This week’s article comes from the transcript of a presentation I gave recently at the Pemberton Chamber of Commerce. If you would like me to present on this subject at your local chamber, industry association, or other group, please don’t hesitate to reach out!
AI is the buzzword du jour. Everyone in small business, particularly marketing, is hearing about it. There’s a lot of hype, a lot of talk about it being the “next big thing,” almost like magic. Let’s be clear: it’s not magic. However, it is a seriously impressive technology, and with a clear strategy, it can be transformational.
The best way I’ve found to explain AI is with a low-tech analogy: think of it as the world’s biggest library. You ask it a question, it races off to the stacks, finds a book, and brings it back. That’s your answer. The catch? Ask a generic question, you get a generic book. Ask about “farming,” and you might get a nursery rhyme with Old McDonald on the cover.
The key is learning to ask really specific questions. Use technical terminology relevant to your field. If you do, the AI will return with a book that feels like it was written just for you.
Choosing Your AI Interface: Free vs. Paid (and What You Risk)
There are plenty of AI interfaces to choose from. ChatGPT led the charge a few years ago. Google Gemini is another major player, and Microsoft offers Copilot. Even IBM is in the game with Watson X. They all operate similarly, but key differences exist, mainly concerning the free versus paid versions.
Here’s a hard truth: if you’re not paying for it, you’re the product. The terms and conditions of free versions often state that any information you share can be used however they choose. It doesn’t guarantee they will use your secrets, but the option is there.
ChatGPT, Google Gemini, and Microsoft all offer paid versions with stronger data privacy promises. However, even then, you’re still sharing data with a third party. While they promise security, accidents happen.
If you’re only using AI to tweak marketing copy, the free version might be fine. But if you’re using AI to make important, data-driven decisions, tread carefully. Sharing medical or financial data, even with promises of security, can raise privacy concerns.
Take Control: Open Source AI and Running Models Locally
Fortunately, options exist that allow you to take control of your data. Companies are releasing “open source” AI models, like DeepSeek, that you can download and install directly on your computer. Run it locally, and you eliminate the need for an internet connection and third-party data sharing. In most cases, a high-end laptop can handle the processing.
While the chat interface (where you type your prompts) is like the steering wheel of a car, the real work happens under the hood with models like Claude 3.5 Sonnet, Gemini 2 Flash, and so on.
You don’t need to be a data scientist to get the most out of AI, any more than you need to be a mechanical engineer to drive a car. But understanding the basics of how these models are trained is essential for maximizing their effectiveness. So, that’s where we’ll start.
How AI Really Works: It’s All About Relationships
When we talk, we understand language intuitively. Computers don’t. So, how do they do it?
They’ve ingested the sum total of words on the internet – about 10 trillion! That’s enough to fill a bookcase stretching around the world twice. These computers look for the relationship between words and convert them into statistics, using them to predict the next word in a sentence. We do it instinctively; they do it mathematically.
Consider the sentence: “I’m brewing the tea.” We all understand it. But change one word: “I’m spilling the tea.” Suddenly, it means gossiping! That single word change alters the entire meaning. That’s what the math behind AI seeks to understand: the relationships between words. AI uses massive computing power and statistical analysis to predict what word comes next.
Here’s the kicker: if you ask the system, “What is 1 + 1?” it answers “2,” not because it did the math, but because “2” statistically follows “1 + 1 =” most often on the internet. But it’s not always correct! Somewhere, someone has probably written, “1 + 1 = 5.”
That’s the essence of how it works. Now that you understand that, you need to learn how to “speak the librarian’s language” to get the best results.
R.A.P.P.E.L.: Speak the AI Librarian’s Language
Here’s a model we use at SATM to get fantastic results. Full credit to the data scientists who developed it. It’s called R.A.P.P.E.L.:
- Role: Tell the AI what kind of expert it is.
- Action: Define what you’re going to do today.
- Prime: Ask the AI what it already knows.
- Prompt: Give the AI the specific task.
- Evaluate: Provide feedback and refine the results.
- Learn: Convert the process into reusable instructions.
Let’s break it down:
- Role: Instead of dealing with the head librarian (who knows everything about everything), tell the AI to find the area specialist. For example: “You’re a seasoned marketing expert specializing in hospitality. You understand the fundamentals of marketing including the four Ps, brand development, etc.” This narrows the focus and improves results.
- Action: Define the task: “We’re going to write content for our website. I’ll give you information about our business, our accommodation offering, our USP, etc. Your task is to review this information and write a travel guide-style article about why someone should visit this region, stay with us, and experience the local attractions.”
- Prime: Before you ask the AI to start creating, ask it what it already knows. “Before we begin, what do you know best about best practices for travel writing, specifically for online publications like websites? What would an expert on this subject know that a novice writer would make? What are common mistakes a novice writer would make that we should avoid?”
Remember, AI works best when it “talks it out.” Filling the chat window with relevant jargon improves the results. And you don’t have to do all the heavy lifting; the AI already has a PhD in everything! - Prompt: “Based on everything we’ve discussed, including best practices, expert tips, novice mistakes to avoid, and the information I’ve provided about our business and the local region, write an article for our website.”
