide to reveal what ChatGPT advised:
The most effective form of prompt for ChatGPT depends on the specific task or goal you have in mind. Here are a few approaches that can help improve the effectiveness of prompts:
- Clear and specific questions: Asking direct questions with clear objectives tends to yield better results. Instead of asking, “Tell me about the weather,” you can ask, “What will be the weather forecast for tomorrow in New York City?”
- Providing context: Giving relevant context before asking a question can help ChatGPT understand the specific information you’re seeking. For example, if you’re asking about a specific movie, you can provide details like the title, director, and genre.
- Instructional prompts: Instead of asking an open-ended question, you can provide specific instructions to guide the model’s response. For example, you can ask ChatGPT to list the steps for a recipe, provide pros and cons for a particular topic, or explain a concept in simple terms.
- Systematic exploration: If you’re unsure about the model’s knowledge or want to explore its capabilities, you can start with a broad question and then refine or narrow down the focus based on the initial response. This iterative process can help you gradually get the desired information.
- Multiple-turn conversations: Instead of using a single prompt, you can engage in a back-and-forth conversation with ChatGPT. Reference previous responses to maintain context and clarify any misunderstandings. This can help build a more coherent and detailed conversation.
Remember that while these approaches can improve the effectiveness of prompts, ChatGPT still has limitations and may not always generate accurate or optimal responses. It’s important to evaluate the information critically and verify it from reliable sources when needed.
This live demonastration caused an audible gasp in the room. Be aware the responses from ChatGPT are almost instantaneous. No doubt, given time, an MBA student or savvy small business owner could come up with similar information. But why wait? Here’s the response from ChatGPT:
Certainly! Here’s a Business Model Canvas for a kombucha start-up:
- Customer Segments:
- Health-conscious individuals seeking natural and organic beverages.
- Individuals interested in gut health and probiotic-rich products.
- Fitness enthusiasts looking for a refreshing and energizing drink.
- Value Propositions:
- Locally brewed, organic kombucha that promotes vitality and digestive wellness.
- Wide range of flavors and varieties to cater to diverse tastes.
- High-quality, living probiotic cultures for a beneficial gut microbiome.
- Direct sales through a dedicated e-commerce website.
- Local partnerships with health food stores, gyms, and yoga studios.
- Participation in farmers’ markets and local food festivals.
- Customer Relationships:
- Personalized customer support through email, chat, or phone to address queries and concerns.
- Engage with customers through social media platforms and maintain an active online presence.
- Gather feedback and suggestions through surveys and customer reviews.
- Revenue Streams:
- Sale of kombucha bottles and multipacks through various channels.
- Subscription models for recurring sales and delivery services.
- Collaborations with cafes and restaurants for bulk orders.
- Key Resources:
- Brewing facility with fermentation equipment and quality control measures.
- Access to organic ingredients and living probiotic cultures.
- Skilled brewmasters and staff knowledgeable about kombucha production.
- Key Activities:
- Brewing and fermenting kombucha batches.
- Flavor experimentation and product development.
- Quality testing, packaging, and labeling.
- Marketing and brand promotion activities.
- Distribution and logistics management.
- Key Partners:
- Organic ingredient suppliers.
- Packaging and label suppliers.
- Local retailers and distribution partners.
- Health and wellness influencers or bloggers for collaborations.
- Cost Structure:
- Raw material costs (organic ingredients, packaging materials).
- Production and brewing expenses.
- Marketing and advertising costs.
- Staff salaries and operational costs.
- Distribution and logistics expenses.
- Regulatory and compliance costs.
Please note that this is a general outline, and specific details may vary based on the unique circumstances and strategies of the kombucha start-up.
Understand you don’t have to accept all this, but are free to cherry pick whatever is most useful. Again, use the tool as a smart virtual assistant available 7×24 at no charge. And, as cautioned above, “it can generate plausible-sounding responses without having a deep understanding of the topic.” So use common sense when evaluating the output.
There’s much more to my KKON presentation, with examples of other killer prompts you can use to help grow your business. Plus a look at using generative AI to create marketing graphics, as Wildflower Kombucha did for their logo.
Purchase the conference recordings on the KBI website.
This sample of my presentation is posted with the express permission of KBI.