2025 has been the year of AI everything. AI-enabled ovens are now a thing. I don’t know why but they are. AI assistants are popping up everywhere. Microsoft is rebranding everything with Copilot. They’ve even renamed the Office software suite yet again to reflect Copilot’s integration. Office is now being officially marketed as Microsoft 365 with Copilot.
Microsoft is embracing the age of the Agentic. Agentic AI refers to artificial intelligence systems that can independently plan, make decisions, and take actions to achieve specific goals. Agents are AI that can “figure it out” when given a problem. Copilot is really a series of AI agents built for specific purposes and integrated into Office products. And some of these agents are really impressive.
One of those is the Analyst agent. Consider this scenario. I leveraged the Copilot Analyst Agent to streamline the process of comparing two large files. One contained a comprehensive inventory of all managed devices, and the other a listing of devices that already had a critical software package installed. Traditionally, cross-referencing these lists would require time-consuming manual effort in Excel.
By utilizing the Analyst Agent, it took very little effort. The agent imported both files, analyzed the data sets, and performed a match between the device identifiers in each file. It then generated a clear and actionable report highlighting which managed devices from the master list did not appear in the list of devices with the software installed.
This is cool. But also, not terribly difficult to do with some Excel skills. Two lists, one common identifier between them. Tie them together and create a new list with everything. There are a few ways to achieve that with Excel. But here’s the really impressive part. Both files weren’t Excel spreadsheets. One was a nice clean Excel file. But the other was a 95 page PDF file formatted as a human readable report. It contained lists, and sub-lists, and different sections with different lists, headers, footers, completely irrelevant descriptions, et al.
The data was spread out and couldn’t be any further from flat data. The Analyst agent just parsed this PDF file for the required data. It wrote Python scripts to do the work of parsing. It would have taken the better part of a productive day to create custom tools that would parse that PDF file and probably another day debugging it. The Copilot Analyst Agent didn’t bat a digital eyelash and all I had to do was hand it both the files and tell it what I needed. I handed it two files and a few paragraphs of prompt explaining the data and requirements.
The Copilot Analyst Agent executed 84 actions, wrote 18 Python scripts, and spent 487 seconds solving the problem before handing me a new spreadsheet with the exact data I needed.