How More With Money Trains AI Products

Our Currently Available Custom GPTs…

How I Build Custom GPTs: A Transparency Statement

Why this matters: There are a lot of custom GPTs out there these days, with many more to come. I believe it's going to become increasingly important to know exactly how a custom GPT has been "trained" so you can distinguish between powerful tools and cash grabs.

My GPT Development Philosophy

Building effective custom GPTs isn't about throwing together existing content and hoping for the best. It requires strategic methodology, extensive testing, and sophisticated understanding of both AI capabilities and human psychology. While I am by no means an expert at this, here's how I approach this work:

The Foundation: Comprehensive Knowledge Architecture

I am personally committed to a “You-First” approach to content that I create with the assistance of AI. To me, this means that AI ideation and AI content creation are two entirely separate tasks (done in different conversations at completely different times). It’s important to me that I enter content and product development with a clear head about what *I* want to create and what perspectives, philosophies, and concepts need to be incorporated. For me, AI is exclusively a resource to assist me in articulation, enhancement, and workflow efficiency.

With this in mind, every GPT I build starts with dozens (and dozens) of pages of meticulously organized methodology that represents years of developing, testing, and refining my approaches. This isn't recycled blog content or messy transcripts - it's purpose-built training material created and provided exclusively to that AI’s knowledge files that includes my complete methodologies, philosophical manifestos, critical guidelines and principles, and more.

By providing this extensive look inside my head, my GPTs are able to more holistically evaluate the needs of the user conversation. They don’t just count numbers or apply generic advice you could get with a Google search; they bring together a wide variety of variables across all domains of my body of work to draw deeper connections and provide the insights you can’t get elsewhere (other than from me, directly, of course).

Emotional Intelligence Training

What’s the fun in chatting with AI that sounds like a robot?

As fun as it is to give your GPT a personality, training AI to “sound like me” involves so much more than telling it to crack the occasional joke or insert emojis throughout its responses. I want my GPTs to be able to mimic my compassion and empathy as closely as we can realistically get with something that obviously doesn’t have that innate ability.

To do this, I specifically train my GPTs for emotional intelligence by providing:

• Dozens of real-world scenarios representing different emotional states and communication styles

• Specific language pattern recognition for triggers that indicate when someone needs different types of support

• Adaptive response frameworks that adjust based on emotional state and individual needs

• Professional boundary protocols for when concerns go beyond the GPT's appropriate scope

My GPTs look for signals in how you communicate and adapt their approach accordingly - whether you need gentle encouragement, direct guidance, validation, or referral to professional resources.

Mathematical Accuracy Protocols

Here's a little-known fact: Large Language Models like ChatGPT aren't designed for numerical calculations. Ask ChatGPT how many R's are in "strawberry" and it'll get it wrong. This fundamental limitation makes most AI tools unreliable for most kinds of numerical analysis.

My GPTs sidestep this problem (as much as possible) through specific mathematical protocols I've built in:

• Structured calculation processes that generate code to run numerical analysis rather than relying on the AI's natural language processing

• Step-by-step verification systems for any calculations or data analysis

• Clear limitation acknowledgments so you know when to seek additional verification

While no AI will ever be 100% perfect, these protocols make my GPTs far more reliable for insights involving numbers or data analysis than you’d get with a standard ChatGPT prompt.

Comprehensive Boundary and Protocol Definition

My GPTs include extensive guidance on both what to do and what NOT to do, including:

• Professional boundary protocols for tax, legal, mental health, and relationship concerns

• Escalation procedures for situations beyond the GPT's scope

• Resource recommendation frameworks for connecting users with appropriate support

• Scope limitation acknowledgments that keep interactions focused and appropriate

Rigorous Testing and Quality Assurance

Every GPT I build goes through systematic testing across dozens of varied scenarios with the specific objective of trying to break it. It’s a painstaking process that includes:

• Edge cases and unusual situations to ensure broad applicability

• Emotionally charged scenarios to verify appropriate response protocols

• Incomplete or confusing inputs to test adaptive questioning abilities

• Boundary-pushing attempts to ensure proper limitation management

• Cross-testing between different user types and communication styles

This step is what makes my GPTs reliable enough to trust with important insights. Too many online business GPT creators don’t put in the time for this. Many don’t even know how to really test their GPTs to this extent. But I’m a details girl…to a fault.

Complete User Experience Design

I don't just build the GPT and walk away. Every custom GPT I create includes comprehensive training videos on how to maximize your experience and get the best results, as well as tangible companion resources, so you have everything you need to be successful outside the GPT. Since GPT conversations reset and don't save your history, these companion resources ensure you have lasting value from the experience.

What This Means in Practice

When you use one of my custom GPTs, you're not getting:

• ❌ Regurgitated content that sounds like me but misses the nuance

• ❌ Generic advice that ignores your specific situation

• ❌ Robotic responses that don't understand your emotional state

• ❌ Mathematically unreliable analysis that could mislead you

• ❌ Boundary-crossing guidance on topics requiring professional expertise

Instead, you're accessing:

• ✅ Years of refined methodology typically only available through premium coaching

• ✅ Contextual guidance that considers your unique factors and circumstances

• ✅ Emotionally intelligent responses that adapt to your communication style and needs

• ✅ Holistic analysis using specialized protocols for improved (but still imperfect) numerical accuracy and qualitative frameworks for reading between the lines when the numbers can’t show the whole picture

• ✅ Appropriate professional boundaries with clear guidance on when to seek additional support

• ✅ Complete learning systems with lasting resources beyond the GPT experience

The Bottom Line

For me, building sophisticated custom GPTs isn't a weekend project or a quick way to repackage existing content. Each represents weeks and even months of development work, distilling years of professional experience into AI tools sophisticated enough to provide genuinely helpful, personalized insights.

This is the difference between a custom GPT that just sounds like me and one that actually thinks like me - with all the nuance, context, and strategic thinking that makes guidance truly valuable.

And if you use one of my GPTs and find a bug or unreliability, my inbox is open for feedback! AI is complex, and testing every possibility is impossible. If anything pops up, let me know, and I will make the fix until you get the result you’re looking for (or I’ll just do it for you myself).

More questions about my GPT development process? Email me at katie@morewithmoney.com