🤖 Use AI to develop & test a model, using the "Startup Core"

A key to startup success is maintaining and iterating on a testable model. It’s harder than it sounds. Fortunately, your AI co-founder can work with you to define it and stress-test it.

Hey friends 👋

This is Founding with AI, your complete guide to starting, running, and scaling successful companies using AI as your co-founder.

In this inaugural post, we’re talking about one of the most important and least performed acts of startup founders: developing a mental model of the business.

We’ll start with the theory behind modelling, then move into how you can use your AI co-founder to supercharge your effort.

But no FWAI issue would be complete without ChatGPT prompt templates you can put to use right now.

So let’s go 👇

I’m kicking off this newsletter with the first part in a series on defining what your startup “is”.

Most founders think they know:

• we’re X for Y

• we have an app that does A

• we help B with C

• etc.

But that’s too shallow to be useful, so let’s take a step back and talk about why we need a model at all.

The 3 rules of being a startup founder

The work of the startup founder is following three key rules, which form an iterating loop:

1. Maintain a model

2. Separate fact from conjecture

3. Test the riskiest conjecture

It’s not complicated.

You first have to know what the thing is. Then you have to know what of that thing is known, versus what is assumed. Then, you have to test the assumptions that are key to the success of your business but which are the least known.

Finally, rinse and repeat.

But what do we mean by “model”, exactly?

When I talk about models with startup founders, they usually think spreadsheets with financial projections, and hockey stick graphs, and balance sheets, and so forth.

And it might be that… eventually.

But at first, it’s just a mental model. It’s just what we think the startup “is”.

Later, your model might be driven by pirate metrics, with an eye toward the earliest metric for which we find risky conjecture.

And, sure — it’s eventually a robust financial model that we’re testing in the marketplace.

But regardless of your stage, it’s something that we can use to tell us what the thing “is” that we’re talking about, and what it means to be on the right track.

After all, one of the biggest experimental mistakes founders can make is to take an action in order to “see what happens”. Instead, take an action in order to prove what you think will happen because your model is correct.

Always know your True North, stress test it constantly, and incessantly revise it.

So what kind of model should you develop?

The simplest model is usually most impactful.

It’s easy to overcomplicate a startup, particularly when we start dropping terms like “model”. It takes persistence and grit to keep it as simple as it needs to be to remain both useful and agile.

I use what I call the “Startup Core”, and it couldn’t be simpler.

It’s based on an academic definition of a business model called the Magic Triangle, which I’ve augmented into a model that you can use starting on Day 1, but still works on Day 1000. In other words, it can be a mental model at the outset, and it can power a robust financial model when you're ready.

The Startup Core answers four key questions about your business:

• Who is the customer? Do they exist? Do they have the problem you think they do? Are there enough of them out there to make this worthwhile?

• What is the value you’re creating for the them? Do they want a solution to this problem? Do they want yours? Do they need it with sufficient severity and urgency to drive a purchase?

• How is that value created and delivered? Do customers want it in the way you’re offering it? Are you able to get the offer in front of them cheaply and efficiently?

• Why will this make money? Will they pay what we need to turn this into a viable business? Is this a large enough, sufficiently growing market to make this worth your time? Is now the right time to chase the opportunity?

There’s a lot packed into those 4 questions, but want to know what the cool part is?

All the models I mentioned above (mental, pirate, and financial) are all based on the same core.

Over time, we go in greater depth and add greater specificity, but they all start with the same four questions. In fact, the Business Model Canvas is just a 9-box representation of the above four attributes.

The Startup Core starts out simple, but grows with you. As your assumptions become validated learning, you add more detail and measure with greater specificity, which automatically becomes more robust model.

Boom!

The Startup Core powers everything you do.

I said earlier that the first rule of being a startup founder is to maintain a model.

Not to have a model, but to maintain one.

This is implies a level of dynamism:

• Use the Startup core to document your model

• Identify the riskiest assumptions within that model

• Run experiments to turn assumptions into validated learning

• Rinse and repeat.

Because, after all, that's all a startup is.

But it all starts with having a Startup Core that’s solid. A flabby, flimsy model produces flabby, flimsy results. The good news is that we can use our AI co-founder to stress-test our model.

Here’s how 👇

Workshop your Startup Core with your AI co-founder.

We’ll tackle this in 4 simple steps:

Step 1: Train the AI

Grab your Startup Core notes, your elevator pitch, or your random musings on your startup idea and drop them right into this prompt.

It takes your notes as input, and then subjects it to the four questions that make up the Startup Core. It will provide feedback on what it thinks you might be missing, as well as suggestions for improvement to close any of the gaps it finds.

This should work in both free and paid ChatGPT, as well as in Claude and Gemini, but the examples I use are from ChatGPT 4o.

Here’s the prompt: