Here's something almost nobody knows about ChatGPT.
There's a mode that most finance pros have never turned on. And it's the difference between getting a basic description of your file and getting a full consultant-level analysis you couldn't build yourself.
When I show this in training sessions, people ask: "Wait, ChatGPT can do that?"
It can run multi-step analysis, check your data consistency, spot patterns in your actuals that you'd miss manually, and deliver prescriptive recommendations. All while showing you exactly how it got there.
I'm talking about Agent Mode.
Before we published our AI Finance Club article on Agent Mode a few months ago, we asked members: "Have you tried using Agent Mode yet?"
75% said no.
This is understandable, because it's hidden in the Tools menu (plus people don’t know the use cases.
But, if you've never activated it, you have no idea what you're missing.
So let me show you what it actually does.
Your Own AI Virtual Assistant
|
|
Agent Mode in ChatGPT doesn’t just make it think harder (like we discussed in my newsletter here), it completely changes the way ChatGPT carries out a task.
Regular ChatGPT: You do the work with the help of AI.
Agent Mode: AI does the work whilst you supervise it.
It uses a virtual computer with a browser, terminal, file system, and connectors. So it can:
- Navigate websites and extract data
- Run code to transform your files
- Build Excel models with formulas you can audit
- Create presentations from your spreadsheets
- Execute multi-step workflows while you stay in control
Christian Martinez, one of our AI Finance Club experts, used it to build a full DCF valuation model for an energy company – complete with sensitivity analysis, Excel files, and a board-ready presentation – in under 5 minutes.
Here’re some other use cases he covers.
- Adding variance commentary directly into Excel
- Building weekly KPI packs with charts
- Bank reconciliations with proposed adjustments
- Vendor pricing benchmarks with negotiation briefs
- Budget variance analysis by department
ChatGPT Agent Mode can help you work at a level that you’ve not been able to before.
Here’s how.
How to Use Agent Mode (Step by Step)
I'm going to show you the pattern, then give you some example copy/paste prompts you can try.
IMPORTANT – Only upload sensitive data to ChatGPT Business or Enterprise. Do not do this with consumer versions.
Step 1: Know When to Use Agent Mode
Use Agent Mode when:
- You need to BUILD something (a model, a report, a consolidated file)
- The task has multiple steps (extract → transform → analyze → present)
- You're working with messy data that needs cleaning
Don't use Agent Mode when:
- You just need advice or brainstorming
- You want a quick answer to a question
- The task is simple enough for regular ChatGPT
From the YouTube video, I showed how regular ChatGPT with thinking mode handles financial analysis. But when that doesn't work—when you need actual file transformation, consolidation, or automation—that's when you switch to Agent Mode.
Step 2: Activate Agent Mode
Open any ChatGPT chat. Click Tools → Agent mode (or type /agent).
|
|
You'll see it turn blue.
Step 3: Write Your Prompt Like You're Delegating
Don't just write: "Analyze this file."
Instead, write like you're briefing a junior analyst:
Bad prompt:
|
|
File used: License Analysis
Good prompt:
With the good prompt, you gave:
- Context (who you are, what you need)
- The problem (different formats)
- The desired output (specific columns)
Let me give you some inspiration for prompts you can use with ChatGPT Agent.
Step 4: Real Finance Use Cases (Copy/Paste These)
Don’t forget to add context to the below as per the good prompt example above.
Use Case 1: Add Variance Commentary to Excel
Upload your actuals vs budget file, then use this prompt:
What you get: Your file back with commentary added directly to the sheet. Review it, then adjust it as needed.
Use Case 2: Weekly KPI Pack
Upload your monthly data file (CSV or Excel), then use:
What you get: An updated Excel file and a 3-slide deck that you can review and tweak.
Use Case 3: Budget Variance by Department
What you get: A full variance analysis with notes ready for your management review.
Use Case 4: Bank vs GL Reconciliation
What you get: A reconciliation file with proposed adjustments you can review and post.
Step 5: Review Like a CFO
When Agent Mode returns your output, don't just accept it. Review it like you're checking a junior's work:
- Spot-check the numbers (does the math work?)
- Verify the logic (does the analysis make sense?)
- Check the sources (if it pulled data, is it correct?)
- Iterate if needed ("Great, now tighten the commentary and add a chart")
This is what’s good about Agent Mode, you can see the work it’s done, which makes it much easier to audit.
Plus, when you use Agent Mode, this means that you are spending your time on judgment:
- "Is this variance real or timing?"
- "Should we flag this to leadership?"
- "What's the decision here?"
Not: "Did I copy this formula correctly across 47 rows?"
The one thing to remember
Agent Mode is not about doing work faster. It’s about being able to do work that would have been too hard to do by yourself before.
Now, in 15 minutes, you can do this work. And you understand every number because you directed the work whilst the AI was doing it.
So here's my question: what's one workflow you do every week that you could delegate to Agent Mode?
Take 30 minutes this week. Pick one task and try it.
You’ll be amazed by what you see.
Best,
Your AI Finance Expert,
Nicolas
P.S. – What did you think of this approach? Hit reply and let me know if you're planning to try this for your team (I read all replies).
P.P.S. – I cover this in my YouTube Video ‘3 Practical Ways to Use AI as a Finance Pro in 2026. For this and more tips. Watch it now!
|
|