Mastering the Language of AI 🤖

🧭 Decoding the Mystery: What is Prompt Engineering?

I think we can all agree that the term “Prompt Engineering” comes off as a little pompous. Calling it “engineering” is quite a stretch - it’s a straightforward process once you learn the basics. At its heart, prompt engineering is a framework for crafting precise, strategic instructions for AI systems. There is a reason for the name to sound so officious: we need to understand that talking to AI systems is fundamentally different from how we’ve communicated with computer systems in the past. To understand that, we first need to take a look at how these AI systems work.

🤖 The Digital Brain: How AI ‘Thinks’

First, “Artificial Intelligence” is a really broad concept. What we’re really talking about here are the Generative Pre-Trained Transformers, or GPTs. You probably recognize many of these, like OpenAI’s Chat-GPT, Anthropic’s Claude, and Google’s Gemini. These are the models we can talk to in plain language, and that can respond to us in the same (or different if you’re translating!) plain language.

A robot head, cutaway, to see more robots in the brain area

That’s the important distinction: plain language. GPT models work best with natural language: full sentences, grammar, and context. At an extremely high level, these models are pre-trained on immense set of data, and they work by predicting the most likely next words based on the preceding words based on the context of the conversation. Think of a sentence starting with “This candy is….” The end of the sentence depends on context! Coming from a child, it might be “…tasty!”, but from someone with diabetes it might be “…sugar-free!” Providing that context is how we can get the best results.

🌈 Transforming Instructions into Intelligence: Our AI Interaction Toolkit

This is what we mean by prompt engineering. We need to ensure that our AI models have all of the context they need to give us the best replies. And you can already see how this is going to be completely different from how you’re used to doing internet keyword searches. In Google, we’ve been trained to over the years to remove everything except the most important key words from our search. If that search term had more than four words, there’s no telling what sort of results we’re getting!

So if we’re creating a marketing plan for our B2B software-as-a-service startup, we’d probably use a simple keyword search like marketing strategies. Maybe add one or two more words for our industry, like B2B SaaS marketing strategies if the first search was too broad. Then we’d start sifting through the responses to see what fits.

It’s the exact opposite with GPT models. We want to provide a good context for the model to understand what we’re looking for. So our prompt here might be something like I'm launching a B2B SaaS startup targeting mid-size manufacturing companies. Create a targeted marketing strategy that emphasizes our AI-driven predictive maintenance solution. Focus on pain points around equipment downtime and maintenance costs. Write in a professional but slightly conversational tone that speaks directly to operations managers.

This sets context in the first sentence, gives the prompt in the second sentence, adds clarifying instructions in the third sentence and directs the tone and target of the output in the fourth. That’s the “prompt engineering” framework in a nutshell - crafting intelligent prompts that gives the model everything it needs to give you the best results.

🎨 The Prompt Engineering Palette: Strategies & Examples

A colorful palette of robots

Prompt engineering isn’t a one-size-fits-all approach. Different scenarios call for different prompt strategies:

🎯 Precision Prompting: Crystal Clear Communication

Ever tried asking a friend for help and gotten a completely different result than you expected? That’s exactly what happens with AI when you’re not crystal clear. The magic of prompt engineering is all about providing context, details, and specific instructions.

  1. Vague: Explain quantum computing
  2. Specific: Explain quantum computing in a way a high school student would understand, using analogies from everyday life

📘 Learning by Example: The Power of Guided Instruction

Sometimes, showing is better than telling - and that’s exactly what one-shot and few-shot prompting are all about. By providing examples, you’re essentially teaching the AI your preferred style, format, and tone. It’s like training a new team member by showing them exactly how you want a task completed.

Create a customer persona template. Here's an example structure:`
   - Name
   - Professional Role
   - Key Challenges
   - Motivational Drivers

Generate a project timeline using this format:
   - Phase Name
   - Key Objectives
   - Estimated Duration
   - Critical Milestones

🧩 Breaking Down Complexity: Mastering Step-by-Step Strategies

Complex problems can feel overwhelming, whether you’re tackling them yourself or with an AI. The step-by-step approach breaks down intricate tasks into manageable chunks. It’s like solving a puzzle - you don’t try to fit the whole thing at once, you work on one section at a time.

  1. Solving a business strategy challenge:
    • Identify current market position
    • Analyze competitive landscape
    • Develop potential growth strategies
    • Evaluate risk and potential ROI
  2. Developing a new product concept:
    • Research market needs
    • Brainstorm initial concept
    • Create preliminary design
    • Assess technical feasibility

🎭 Shifting Perspectives: The Art of AI Role-Playing

Imagine having a conversation with an expert who can instantly switch perspectives. That’s the power of role-playing prompts. By asking the AI to assume a specific persona, you can get insights from different viewpoints, almost like conducting a mini focus group in your own workspace.

  1. Provide feedback on a business strategy from the perspective of a seasoned industry consultant
  2. Analyze a technical challenge as if you were a senior engineer with 20 years of experience
  3. Evaluate a marketing approach from the viewpoint of a potential customer

🔄 If At First You Don’t Succeed: Iterating your Prompt Techniques

Prompt engineering is an art of refinement. Your first attempt is rarely perfect, and that’s okay! Each iteration is a learning opportunity, helping you understand how to communicate more effectively with AI. It’s like learning a new language - you get better with practice.

