For the last few weeks, I’ve been testing Microsoft Copilot and Gemini alongside Claude and ChatGPT, and…I’ve been less than impressed with them.

Then I started using them differently from the way I’d use Claude and ChatGPT and started to be surprised.

To be fair, the use cases I’m finding for these two are pretty limited in comparison to the others, but they’re standing out in their niches.

I had a large dataset that I was analyzing, and before going the route of pivot tables, I thought I’d test Copilot’s abilities to analyze the data and generate charts.

Given my experiences prior to this, I wasn’t holding my breath.

So, imagine my surprise when the chart I requested was created, and the math worked out correctly! I then needed to have the colors updated to match a brand’s look and feel, so I uploaded a chart with their look and feel and…it matched it perfectly.

WUT?!

To keep testing, I asked for more analysis of another few questions, and it came back with why the request wasn’t possible (I was asking for too many answer options that hadn’t been presented to the people taking the survey).

So - no made-up answer to a query it couldn’t complete. Just a “can’t do that.”

Refreshing, no?

Just for fun, I decided to try out a little TURF analysis - and it generated it. I checked against an analysts’ version and the top 2 were the same, just swapped. But the fact the top 10 items were the same impressed me a LOT.

That said, advanced analysis would still need someone to review for correctness until I felt like it was actually capable of consistent good answers.

I’ve tried Copilot in PowerPoint and been underwhelmed by the experience. It’s earning its place in Excel. Could Claude do the same, maybe a bit better? Possibly, but if you have access to Copilot and not Claude, especially for doing data analysis inside an Excel file, I think it’s pretty good!

I know I’m a bit late to the party on this one, but the Notebook IS actually really good at analysis of a specified dataset. I tried it on a big group of papers I’d found and wanted to query. I can see why people turn to it for this. The analysis is contained to the data inside the project space, reducing the chances the system is going to make up answers like Gemini did when creating a draft report on some synthetic data I’d given it last week. It felt like a mini-web search, where I got to define what was a valid resource.

That’s the nice part for researchers.

For freelancers and small business owners, there’s another way I used it that really surprised me.

I have a Google Workspace Business Standard account, which means I have Gemini Pro as part of my subscription (and a default to not sharing my conversations to train their models). I needed to take all of my invoices and expenses and create my own balance sheet from the bookkeeping entries I’d compiled. By bookkeeping entries, I really just mean copying from my business credit card and invoices and putting it all into a running spreadsheet. I decided to try having Gemini create a balance sheet and a profit and loss statement based on all of the data in the spreadsheet. It created the template, showed me what the steps were it would take before taking them, and then generated both while I went and did other work.

I stared at the results for awhile. It was actually helpful!

I’ve said in my other newsletter that insights professionals are increasingly being called on to know what tools to use for which use cases. I’m starting to see where Gemini and Copilot fit into the suite.

Gemini: Notebook for analysis of written works, including anonymized transcripts; Gemini in Google Sheets for help with turning somewhat chaotic business information into usable templated forms needed to keep your business records clean.

Microsoft Copilot: Excel support for analysis.

Limited, yes, but I’m hoping I’ll continue to find where these tools perform well, both as a business owner and a market researcher.