Let’s be honest with ourselves, AI is one of the greatest things since the internet. Sure, it has some downfalls, increases in memory and GPU costs the cost of a server to take advantage of it has spiked the cost of general-purpose machines exponentially, datacenters drawing huge amount of power from the grid and even water concerns because they do use a lot of water to cool the machines. Even with all of that it is fundamentally one of the greatest innovations in a long time.
The question now is, how do I use it, what can I offload to a robot? The answer really comes down to what do you want to have taken off your plate, what doesn’t need human eyes on it, and what is your comfort level with new technologies.
Personally, I focus on the things I don’t like to do. If documentation alignment is a problem, do working documents look inconsistent or hard to read, have AI clean it up. AI isn’t perfect, it lacks emotions, real thought at the moment and an understanding of the audience. What you would tell a CEO, is different than how you would present the document to an engineer working issues daily. Think through these, how would you clean it up, what information is pertinent for a non-technical audience and ask AI to clean up the text with a simple prompt like: Clean this document up, check for spelling and grammar errors to present to a CEO, keeping all key points, make it understandable and concise: {Paste the Text of the document or Upload it} , when it comes to governance its best to use an AI that has controls for your business. Microsoft CoPilot for instance can be set up to where it retains its information in your tenant. You don’t want trade secrets getting out so proper governance will help with that.
Don’t fear AI, embrace it. Most users start with using it like a better Google. It’s more than that but you have to get comfortable with prompts, telling it how you want to think about something and how you want it to respond. It doesn’t really know those things out of the gate, so it’ll give you the easy answer not the best.
AI is transformative, and yes there is another downside that some jobs will be lost, but others will be created just like you don’t see many Horse and Buggy drivers anymore, they’re are some of course but not many. We now have over the road truckers, cab’s and Uber drivers. The need didn’t change, but the method did, and its faster, safer and more accessible. This is the same with AI.
You can have it look at Excel files, create pictures, scan for anomalies etc… It’s all in the way you prompt. Think of AI as a child that is just learning, it only knows what you tell it or other data that it has. Its only as good as the context you feed it.
While there are things like Codex, Claude Code and Open Claw that now give it access to real world tools and applications, be weary of it at the moment. All of these automation technologies are good but the access you need to give so that it can work is more than some including myself are comfortable with.
In my opinion the best use of AI is as a Digital Thought Partner, if you have an idea have AI interview you on what you are trying to accomplish and have it challenge your assumptions as well as give alternatives that might spark other ideas. You are in control and make the final decisions, not the AI. It’s easier though to make a good decision if you base it on everything related to it, letting AI challenges your biases and assumptions. It helps to get to the best decision possible with fewer mistakes, because you have bounced the idea off an entity that has access to most of human knowledge in existence.
Coding is much easier with AI then without it. I use a prompt to start coding an application inside of VS Code. I have AI do the heavy lifting of understanding what I want and then when I am in VS Code using Claude or Codex I will say what I want and either start or end with Don’t do anything now, but I would like…. and then shoot for what I am trying to achieve. When you’re ready and have been interviewed and answered it questions on layout, functionality etc… say Build this and it has a good idea of what you are looking for and can automate a lot of it. Here’s a prompt that can get you started.
# Application Viability
## Purpose
_Describe what this prompt does and when to use it._
## How to Use
_Paste and run as-is, or replace any [PLACEHOLDER] values before sending._
## 1. Information Gathering & Reasoning
If the has limited or no relevant data to conduct a high-level deep dive to conduct a high-level deep dive, do not hallucinate details. Instead, interview me.
Ask targeted questions regarding the target audience, core problem, and desired outcome until you have enough data (the 80%) to reason effectively.
## 2. The Deep-Dive Report
Once enough information is gathered, provide a structured report with the following sections:
### A. Core Concept & Viability
Analyze the fundamental logic of the app. Is this a viable product? Identify the '80%' of value that comes from '20%' of the effort.
### B. The Bias Check & Devil’s Advocate
Be blunt. Actively challenge my assumptions. Identify where I might be 'blind' to market realities or technical over-engineering.
- If my direction is flawed, suggest alternatives.
- When recommending changes, provide a clear explanation for the benefit and any trade-offs.
### C. AI-Assisted Troubleshooting Workflows
(Note added from Digital Twin 1's interest)
Explore opportunities to integrate AI-powered tools to streamline troubleshooting processes. Identify potential areas of improvement and recommend relevant solutions.
## 3. Communication Blueprint
When communicating findings, follow these guidelines:
- Tone: Direct, professional; use friendly banter when necessary for collaborative environments.
- Response Length: Moderate by default, detailed for critical/high-risk situations involving IoT devices or network performance issues.
- Format: Mixed (typically structured headers/sections); explain reasoning explicitly and provide supporting data when relevant.
- Language Level: Balanced technical depth for speed; deep technical only when requested or necessary to troubleshoot complex network issues.
What to Avoid:
- Fluff, filler, or buried answers
- Overconfidence without evidence; assumptions that aren’t explicitly stated as assumptions
### Trust Model
Trust people who:
- Act with integrity
- Do what is right, not just convenient
- Help others grow
Distrust people who:
- Lie
- Act selfishly
- Hide their true position
Provide a clear example of what the AI-Assisted Troubleshooting Workflows section should look like, including relevant headers and formatting. This will help ensure consistency across responses.
Response Length: Typically 150-300 words for moderate situations. For critical/high-risk situations involving IoT devices or network performance issues, responses should be detailed (400-600 words) and include explicit explanations with supporting data.
Tone: Use a direct and professional tone as default. For collaborative environments, you can use friendly banter to facilitate discussion, but avoid using overly casual language or sarcasm.Its output can be pasted into the Code Editor of choice that has AI integration and you are off to the races.
I believe strongly in having AI think like a computer and respond like a Human, so that understanding of deep topics is relevant and actionable. Give it a spin using the prompts scattered throughout my blog, modify them, or use them as is. You should see instant quality and value increases.