  1. First prompt: Help me understand market trends
  2. Refined prompt: Analyze current market trends in the B2B technology sector
  3. Further refined: Provide a comprehensive analysis of emerging technology trends in the B2B sector, focusing on potential disruptive innovations

🤝 Your AI Coaching Companion: Collaborative Problem Solving

Here’s a meta-prompting trick that might blow your mind: sometimes, the best way to get a great response is to ask the AI how to ask a great question. It’s like having a communication coach who can help you articulate your needs more precisely.

  1. Help me craft the most effective prompt to explore a complex business challenge
  2. What additional context would help you provide a more nuanced and accurate response?

🛠️ Prompt Engineering for Small Business Owners

🧰 Crafting Your AI Communication Toolkit

Think of AI like a new team member who needs to understand your business inside and out. The same way that you’d onboard a new employee with a comprehensive briefing, you need to create a communication toolkit that helps AI understand your unique business voice and approach.

Many entrepreneurs make the mistake of treating AI like a generic tool. But the real magic happens when you develop foundational documents that capture the essence of your brand. This is the ‘context’ that AI needs to generate work that’s on-brand, and should be included as part of every prompt you submit.

Your AI communication toolkit should include:

  • A crystal-clear Brand Positioning statement, including your company’s mission, vision, and values.

  • Detailed Brand Messaging guidelines, including your key messaging pillars, an elevator pitch, your tagline, and key words/phrases.

  • A comprehensive Brand Voice description where you describe the tone, style, and structure of your voice across different mediums (ad copy, articles, youtube videos, etc)

🔬 Crafting Reusable Prompt Frameworks That Speak Your Language

Imagine having a personal assistant who knows exactly how to draft your emails, create social media content, and develop marketing strategies - without you having to explain everything from scratch every single time. That’s the power of custom prompt templates.

The key is to create reusable frameworks that capture your unique approach. These aren’t fill-in-the-blank templates - they’re sophisticated guides that help AI understand the nuanced way you approach different business tasks.

Essentially, any time you find yourself trying to remember how you worded this prompt last time - that’s when you need to save a custom prompt. There’s plenty of things that you’ll want to use AI for repeatedly: editing blog posts, analyzing product reviews, helping with social media posts and marketing personas. And definitely make sure to include your branding documentation discussed above!

A robot reading a book titled Brand Voice

✨ Practical AI Integration: Tools and Workflows

I use Copilot for Obsidian for most of my day-to-day work; you can read more about how I chose this tool [here]. I can easily try the same prompt with different GPT models, I can adjust the model ’temperature’ and the amount of included context, and it even has a mode where you can query all of the notes in your vault.

I like to run Copilot in the right sidebar beside the note I’m working on. The interface is very straightforward - type your prompt into the box at the bottom, and your conversation will scroll up above with the AI model’s responses like a messaging app. You can insert a reference to the open note in the prompt by typing ‘{activeNote}’, and you can reference any other note by typing ‘[[’ which will bring up a list of all notes you can choose from. You can easily copy the AI responses to the clipboard or insert them at your cursor in the open note.

🧠 The AI Memory Map: Navigating Context Windows

Before we go any further, let’s take a minute to talk about context windows. Think of a context window like a conversation’s short-term memory. Most AI models can only “remember” a certain amount in a single conversation. When you reach that limit, the AI model forgets the earlier parts of the conversation. So its responses become less accurate or coherent. The context window can vary significantly across different models. Google’s models have a 1 million token context window, while Open AI’s are 128k. One english word is roughly 1.3 tokens, so that’s still a big window! Keep in mind that any documents you link to will also weigh into the limit.

A long document scroll with the lines disappearing at the end

So make sure not to carry a conversation on for too long, or link to too many documents. As soon as you’ve finished what you’re working on, start a new chat and make sure to reference your branding documents, or any other documents pertinent to the new topic.

🛠️ Personalized Prompting: Your Custom AI Toolkit

In Copilot for Obsidian I have the “Add Custom Prompt” command assigned to a hotkey for easy access. I have my branding documents saved in a folder in my vault, and most of my custom prompts start something like this:

You are an expert at digital content marketing. I have a business called Interrupt 13 which provides education and tools for professionals that are interested in starting their own businesses. Here are my Brand Positioning and Brand Messaging documents with my business's core identity: {Brand Folder}

And yes, you can link directly to documents or entire folders from the custom prompts. Or have it refer to the currently active note. Then, it’s easy to access the custom prompts from the chat window by simply typing the forward slash ‘/’ and picking the custom prompt you want.

🚀 Elevate Your Prompt Engineering Skills with Interrupt 13

Your journey into AI mastery doesn’t end here. At Interrupt 13, we’re committed to transforming tech professionals into innovative entrepreneurs. Are you ready to turn your expertise into a strategic advantage?

Next Steps:

The most successful entrepreneurs won’t be those with the most advanced AI tools, but those who know how to communicate with them most effectively.

Communicate. Innovate. Elevate. Your AI Journey Starts Here. ✨