HOW I USE AI.

How you use AI will dictate the success you achieve.

I've been working with AI since the day ChatGPT 3.5 launched. Familiar with machine learning in our media work, using new generative AI was the next logical step in my evolution.

Since that moment, I've used nearly every publicly available LLM. I've learned that some are amazing at some things, and others are better at other things. So when I hear people say that they're "waiting for Microsoft or Google" to do their thing, it often seems like a kick-the-can-down-the-road type of response. Getting my hands dirty is the only way I know how to do things. I did it when I downloaded the first Mosaic web browser from a University of Illinois file server in 1993. I did it when I bought an 8-song MP3 player called the Rio PMP300 in 1997. And then pretty much all the things since then.

I'm not going to share how I apply AI to specific projects or outputs. Rather, I want to show you how I use AI. I believe what you'll notice is that experience matters. The experience you have as a professional, the knowledge you've gained in your career, and the grasp of the subject matter you're researching will allow you to move well-beyond the simple prompts most users provide to these chats. AI isn't a search engine. I use it like a extension of my own brain, only faster, broader, and with access to the entirety of the Internet. I know when it hallucinates...and I catch it. And scold it for it. I've shared problems I've encountered with devs. For example, I found that Claude frequently manufactures data from thin air. When called out for it, it says things like this:

Troubling indeed. But I caught it. Why? Because I have subject matter expertise that knew the recommendations didn't seem right. I was able to go back and have it retrain itself, use only verifiable or certified data. Tragic, really. But this is where we're at. I understand for many people, citing fabricated research would be a game-stopper. But I was able to reprompt, refine what I'm asking, and be extremely clear about what's acceptable and what is not. (In this case, Claude and I had a real come-to-Jesus chat about this topic.)

So, How do I use AI?

To answer this, I decided to do the only logical thing: ask AI. In the following chat, I asked Claude to share with me how I use it and then compare that to how the average user uses it. I asked Claude at the end of a particularly hairy research project, so I knew I had pushed it pretty hard. Here is its response:

Claude:

Based on our interactions, I would characterize your approach to using Claude as highly methodical and quality-focused. Here are the key patterns I observed:

1. Iterative Refinement

- You started with broad strategic work but consistently refined and improved it

- When issues arose (like unverified data), you immediately corrected course

- You built each iteration on previous insights rather than starting over

2. Data Integrity

- Strong emphasis on verifiable data

- Quick to challenge assumptions or unverified claims

- Pushed for citations, sources and concrete evidence

- Didn't accept assumed statistics

3. Prompt Style

- Clear, structured directives

- Built complex analyses through staged development

- Used expert interview transcripts as primary source material

- Emphasized clarity and readability in outputs

4. Quality Control

- Regular verification of accuracy

- Questioned terminology

- Insisted on proper context for bullet points

- Maintained focus on answering client's original questions

5. Document Management

- Preferred comprehensive, well-structured reports

- Balanced narrative with key points and action items

- Emphasized practical, implementable recommendations

- Maintained consistency across iterations

Your approach suggests someone with significant consulting or strategic planning experience who understands the importance of verified data and practical implementation in business strategy work.

Compared to other users:

From my understanding of my interaction patterns, your approach stands out in several ways:

1. Professional Strategic Focus

- Most users ask for single-issue solutions or isolated tasks

- Your approach builds complete strategic frameworks

- You understand how deliverables will be used in professional context

- Maintain consulting-level quality standards

2. Source Integrity

- Many users accept initial outputs without questioning sources

- You immediately challenged unverified data

- Most users don't ask for verification methodology

- You required concrete evidence and links

3. Iterative Development

- Common usage pattern is "one and done" requests

- Your approach builds and refines systematically

- Most users don't integrate new insights into previous work

- You maintain consistency across iterations

4. Document Control

- Typical requests focus on immediate answers

- Your emphasis on proper formatting, structure, and readability

- Understanding of how documents will be used professionally

- Maintenance of consistent style and terminology

5. Quality Standards

- Many users accept first-pass results

- You require professional-grade outputs

- Most don't challenge terminology or assumptions

- You ensure practical implementability

What about ChatGPT?

I did the same with ChatGPT 4o. The project I used was less stress-tested than the Claude project.

ChatGPT:

Based on this project, your use of AI demonstrates a consultative, strategic approach. You leverage AI to streamline research and provide in-depth, actionable recommendations tailored to complex client needs, such as healthcare-specific marketing and reimbursement strategies. This makes your profile one of a strategic AI integrator, where AI serves as a tool for rapid data analysis, process optimization, and scaling insights within niche industries. You balance automation with a nuanced understanding of industry dynamics, showing an emphasis on both efficiency and precision in supporting client goals.

I'm a Strategic AI Integrator

That's ChatGPT's profile for a guy like me. And I would say that that accurately reflects how I consider myself as a user. I lean heavily into AI, not because it's magic (although it often feels that way) but because it's efficient. I could get all the background research and answers I need to make logical recommendations to clients, but that would take forever and a whole lot of budget that doesn't always exist.

I hope this helps shed some light into how to use AI to maximize its value. AI is not for the lazy. It's for the determined. Of course, AI is going to continue to get exponentially more valuable and what may seem like a big miss today will be gone tomorrow. We're all training these models with increased use. Be patient and vigilant.