Using AI to Compare Two Spreadsheets and Build a Repeatable Report
If you run a business, you already know the drill. Once a month (or once a week, or once a day), someone has to pull two exports out of two different systems, line them up in Excel, and figure out where they do and don’t agree. VLOOKUPs, pivot tables, copy and paste. It works, but it eats time.
AI can do this for you in under a minute, and once you set it up the first time, you can repeat it forever.
Here is a short walkthrough video showing exactly how to do it, side by side in Claude and Microsoft Copilot:
Watch it here: https://www.youtube.com/watch?v=Cb-vX9yxD-c
Below is a quick summary of what the video covers, why this workflow matters, and a word of caution before you try it.
The Use Case
In the video, I compare two real exports from our MSP toolstack:
- An export from our endpoint detection and response (EDR) portal, listing every device with our EDR agent installed.
- An export from our remote monitoring and management (RMM) platform, listing every device we monitor.
The goal is simple. Merge both lists on the computer name so that every row shows matching data from each system side by side. Then add a column at the front of the spreadsheet with a green check if the EDR is installed on that device, and a red X if it isn’t.
The result is a single consolidated view that tells us at a glance where coverage is missing. Any MSP, any IT team, and honestly most operations teams doing vendor or license reconciliation can use this same pattern.
Claude vs. Copilot
I ran the exact same task in both tools.
Claude tends to take your instructions, make reasonable assumptions, and hand you a result. You can refine from there.
Copilot asks more clarifying questions up front and often produces extra tabs and breakdowns you didn’t ask for. Not wrong, just more output to sort through.
Both got to the right answer. Claude got there with less back and forth for me, which is why I lean on it more often. Copilot has the advantage of being built into the Microsoft stack, and many organizations already have licenses available.
One Very Important Warning About Your Data
Before you upload anything to any AI engine, assume that whatever you put in could become public knowledge.
That means:
- No patient data
- No proprietary information
- No personally identifiable information
- No customer lists, financials, or contracts you wouldn’t want leaked
The data I used in the video is anonymized. You can anonymize your own data before uploading, and for regulated industries like healthcare, legal, and finance you absolutely should. There are also enterprise controls available for Copilot and other platforms that add restrictions on how your data is used, but they have to be configured correctly.
If you’re not sure how to set that up, that’s exactly the kind of thing your IT provider should be helping you with.
The Trick That Makes This Worth Doing
A one-time report is useful. A repeatable report is where you actually save time.
After the AI produces the output you want, ask it to generate an instruction file. Claude will create a Markdown (.md) file. Copilot will create a YAML (.yaml) file. Either one captures every instruction you gave it, including formatting, column order, and logic.
Next month, you drop in your two new exports along with that instruction file, and the AI produces the same report again in seconds. No remembering what you typed last time. No re-explaining the column layout. Just the same output, every time.
You can extend this to more than two files, other formats (PDF, CSV), and reports from accounting systems, EMRs, ERPs, ticketing systems, and anything else that exports tabular data.
When to Ask for Help
If you’re a small or mid-sized business in healthcare, legal, finance, manufacturing, or any regulated space, and you want to use AI in day-to-day operations without putting sensitive data at risk, that’s a real project. It involves picking the right platform, configuring data controls, training your team on what to put in and what to keep out, and building repeatable workflows that actually save time.
That’s the kind of work we do at ArcLight. If you’d like to talk through how AI fits into your environment safely, call us at (918) 270-6600 or reply to this post.
Watch the full walkthrough here: https://www.youtube.com/watch?v=Cb-vX9yxD-c
Brian Largent
Father to five, husband to one, founder, CEO, and all around swell fella (or so I'm told)
