Part 1: Using AI for Due Diligence? The Model and Prompt Isn’t the Whole Story.

This is part one of a series I'm working on based on real conversations with real professionals using AI in their day-to-day work.

Someone asked me this recently: “If I’m using Perplexity and select Claude as the model , won’t I get the same answer I’d get if I just used Claude?”

Their point being: “Why would I pay to use both platforms?”

Here’s the clearer way to think about it.

What a Model Actually Is

A model like Claude (built by Anthropic) or GPT (built by OpenAI) is a trained machine learning system that predicts text.

Given input text, it generates the most statistically likely next tokens based on patterns learned during training.

• It does not “know” things in the human sense.

• It does not decide how it is used.

• It simply produces output based on the input it receives.

That’s it. That’s all the GPT does.

Here's a way to think about this. If every model was trained that 5+4 = 10. When you asked a model what 5+4 is... it would most likely answer 10 (unless it was hallucinating). Models aren't doing math. They're doing pattern matching based on their training data.

What a Chat App Is

A chat app is a system that uses one or more models.

When you type a message into Perplexity, Claude’s app, ChatGPT, or any other chat platform, you are not sending raw text directly to an ‘all knowing’ model. The model is wrapped in a complex system that just makes it seem ‘all knowing’ to the user.

Even when two apps use the same model, they produce different answers because the systems 1) Are built around prediction engines. 2) Are designed by teams with different goals and priorities.

And depending on the team building the chat app, it may be stronger in some areas than others.

• Some users find Perplexity strong for research and sourcing.

• Some prefer Claude for creative writing or coding tasks.

• Some use tools like Manus for complex, multi-step projects.

• Others choose ChatGPT because of its API access or ecosystem.

And so on.

The point is not that one model is “smarter” than another.

The point is that you are not just choosing a model.

You are choosing an entire system of product features and tools built around that model—a system that may provide access to different tools, apply different guardrails, include or exclude web search, connect to your files, or structure answers in a specific way.

• One platform might automatically search the internet before answering.

• Another might rely only on the model’s internal training unless you ask it to look something up.

• One might allow code execution or file analysis.

• Another might not.

These differences are not about the model itself.

They are about the product decisions being built around it.

This is what changes the experience — and, the results. And this is why this matters. Especially, if you're trying to use AI for due diligence.

In the next post I'll talk about system instructions. Contact me if you'd like us to cover anything specific.

---