Free preview8 min

Finance Prompting Principles

Reviewed by Human · Updated October 6, 2026

✗ Wrong way

What's Apple's stock price right now, and will it go up next quarter? Give me a target price.

Two fatal asks in one prompt. The model has a training cutoff and no live market feed unless it's connected to a data tool, so any "current price" it gives you is either stale or invented. And no model can forecast a quarter's price move—if it could, it wouldn't be answering your prompt. The confident-sounding number is the dangerous part: it looks like data but it's fiction.

✓ Better way

Here are Apple's last four quarters of revenue and operating margin [paste figures]. Walk through what these trends suggest about the direction of the business, and list what else I'd need to build a defensible valuation. Do not estimate a share price.

You supplied the data, so the model isn't guessing at facts. You asked it to reason over what you gave it and to name its own gaps—which is exactly the work it's good at.

Most people point AI at the one thing it's worst at. Guess which layer of finance work it's genuinely great at?

Can AI Give You Real-Time Stock Prices?

No—not on its own. **Training cutoff**: the fixed date after which a model has no knowledge of the world. **Retrieval / tool use**: connecting a model to a live source (an API, a spreadsheet, a search tool) so it can pull current data instead of recalling it. Unless a tool is explicitly wired in, a model answering "what's the price today" is recalling a snapshot from its training data or fabricating something plausible. It will rarely warn you which. So the first principle of finance prompting is simple: **you bring the numbers, the model brings the reasoning.** Give the model the context it needs—the actual figures, the period, the units—and ask it to analyze, explain, or structure. Never ask it to *know* the market.

Rule of thumb: if the answer depends on a fact that changes by the minute, the model shouldn't be the one supplying that fact.

Your turn

Below is a prompt that asks the model to be an oracle. Rewrite it so the model becomes an analyst working on data you provide.

Reflect

Notice how much calmer the rewritten prompt feels. You stopped asking for a verdict and started asking for help thinking.

You need Q3 revenue figures for a competitor in your report. What's the safest way to involve AI?

📚

Go deeper

end of module

You've finished this module.

Mark it complete to earn your XP and keep your streak alive.

Progress saved locally · Sign up to earn XP

Financial Modeling Support→