The Personal Agent That Does the Work: What That Means

Ask a chatbot for a slide deck and you get a list of bullet points. Ask it to analyze a spreadsheet and you get instructions on which formulas to try. The advice is fine. The work is still yours.

That gap is the whole idea behind a personal agent. An agent does not stop at the explanation. It takes the file you hand it and hands a finished file back. This post walks through what that looks like in practice, with a real example: a raw spreadsheet goes in, an analysis comes out, and then a .pptx you can present.

What “agent” means here, exactly

The word agent gets used loosely. Here is the narrow definition we mean:

• It reads files. Drop in xlsx, xls, csv, docx, pptx, or txt and it works with the actual content, not a summary you paste back.
• It produces files. Ask for the analysis as a spreadsheet, or the findings as slides, and you get a real .xlsx or .pptx to download. Not text formatted to look like one.
• It remembers. Come back tomorrow and it still knows what you were working on. You do not re-explain the project every session.
• It starts without an account. Open the page, type, attach, done. No signup, no card, no ads, no tracking. The infrastructure is in Germany.

That combination, files in and files out with memory across sessions, is what separates an agent from a chat window. Chat alone is useful for thinking out loud. The agent is for finishing the task.

A real walkthrough: from raw export to finished deck

Here is the flow end to end, the same one you can run right now.

Step 1: Upload the spreadsheet. Take a raw export, say a Q3 sales sheet with columns for region, product, units, and revenue. Attach the .xlsx to the conversation. No preprocessing needed. The agent reads the sheet directly, including multiple tabs if it has them.

Step 2: Ask for the analysis. A question like “Which regions drove the quarter and where did we lose ground?” gets an answer with the math already done: totals by region, month-over-month movement, the two products that carried the quarter and the one that dragged it. If the numbers need cleaning, duplicates or blank cells, the agent handles that too and tells you what it fixed.

Step 3: Ask for the deliverable. Now the part a chatbot cannot do: “Put this into a short deck for the Monday meeting.” You get a .pptx with the headlines as slide titles, the numbers in tables, and a chart or two. Download it, open it in PowerPoint or Keynote or Slides, edit it like any other deck. The same works in the other direction: ask for the cleaned data as a .xlsx or the write-up as a .docx.

The point is not that any single step is magic. It is that the loop closes. You went from a messy export to a presentable file in one conversation, and you never opened a formula editor or a slide template.

What it will not do (yet)

Honest limits matter more than feature lists. The agent works on the files you give it. It does not connect to your email, your calendar, or Google Drive, so it cannot send that deck for you or pull the numbers straight from your account. It does not book meetings. If a task ends with “now email this to the team,” that last step is still yours.

Everything it does claim, it ships: files in and out across xlsx, csv, docx, pptx and txt, memory between sessions, images, video, music, code, and deep research in the same conversation. Eight interface languages, free to start, hosted in Germany.

Who this is for

If your week involves turning raw information into something someone else reads, a spreadsheet into a summary, notes into a document, data into a deck, an agent removes the busywork in the middle. If you only want to talk through an idea, chat is fine and always will be. The difference shows up when the deliverable is due.

FAQ

Is it really free? Yes. ngini is free to use, with no account required to start and no ads. Open the page and work.

Do I need to install anything? No. It runs in the browser. Upload a file, ask, download the result.

Can it read my spreadsheet? Yes. xlsx, xls and csv, including multi-tab workbooks. It reads the real content and does the math on it.

Does it generate real files, or just text? Real files. Request slides and you download an actual .pptx; ask for data output and you get a .xlsx. Both open in the normal office apps and stay editable.

What happens to my data? Prompts are not used for training, there is no ad tracking, and the infrastructure is hosted in Germany.

How is this different from ChatGPT free? ChatGPT free needs an account and ends on advice; ngini ends on the file. See the full comparison or start with the no-account walkthrough.

If you want one page that covers every capability in one place, the all-in-one overview lists what the agent handles, or go straight to the chat and hand it something to do.