Three months ago, I wrote a newsletter telling you to go ecosystem-first with your AI tools.
Copilot if you are on Microsoft, Gemini if you are on Google. And then specialist tools (like ChatGPT or Claude) only for the 20% of advanced work.
I still stand by that advice (which you can read here).
But, last week I looked at our internal tool usage data and something jumped out at me.
Inside my own team, our Claude usage has gone through the roof. We went from using it for only a few tasks to it being open on everyone's screen, all day.
And I am getting asked every week this same question:
"Should I switch from ChatGPT to Claude?"
So let me help you with this decision.
Quick Comparison – March 2026
Before we go deeper, here is the snapshot.
|
|
Both tools are really good.
At the 90-95% quality range, they perform similarly on most tasks. So, the question is not "which is better" it is "what does my team need?"
And for me, only 3 things that matter.
1. Cowork
|
|
This is where Claude has pulled ahead in a way I did not expect.
Cowork is a desktop agent that works with real files, on your computer. You describe what you want done, and Claude does it. Reading your documents, creating new ones, organizing your folders. A lot of cool stuff!
I showed this recently in a masterclass with a bank reconciliation use case. I gave Cowork access to files and folders and it:
- Organized the messy invoices
- Read every invoice and extracted amounts, dates, vendors, invoice numbers
- Matched each invoice against the bank statement transactions
Was it 100% accurate? No. More like 70-80%. But that is 70-80% of the matching done automatically, and the output tells you exactly where to focus your time.
If there’s 1 thing you do right now, is to get a Claude license so you can experiment with this (without sensitive data) as it’s super powerful and the main differentiator against ChatGPT right now.
One more thing. If you are a Microsoft 365 shop, you should also know that Copilot Cowork now exists. It is basically the same concept just within Copilot instead of Claude. Just be aware it is cloud based, so won’t run with the local files on your computer (although this is likely more secure).
2. Skills
This is the feature that I think most people do not understand yet.
It is a set of instructions that tells Claude how you do a specific type of work. Your process, your format, your checks, your output requirements. You build it once, and Claude activates it automatically whenever it detects a relevant task.
Let me give you two use cases from our community.
Financial Modeling Skill: One of our AI Finance Club instructors, Anna Tiomina, created a custom skill called "Financial Modeling." It took about an hour of back-and-forth with Claude to define. She described her exact process for building models for different client types (SaaS companies vs construction vs manufacturing). Now, when she needs a new model, the skill activates. Claude already knows how to structure the assumptions tab, how to link the three statements, and how to format the output. That saves her hours of setup time on every new engagement.
Variance Analysis Skill: We also built a skill for monthly variance analysis. It defines the sequence and start with top-line verification, then data integrity checks, then KPI calculation, then commentary drafting. When you upload a new month's actuals file, Claude runs through the entire process in order and produces a draft commentary with the right structure. You review, tweak, and send.
The way skills work is smart.
Claude scans your available skills, finds which ones match your current task, and loads only the relevant ones.
Plus you can run multiple skills in the same conversation. So if you are doing a monthly close and have skills for both reconciliation and variance commentary, you can activate both skills as part of one workflow.
ChatGPT has Custom GPTs, which serve a similar purpose. But there is a difference: Custom GPTs are separate chat instances you have to navigate to. Claude skills live inside every conversation and activate on their own when they are relevant.
You build them once and they follow you everywhere. In the standard Claude Chat, in Projects, in Cowork, and even in Claude Code (but this is for another day)
For finance teams, this is the closest thing I have seen to standardizing how your team uses AI.
You define the process once, then everyone runs the same skill, and the quality stays consistent.
3. Claude in PowerPoint
|
|
Claude in PowerPoint is quite new. It reads your slide master, layouts, fonts, and colors, and generates native editable slides that match your template.
This matters because every other AI tool I have tried creates slides that look really bad.
Here is an example.
Every week, I prepare slides for our internal AI learning session. Last Thursday, I was running late. So I opened Claude on my phone while brushing my teeth, gave it some context, and asked it to build a 6-slide deck using our company template, and have it ready by the time I sat down.
Eight minutes. The slides needed a few tweaks, but the structure, content, and formatting was maybe 80% there.
That is 30-40 minutes of slide work I did not do. While brushing my teeth!
For finance specifically, the use cases we are seeing from our community are:
- Board decks – Upload your monthly data and ask Claude to build slides with the key trends, variances, risks, and recommendations already structured
- Investor presentations – Generate slides from financial analysis that Claude already did in the same conversation, so the data flows straight through
- Monthly reporting packs – Create a consistent set of slides each month with updated numbers, and because Claude reads your template, the formatting matches every time
In one of my recent masterclasses on Claude in Excel, I also showed Caude in PowerPoint – we had some really great feedback.
|
|
The decks Claude produces are better structured and look a lot better than what most other tools give you.
It’s not perfect, but for getting a good first draft that matches your branding in under 10 minutes? Nothing else I have tested is as good.
So Which One Should You Use?
Same answer I gave at our February masterclass when someone asked directly.
It depends on your setup.
But let me be more specific than this:
A) If you are not using any AI tool right now and you’re not locked into Google or Microsoft workspace?
Try Claude. With Cowork and Skills it is one of the most powerful tools available.
B) If your company runs on Microsoft 365?
Go with Copilot. Copilot Cowork (although still in early stages) has similar capability to Claude Cowork.
C) If your company runs on Google Workspace?
You already have Gemini. Use it.
D) If you already have an AI tool but you are only using basic chat?
Stop. Before you even think about switching to Claude, go learn the features you are paying for and not using.
I am talking about Custom GPTs, Copilot Agent, Gems, Projects in ChatGPT, Notebook in Copilot, NotebookLM in Google, ChatGPT in Excel, Edit with Copilot in Excel, Data Analyst in Google Colab for Gemini, Deep Research – You probably have access to powerful features you have never used, so start there.
E) If you are already and advanced AI user and want to see what Claude can do?
Try it for one month. But do not just stay in Claude Chat.
Use Cowork. Use Claude for Excel. Use Claude for PowerPoint. Build AI-powered Artifacts and create Skills.
And remember what is super important.
Do not fall into the trap I warned about three months ago. Analysis paralysis.
Do not spend three weeks debating which tool to use instead of using either one.
Experiment, but do not let it become a distraction.
Actually using (any) AI is the most important thing.
Best,
Your AI Finance Expert,
Nicolas
P.S. – Are you using ChatGPT, Claude, or both? Hit reply and tell me your setup – I am really curious how finance teams are splitting their usage right now (I read all replies).
P.P.S. – First time hearing Cowork? Don't worry, I've just got the tutorial for you! You can check out one of my latest videos on "How to Use Claude Cowork For Finance Like a PRO"
link