the art of prompting

To ask the right question is already half the solution of a problem.
— Carl Jung

Today I’ll show you how you can instantly level up your AI prompts with just a few simple tweaks.

This is a key skill to master as AI is becoming more and more popular by the day.

Skillful prompting will:

  • avoid generic-sounding outputs

  • increase productivity

  • greatly aid your creative process

It’s also vital to understand good prompting etiquette when designing automation workflows to get the most out of them.

By the end of this newsletter, you will have a better understanding of prompting than 95% of people using AI.

Most people don’t take the time to learn prompting and as a result they will only get mediocre results when working with AI.

They choose not to learn prompting because:

  • They think it’s difficult to learn.

  • They think good writing skills are enough.

  • They believe it will become irrelevant as AI evolves.

  • They actively resist change and dislike the novelty of AI.

Luckily, prompting is not difficult to learn and anyone can get good at it quickly.

All it takes is the willingness to learn and some time to actively apply the knowledge.

In this newsletter, I’ll cover the basics of good prompting and provide templates and examples.

We won’t have time to go into every aspect of prompting. But nonetheless by the end you will have a thorough understanding of the basics.

the basics

First we need to understand that AI doesn’t know anything about who we are and what we are trying to achieve.

Therefore, specificity will be key when crafting our prompts.

Consider the following points to immediately improve the quality of your prompts:

  • Define the task using an actionable verb (like “Generate“, “Analyse“ or “Summarize“)

  • Provide relevant context such as background information, constraints, and requirements

  • State the intended audience (who are we writing to?)

  • Specify the writing style (you can have AI take on a relevant persona)

  • Indicate the output format you prefer

Another helpful practice is to provide AI with an example of what you are looking for.

Examples written by humans can greatly help AI produce more human-sounding responses.

Here’s an example of what a prompt like this would look like:

“As a seasoned YouTube strategist, analyze the effectiveness of 3 common YouTube thumbnail styles for a tech review channel. Consider click-through rates, audience retention, and overall channel growth. The intended audience is aspiring tech YouTubers. Present your analysis in a concise report format with bullet points for key takeaways.”

It’s important to adopt the mindset of a mad scientist when you craft your prompts.

If you don’t get the desired result on the first try, change the wording in your prompt.

Sometimes one word can make all the difference. Iteration is key when building your prompt.

Keep experimenting and take note of what is working and what is not.

level up

We now have a good understanding of the basics. So let’s take things up a notch.

One key aspect to consider when crafting your prompt is to keep it simple. AI typically doesn’t do very well with large complex tasks because it only has so much cognitive power per request.

If you use up all that cognitive power early on in a long prompt, the quality of your output will suffer.

Therefore it’s a good idea to break down your tasks into sub-tasks.

This is something that’s called “chain-of-thought prompting” or “progressive prompting”.

Below you can see what a progressive prompting sequence could look.

In this example we are asking AI to develop a content strategy for growing a personal brand:

Progressive Prompt: Develop a Content Strategy for Personal Brand Growth

  • As a content strategy expert, what are the key elements to consider when developing a content strategy for a personal brand in the digital space?

  • Given the elements you mentioned, how can a content creator effectively identify and refine their unique voice and niche?

  • Based on the identified niche and voice, what types of content formats and platforms would be most effective for building a strong personal brand? Provide specific examples.

  • Considering the content formats and platforms you suggested, estimate the time investment required and potential impact on brand growth. Which approaches offer the best balance of effort and results?

  • Based on the insights gathered, act as a personal branding consultant who specializes in digital content strategy. Recommend a comprehensive content strategy for a content creator looking to grow their personal brand.

  • Include target audience definition, content pillars, platform selection, posting frequency, and key performance indicators. Output your results using markdown including headings, bold, and bullet points.

This method ensures AI uses its limited cognitive power to the fullest.

Speaking of clarity. Another thing you want to avoid is providing AI with one big block of text.

This makes it hard for the AI to make sense of your request.

AI love things like ##, “” or line breaks to indicate different sections of your prompt. When you provide an example in your prompt, you can indicate to AI that the example is within ““.

Now AI can easily identify what part of the prompt is your instructions and what part is the example.

AI also loves bullet point formats when you provide instructions. Again, it makes it easier to digest the information.

With these points in mind, I have created a template you can use:

[Instruction/Question]

Context: [Provide relevant background information]

Persona: [Give AI a relevant role]

Example output:

“[Include a specific example of what you want]”

Requirements:

  • [List any specific requirements]

  • [Such as word count, tone, format, etc.]

Here’s an example of what it could look like using the above template:

Create an engaging YouTube video script for a product review.

Context: You’re a tech influencer reviewing the latest smartphone. The product has innovative features but also some drawbacks. Your channel typically blends humor with technical analysis.

Persona: You’re a no-nonsense tech reviewer known for brutal honesty and occasional sarcasm.

Example output: “Hey tech warriors! Today we’re dissecting the HyperPhone X – the smartphone that promises to do everything short of making your coffee. Let’s cut through the hype and see if this gadget is worth your hard-earned cash or if it’s just another overpriced paperweight.”

Requirements:

Script length: 800-1000 words

Include at least 3 pros and 3 cons of the product

End with a clear recommendation: buy, wait for next version, or skip

Incorporate 2-3 sarcastic one-liners

Structure: Intro, Features Overview, Pros, Cons, Performance Test, Conclusion

No sugarcoating. If it’s bad, say it’s bad.

Using the provided prompt template will instantly improve the quality of your outputs.

what to avoid

With a solid prompting strategy in place, let’s finish up with a few things to avoid to get the best possible outcomes.

First you must avoid being ambiguous. This leads to AI having to fill in the blanks and guess what you mean.

This in turn leads to inconsistent results. Make it clear and obvious what you are saying. Use 6th grade writing language for best results.

Building on the idea of keeping things clear and simple: avoid jargon.

No fancy words. No complex terminology.

Keep it simple.

Simple prompts with enough information to provide context are ideal.

You need to find the balance between giving AI flexibility while also being specific enough. This will require some experimentation on your part.

It took me some time to get this right for myself. But it’s definitely a skill worth practicing.

Finally, you need to know the limitations of the AI model you are working with.

For example, if you are working with a large language model (LLM) like ChatGPT or Claude, don’t expect it to do math. Not even the most basic math!

LLMs are not designed for that and are notoriously bad at it. Even asking for a tweet with no more than 280 characters is already too much math for LLMs to handle.

Try it for yourself. Sometimes it gets it right, but more often it doesn’t.

conclusion

If you follow the simple advice outlined in this newsletter, you will be far ahead of the majority of people using AI.

Most people don’t take prompting as a serious skill, but if you’ve read all the way to the end you clearly do.

So well done on taking this skill seriously. I can guarantee it will pay off in the long run.

But like any other skill, you must practice to become good. You now understand the basic theory, but next you need to put it into practice.

I suggest that you spend at least 20 minutes a day experimenting with different prompts. That way you’ll get a first-hand experience of what works and what doesn’t work.

Happy prompting!

Until next time,

Nic

Next
Next

how to future-proof yourself