- Evaluate: After the AI generates the article, provide feedback. Tell it what you like and what you don’t. Ask it to focus on certain aspects and downplay others. Then, ask for a final version. “Now that we’ve edited and improved our article, give me the final version that is ready to be published.”
- Learn: If this is a task you’ll repeat, ask the AI to pre-write instructions for itself, saving you time in the future. “Take our final process in our conversation so far and convert it with prompt engineering into a system instructions for a generative AI system. The instructions should include our writing style, information about a business, local region, write the system instructions in second person imperative.”
This creates a set of reusable instructions you can copy and paste next time.
Prompt Limits: The AI Has a Near-Infinite Memory
You might wonder how far you can push the AI. The prompt limits are surprisingly high. Most models can handle at least 90,000 words (the equivalent of a standard business book). Google Gemini can handle roughly 1.5 million words – the entirety of War and Peace, the Bible, and the complete works of Shakespeare simultaneously!
For the average person writing website content or making business decisions, the AI’s short-term memory is effectively limitless.
Build Your Prompt Library
Create a system for storing pre-written prompts and background information. Whether it’s Joplin, Google Keep, Microsoft OneNote, or a simple Google Doc, the goal is to avoid reinventing the wheel.
AI Use Cases: It’s More Than Just Content Generation
Some say there are six AI use cases; I combine extraction and summarization to make five:
- Generation
- Extraction and Summarization
- Rewriting
- Classification
- Question Answering
Believe it or not, generation (writing blog posts, emails) is the one AI is least competent at. However, it improves dramatically when paired with other use cases.
AI in Action: Making the Most of Each Use Case
- Generation: The basics – write a blog post, write an email. As mentioned, it’s the most overrated use case. Using AI only for content generation is like using a Harrier jump jet to collect milk – overkill.
- Extraction and Summarization: Essential for sifting through overwhelming amounts of information. Imagine you’re a winery in Margaret River or a tour operator in the Great Southern region. You’ve seen reports from Tourism Australia about the future of global tourism, but who has the time to go through all 200 pages? Or consider this: You have a CRM full of data, but it can be hard to use for business decisions.
- Plug it into your AI system of choice (I recommend Notebook LM from Google).
- Ask it to summarize the information.
- Ask it to pull out the bits relevant to your specific business.
- You can do the same with competitor websites, social media, and online reviews. Extract the data, summarize it, and look for trends. What are people saying about your competitors? What are they not saying about you? AI helps you synthesize that data and determine your next move.
- Rewriting: More than just improving professionalism, rewriting can help tailor your message to specific audiences.
- Create an ideal customer profile using your CRM data.
- Have the AI review the content on your website and answer the question, “Would this ideal customer be interested in this? What would their thoughts be?”
- Rewrite the content with that person in mind.
- You can also use rewriting to translate your content for international audiences. Find out where your visitors are coming from. Are you getting a lot of Argentinians? Create an ideal customer profile for Argentinian travelers and ask the AI to translate your English content, taking into account cultural nuances.
Remember, 75% of internet users don’t speak English as their first language. This is a massive opportunity that almost no one is exploiting! - Classification: Use AI to classify large amounts of data into useful categories. I had AI review the Pemberton Chamber of Commerce website from a user experience perspective. It came back with strengths, weaknesses, and areas for improvement.
Also, download and transcribe YouTube videos about your area and plug them into the AI. It can then perform a general sentiment analysis. What are people saying and feeling when they travel to your region? - Question Answering: Use AI to brainstorm and get feedback. Use your CRM data to create a synthetic version of your ideal customer and then ask them questions about your products and services.
Limitations and Risks: Garbage In, Garbage Out (and Monkey Selfies!)
Because AI systems are trained on the public internet, they’re exposed to the good, the bad, and the ugly. The internet is full of biases. That’s why it’s crucial to review AI-generated content carefully and ensure you’re comfortable with the claims being made.
Also, remember the “monkey selfie” case? A British wildlife photographer set up a camera in the jungle, and some monkeys took some photos. The US Copyright Office ruled that works created by non-humans (like a monkey) are not copyrightable.
AI is also non-human. Therefore, if AI does all the work for you, you can’t copyright it. It’s essential to have your hand in the creation process.
Succeeding in the AI Era: Be an Amplifier of Good
AI, to paraphrase Captain America, is an amplifier. It makes good things better and bad things worse. To succeed in a world where AI enables more people to create more content, you need:
- The Most Best Audience: Be a mini media empire. Focus on quality content that connects with your audience on an emotional level.
- The Most Best Data: Learn everything you can about your customers and your industry.
- The Most Best Ideas: AI is an amplifier. It allows one person to achieve what used to take a team. You need to stay ahead of the curve with new and innovative ideas.
Ready to Unlock the Power of AI for Your Tourism or Hospitality Business?
AI is a game-changer, but it’s not a magic bullet. To truly harness its power, you need a strategic approach and a deep understanding of your audience and your business.
That’s where SATM comes in. We can help you develop a customized AI strategy that aligns with your goals and drives real results.
Contact us today for a free consultation and let’s explore how AI can transform your business!
