Перейти к содержимому

I Tested GPT-6 Astra on a 5-Year DCF (It Missed One Big Thing) | Financial Modeling

Al Anany

0:00 / 0:00

I Tested GPT-6 Astra on a 5-Year DCF (It Missed One Big Thing) | Financial Modeling

209 просмотров · 2 нед. назад
Al Anany
840 подписчиков
209 просмотров · 2 нед. назад
I Tested GPT-6 Astra on a 5-Year DCF. It nailed the execution, but missed one big thing. A month ago, an AI lab hired me to find flaws in their models' financial modeling. So I know exactly where these tools break. A few weeks ago, I spent two weeks building a financial model for a legal tech client. Then I asked myself: could GPT-6 Astra replicate that work? I couldn't use my client's data, so I built a similar case around Scenefixer, an AI video continuity tool I built for a hackathon. In this video, GPT-6 Astra (high thinking effort) builds: 5-year projections based on real usage & traffic Income Statement Balance Sheet Cash Flow Statement Input sheet (Control Panel) linked to the whole model Valuation sheet Sensitivity analysis (incl. "what if AI traffic disappears?") Scenarios: Base, Upside, Downside, Heavy Usage This is how to build a financial model with ChatGPT (GPT-6 Astra), and what you still need to check yourself. The formulas, links and balancing came out clean. But the logic is where a financial analyst still matters. I break down the revenue assumptions, the funding gap, and one risk the model completely missed. That's the kind of risk an investor would catch in the first meeting. I also compare it with Claude Opus 5.5 and Sonnet to see which model you can trust for financial modeling. If you work in finance, startups, consulting, or FP&A, this shows exactly where AI saves you time and where it can still embarrass you in front of investors. 🔍 Want to audit your own model before an investor sees it? Finixr Audit runs your model through Claude, GPT and Gemini independently and shows where they agree: https://audit.finixr.com/ ⏱️ Timestamps 0:00 Intro – Can AI Replace 2 Weeks of Client Work? 0:57 Setup & Example Company 1:21 The Prompt: 3-Statement DCF, Valuation & Sensitivity 4:51 How to Audit an AI-Built Model (Finixr) 6:48 Reviewing the Model: Revenue Reality Check 9:37 Inputs, Scenarios & Financial Statements 11:53 Sensitivity: What If AI Traffic Disappears? 12:09 Verdict: GPT-6 Astra vs Claude Opus 5.5 13:22 The Risk No AI Flagged Drop a comment: which model do you trust more for financial modeling? Let's discuss!