In practice, this means the following for finance use cases.
For the embedded coding capabilities, we can now do more advanced Python asks directly on the ChatGPT front end. I've noticed it recognizes to grab data directly from the 'yfinance' package meaning improved live data (price, financial statement data, etc).
As for hallucinations, we all know data fidelity has been the hardest problem to solve for reliable financial analysis using AI. ChatGPT-5 is one step closer to recognizing when it has reliable outputs vs when it is making up data.
I tested out creating a detailed comparison of Adobe ($ADBE) using SaaS specific valuation + operating metrics.
I used to take several hours doing this analysis by hand in Excel. Now we don't even need Cursor and can do this one-shot using a single prompt on the ChatGPT front end.
Role:
You are an expert SaaS equity analyst with deep knowledge of comparable company analysis (comps), public and private market valuations, and SaaS metric interpretation. You have access to real-time research capabilities and advanced coding features to create clean, professional data visualizations.
Objective:
I want to conduct a comprehensive comparable company analysis (comps) for Adobe ($ADBE). Make sure you also include Adobe’s own metrics in the tables and graphs. Break the task into three stages:
Stage 1 — Determine Adobe’s Comp Set
Identify a robust peer group for Adobe, including:
Direct design SaaS competitors (e.g., Figma, Sketch, Canva, Corel).
Broader creative software companies with overlapping markets (e.g., Autodesk, Unity).
Public SaaS companies with similar ARR scale, margins, and enterprise penetration.
Provide a table with:
Company Name Ticker (if public) Business Description Rationale for Inclusion Status (Public/Private) Why Relevant as a Comp
Rules:
Explain for each why it is a relevant comp (business model, pricing, growth, market served).
Use credible, up-to-date public data for business descriptions.
If the company is private, indicate “Private” in the Status column.
Stage 2 — Identify Relevant Multiples
For SaaS comps, identify the most relevant valuation multiples for Adobe’s business model. Include:
EV / Revenue (current year, forward 12 months)
EV / Gross Profit
Rule of 40 (growth rate + FCF margin)
P/E or EV/EBITDA (since Adobe is highly profitable and these are relevant for mature comps)
SaaS-specific efficiency metrics (e.g., CAC Payback, NRR) if publicly available
Output Table:
| Company Name | EV / Revenue (CY) | EV / Revenue (NTM) | EV / Gross Profit | Rule of 40 | P/E | EV / EBITDA | CAC Payback | NRR | Data Source |
Include median, 25th percentile, and 75th percentile for each multiple at the bottom.
Clearly note “N/A” if data is unavailable.
Stage 3 — Visualize with ChatGPT-5 Python
Using ChatGPT-5’s Python coding environment, create professional, investor-grade charts:
Scatter Plot: Growth Rate vs EV/Revenue multiple (label each comp).
Bar Chart: EV/Revenue multiples sorted by size.
Bubble Chart: Revenue scale vs EV/Revenue vs Growth rate (bubble size = revenue).
Chart Requirements:
Minimalist, professional design
Labeled axes and clear titles
Investor presentation style
Exportable in PNG format
Rules & Requirements:
Use only credible public sources (e.g., SEC filings, investor decks, financial databases).
If private company data is missing, mark as “N/A” and do not estimate unless from a credible industry report.
Cite sources in footnotes under each table.
Output tables first, then charts.
I was very impressed by the results. From spot checking, the data seemed fairly accurate and ChatGPT-5 fed me where it sourced the data from.
It also gave both qualitative and quantitative outputs (ie, explaining where the business overlaps with other companies + why it's included in the comps set).
We even got a few nice plots. I am playing around with more finance Python examples with the new ChatGPT-5 now that it recognizes the correct packages to use.
Quite powerful stuff! Previously this type of AI analysis required Cursor vibe coding with manual instructions on what packages to use. I'm working on mapping out for you even more use cases we can tackle together using just the new GPT-5 front end...stay tuned.