Posted by |
Getting your Trinity Audio player ready...
|
Apple Podcasts ¦ Amazon Music ¦ Spotify
Hi, and welcome to Try AI for Growth, a podcast out of Make Space for Growth. Here, I share short, and sometimes surprising stories of how I’ve used AI to tackle real challenges at work and at home.
I’m your host, Sara Vicente Barreto, and today I want to tell you about the AI tool that is changing everything for me. Many AI tools change how you work. But now and then, you come across one that makes you question something deeper… What will be of certain jobs? Even mine, when I first started…
Today’s episode is about one of those tools. It’s Claude for Excel — an add-in that brings AI directly into your spreadsheet. It’s incredibly powerful from a productivity perspective… but it is more than that. It reshapes how you approach work. How do you design entry-level jobs? And how many entry-level jobs will be there?
If I had this tool when I started my career as an analyst in banking… I would have certainly slept a few more hours each night. But would I have even had my job? What would it have been like?
I know this podcast is more about applicability and less about reflections. But I can’t help it, because Claude for Excel has my head spinning at the moment.
What is Claude for Excel
At a simple level, Claude for Excel is an add-in. It brings AI directly into your spreadsheet. But the key difference is not the technology — it’s the positioning. The AI is not separate from your work, where you have to go somewhere to ask questions and come back to your workflow to execute. Claude for Excel is inside your work. If you work with Excel, that is.
So instead of:
- Copying or uploading data into an AI tool
- Asking a question or debugging a problem
- Bringing the answer back to solve it in your work environment
You’re doing everything in one place. And the AI is seeing what you are seeing. You don’t need to explain. It can study the spreadsheet, see how it is connected, what is input, and what is output. It can debug with you, suggest improvements, check data and, the whole other step, do it for you. What do I mean by that? It does your formulas and executes changes, all while you are sipping your coffee.
And that’s where things start to shift.
Dealing with Spreadsheets
Let me start with something very familiar. Fixing a buggy spreadsheet. The kind of work that you procrastinate on, but when it hits you, it crashes your spreadsheet when you need it. I picked up one of my crucial accounting files and asked Claude to do a full review:
- Checking formulas
- Alerting me to errors or inconsistencies
- Do a full pivot table review to assess duplications and necessary clean-ups
- Explain the sources of errors and how they can be fixed.
This would be classic analyst work. Painful and time-consuming. At times, one would wonder if even worth it. But it’s also how you learn. It’s how you build intuition.
Working through a large file with multiple tabs, pivots, formulas, and data sources with Claude was a whole different experience. It was as if it was sitting next to me and we were discussing:
- “What’s wrong with this formula?”
- “Why is this slow?”
- “Can you fix this pivot table?”
And it was not just me asking the questions. Claude asks you questions, often in multiple-choice format, which makes it very clear. It clarifies intent, it gives you error alerts, but also things that could be improved, even if not urgent, without you asking for it. And in doing so, it compresses what used to be hours of learning into minutes of interaction. Those hours that used to be my job…
“I have a mistake in the model. I will work through the night to find it and fix it.”
Not only was Claude able to identify problems and solutions, but then execute it for me once I told him to do so.
It could rewrite formulas, clean up structure, and rebuild elements. Always, with my permission. Whilst you can answer the question “can Claude delete data on tab X” with “always”, I have always been saying “once” as I get to know the tool and also see where it can fail me. I am cautious to say it has not failed me yet, but I am watching out for it.
More than fixing… building!
I spent a few days dizzy from how I was experimenting with Claude. It was likely the closest I felt to trying an addictive drug or having a shiny new toy. I stopped those who were willing to listen to come to my Excel and watch. Yes, I am that much of a geek. But I hang out with geeks too.
My project analyst at the charity reminded me we had to build our 2026 dashboard for local reporting. Because I had entirely rebuilt our 2026 file, which was not embedded in it yet (clearly, I did not rebuild it with Claude, but no point crying over that now). And as I thought on how I was going to find the hours to do the detailed dashboard, I wondered if Claude could do the heavy lifting.
This was not about fixing, identifying, or solving. This was about building from scratch. In my normal process, I would have to:
- Clean up and structure the data
- Writing formulas that are flexible to my different usages
- Testing outputs and built-in checks
- Iterating a few times for accuracy
Here is what I did instead.
