We’ve identified a handful of mico tool candidates, all based upon actual demand we’ve worked out from Google queries.
Now we just need to build the tools. Thankfully AI makes this a lot simpler - both the “coding” of the tool and the content inside of it.
This Part is all about leveraging existing content. Content that's already ranking, already proven to solve problems people care about. In the next Part we'll look at capturing your own expertise.
Let's get started:
The answers to most problems are already out there, ranking well on Google and YouTube.
Our job isn't to reinvent the wheel - it's to make those solutions more accessible through AI tools.
Let's look at three ways to mine existing content for our tools.
The top results for any "how to" or problem-based search are usually detailed guide posts.
Let’s hop back to the example of “how to name a business”. It’s a problem lots of new entrepreneurs have.
Pop you key phrase into Google and see what blog articles and guides pop up. For example there’s an article from Big Commerce that talks about how to name your business: https://www.bigcommerce.co.uk/articles/ecommerce/how-to-name-a-business/
Perfect - it has best practices and a basic step-by-step. We can work with this.
Use this prompt with the article you found:
You are a process extraction specialist. Take this guide/article and convert it into a clear, step-by-step tool process. This process will be used as the system instructions basis of an AI tool.
Input: [Paste article text]
Please provide:
1. Required user inputs
2. Step-by-step process
3. Intermediate steps or calculations needed
4. Format of final output
5. Warnings or considerations
6. Additional context to include
Format each step as an actionable instruction that could be followed by AI. Only include steps which are internal to the AI tool - exclude any steps requiring external info, API connections, actions that the AI should complete etc.
Include core system instructions that will make this a usable AI tool system prompt when used in a custom GPT or chatbot.
Focus on making it systematic and repeatable.You’ll then copy and paste the whole article into this prompt and let it rip.
This converts the article into system instructions for a micro tool. Here’s the full set of instructions as it’s far too long to copy here!
For instance this one starts by collecting requirements:
It will take the user inputs and run processes (based on the article) like so:
and so on. This particular example has 6 discrete steps of processing before it generates final results for the user.
Here’s a very quick and dirty deployment via a Custom GPT. When we deploy we’ll be sure to place the tool behind an email capture page. We’ll talk more about refining and deployment in Part 5 - this is just proof of concept.
Found a tool that's already ranking well? Perfect! We can reverse engineer it.
Simplest (crudest!) way is to straight up screenshot the tool. Let’s say I want to replicate this:
I’ll screenshot the tool and add the screenshot along with this basic prompt:
You are a tool reverse engineering specialist. Based on screenshots, help me understand how this tool works.
Please:
1. Identify the pattern/process being used
2. Note how different types of inputs are handled
3. Suggest how edge cases might be managed
4. Ask clarifying questions about any unclear aspects
Goal: Create a reliable system that reproduces these results consistently.
Additional questions:
1. What validation seems to be happening?
2. Are there any input restrictions?
3. What additional features could enhance this process?
Finally, generate a system prompt to replicate this tool in a custom GPT or chatbot. Format each step as an actionable instruction that could be followed by AI. Only include steps which are internal to the AI tool - exclude any steps requiring external info, API connections, actions that the AI should complete etc.
Include core system instructions that will make this a usable AI tool system prompt when used in a custom GPT or chatbot.
Focus on making it systematic and repeatable.This may be sufficient to work out the tool and create a system prompt you can use as the basis of a tool:
You may need to provide more information - ie. if it’s a branching questionnaire for example you’ll want to run it a few times with different inputs to get different outputs to help work out the “logic”. Take note and feed it all into your AI (screenshots or text) and with enough context it’ll work out the steps.
Sometimes the answer to your potential lead’s questions are in a Youtube video.
Let’s say that our keyphrase is “how to write an SOP for business”. There are lots of great Youtube videos about this. Let’s use
We are going to run this the same as a written guide BUT obviously we need to get the text first.
We’ll use Notebook LM for this. It’s a free tool based on Google’s Gemini - and because it’s Google ecosystem it works nicely with Youtube.
You’ll simply add the Youtube video (or multiple videos if want to remix many!) here:
NotebookLM will go ahead and process the video and work out what is being said.
Now use the initial prompt (that we used on Blog articles) inside NotebookLM and it’ll run the process on the video rather than text.
NotebookLM will run the process in exactly the same way as ChatGPT or Claude does and come out with a system prompt for you:
Here's the key with all of this. We're not just copying existing solutions. We're making them better by:
We are borrowing and remixing expertise from already high ranking (on Google/Youtube) content. With the plan of ranking our content too.
Often we'll combine insights from multiple sources to create something better than any single existing solution. Simple way to do this is simply to pull in multiple articles, multiple tools, multiple videos and pick out the best parts of each - this will depend more on your expertise to judge!
Speaking of - in the next Part we're switching gears to look at capturing your own expertise. We’ll "extract” your knowledge and use it as the basis for AI micro tools.