"ChatGPT or Claude?" is one of the first questions a finance professional asks when setting up AI. Comparisons online usually rank features. This one starts from usage: what 140 fractional CFOs, who use AI on client work daily, reported in the 2026 Fractional CFO Study.
The short answer: most use both, for overlapping jobs. What decides whether the numbers hold is not the assistant but where the model lives.
Which Do Finance Professionals Use More, ChatGPT or Claude?
Both, mostly. Of the 140 respondents who named their tools in October 2026:
| Tool | Share of users | Used for over a year | Started this year |
|---|---|---|---|
| Claude | 88% | 70 of 123 | 52 of 123 |
| ChatGPT | 60% | 72 of 84 | 8 of 84 |
| Gemini | 42% | 34 of 59 | 22 of 59 |
| Copilot | 34% | 25 of 47 | 16 of 47 |
Two readings:
- ChatGPT is the incumbent. 86% of its users have had it for more than a year. Very few dropped it.
- Claude is the fast riser. 42% of its users started this year, the largest wave of any tool in the survey.
A sample reached through LinkedIn and the Layerz network leans toward early adopters, so treat the levels with care. The pattern of overlap is the robust part: one assistant rarely replaces the other.
What Are They Used For in Finance?
The same jobs. Among 150 fractional CFOs, last month's AI tasks were:
- Summarising documents (77%)
- Emails or commentary (75%)
- A formula or Excel help (63%)
- Variance analysis (60%)
- Reconciling or categorising transactions (59%)
- A cash or P&L forecast (58%)
Both assistants cover this list. Practitioners describe a split by habit more than by capability: one for drafting and day-to-day questions, the other for longer analysis or work on files. Which one plays which role varies by person.
Where Do ChatGPT and Claude Differ for Finance Work?
On setup and reach, more than on reasoning. As of November 2026:
| ChatGPT | Claude | |
|---|---|---|
| Workspace per client | Projects, with instructions and files | Projects, with instructions and files |
| Repeatable procedures | Custom GPTs and skills | Skills |
| Computation on files | Data analysis (runs Python) | Code execution on uploaded files |
| Inside Excel | ChatGPT for Excel add-in | Claude for Excel add-in |
| External tools | Plugins from its directory, MCP in developer mode | Connectors, remote MCP |
| Coding agent | Codex | Claude Code |
Feature names change often: check the current plans before you commit a team. What matters for a financial model is the last two rows. Both assistants can reach external tools over MCP (Model Context Protocol), which means both can work on the same model.
What Goes Wrong With Either?
The failures the study measured are not specific to one assistant. Among 139 respondents, last month:
- 32%: explaining what they wanted took longer than doing it.
- 27%: numbers were wrong or could not be trusted.
- 24%: the AI forgot the context from one session to the next.
- 15%: it did not connect to their tools or data.
These are problems of state, not of intelligence: no persistent model, no written conventions, no trace of what changed. Switching from ChatGPT to Claude, or back, does not fix them. We describe the layer that does in The CFO Harness.
How Do You Use Both on the Same Financial Model?
Keep the model outside the assistant.
Layerz is a structured spreadsheet, ready for AI. It is an official ChatGPT plugin (also in Codex) and connects to Claude over the same MCP server. One model, its conventions (a FINANCE.md file every session reads first) and its version history serve both:
- Build in ChatGPT: ask it to create a three-statement model from a client's trial balance.
- Review in Claude: the next day, ask Claude to check the balance sheet ties and explain the cash movement. It opens the same model, not a copy.
- Export to Excel with formulas intact for the client.
A deterministic engine computes every figure, so the numbers do not depend on which assistant asked. Every change is versioned with a diff, whichever assistant made it.
So, ChatGPT or Claude for Finance?
Choose by constraint, not by brand:
- The client's rules. 26% of fractional CFOs work only with tools their client approved. If a client standardised on one, use that one on their data.
- What your team already pays for. 51% of fractional CFOs pay for their tools themselves: one subscription used well beats two used half.
- Your working style. Try both on the same real task for a week, and keep the one you reach for.
Then make the model independent of the choice, so that changing assistant next year costs nothing.
FAQ
Is ChatGPT or Claude better for finance? Neither wins across the board. In the 2026 Fractional CFO Study, 88% of fractional CFOs who named their tools use Claude and 60% use ChatGPT, most of them both. Both handle summaries, commentary, formulas and variance analysis.
Is Claude better than ChatGPT for Excel? Not on access: both have an add-in that works inside the workbook, Claude for Excel and ChatGPT for Excel. For a model that has to stay correct over months, what matters more is whether the figures are live formulas and whether changes are tracked.
Which AI do CFOs use the most? Among the 140 fractional CFOs in the 2026 study who named their tools, Claude (88%), Excel (79%), ChatGPT (60%), Google Sheets (51%), Gemini (42%) and Copilot (34%).
Can ChatGPT and Claude work on the same financial model? Yes, if the model lives outside both. A Layerz model is reachable from ChatGPT through its official plugin and from Claude over MCP.
The Takeaway
The finance practitioners who use AI the most do not pick a side. They use ChatGPT and Claude for what each does well, and keep the model, its conventions and its history in one place that both can reach.