- I outlined the headers from last year’s dashboard
- I outlined the structure of the rows from this year’s new activity structure
- I told Claude where the data came from
- And I explained what formulas I thought were best and where I needed flexibility to toggle.
In dismal curiosity, I asked Claude to try it out for the first set of rows. As I watched the formulas unfold, I almost hoped they would be wrong. But deep down, I knew they would not be.
What stood out wasn’t just that it worked. It was how it worked. Once the first attempt was done, Claude did the data checking, which meant it checked my data classifications. The errors found were mostly human and allowed me to fix something now that would be a cumulative error over the next few months. It helped me:
- Clean the data
- Automate classifications where possible
- Validate assumptions and reasons for the differences
With all this, my dashboard was complete. All that was missing was to replicate it to break it down into 2 separate regions. By then, I had run out of tokens. As it was not urgent, I went on with my work elsewhere and, in the evening, whilst I was brushing my teeth, I remembered I probably had the tokens back. I repasted my original ask, and there were the 2 new tabs I needed. Fully formatted, fully consistent and fully checked.
While I brushed my teeth.
How Do You Learn Without Doing?
This is the part I keep coming back to. Because both of these examples highlight something important. A lot of entry-level work is not just about output. It’s about exploring, discovering, and learning.
- Learning how data behaves
- Learning how formulas break
- Learning how systems fit together
But if AI can fix the formulas, build dashboards, understand and structure data, what happens to the learning process? If I think back to my time in banking as an analyst, a big part of my development came from:
- Getting things wrong
- Taking time to fix them
- Slowly building understanding
With a tool like this, that process is compressed dramatically, which is great for efficiency. But raises a real question:
“Where does that foundational knowledge come from now?”
Time will give us the answer. As someone who has learned how to do it, I don’t have blind conversations with Claude. I read through all the steps that it is going through, I discuss where I have doubts on the approach, and I prompt it to challenge me more. But what if I had never been through this sort of problem-solving before? How will this learning be developed in the future? Will we even need it?
Banking without Claude
I can’t help but think. If I had Claude for Excel back then… the hours I would have slept! I would have been faster. No question. Probably more productive. But I do question:
- Would I have understood Excel as a tool at the same level?
- Would I have developed the same analytical instincts?
- Would I be resilient in problem-solving?
And that’s not necessarily a negative. But it does mean the role itself would have been different. Less about doing the work. More about directing the work. Without having ever done it before.
But let’s break with the philosophy and stay on the practical advantages.
Data Stays With You
By the way, there is one thing I love about Claude for Excel that was still breaking my workflow a lot and was preventing me from using Claude or ChatGPT for analysis as much as I wanted. Data seems to be handled differently, or at least that is what it says when I researched it.
You’re not uploading entire datasets into an external tool. Claude works with what you bring into the interaction. If you write it on the Claude chat, this is going into Claude, but the data in the spreadsheet is not being shared. So, for sensitive information, this feels much more manageable.
And much easier to integrate into business applications.
Lessons Learnt
Let me share a few takeaways from this experience
1. Start with structure, not output: If you are building, be clear about what you want to build before asking it to build. It can totally build you a dashboard without instructions, but if you have guidelines, use them early, especially if you are low on tokens.
2. Let it challenge your data: Some of the best value comes from it questioning your inputs and checking your data. In one of my files, I introduced cross-checks with all other tabs, all done by Claude. In case of error, it is easy to get alerted.
3. Use it to learn, not just to execute: This is a big one for me. Don’t just accept the output, understand how it is doing it and how it is working with the data. That way, you will keep the critical thinking on.
Can you try it?
If you are wondering how to start, maybe you are not spending enough time in Excel. Claude for Excel requires a Pro subscription to Claude, which I did with this sole purpose. I am not letting go of it any time soon.
As you are working through analysis, data validation, debugging files, creating outputs, and formatting spreadsheets mindlessly, remember there is a new analyst in town. It is called Claude.
Claude for Excel is an incredibly powerful tool. But what makes it interesting isn’t just what it can do. It’s what it replaces. And more importantly… what it changes about how people learn and develop in their early careers. Because if entry-level work is no longer about execution, then it has to be about something else. And we need to figure out what that is soon.
Thanks for joining me on this episode of Try AI for Growth. If you try something similar, I’d genuinely love to hear how it goes. If you can, share this episode with someone who spends their life in Excel. The more you share, the more we will all learn!
Until next time—keep experimenting and keep having fun.