Why not just chat · 2026

Still pasting chat answers into your files and fixing them by hand? That's a 2020 workflow.

There's one difference most people miss: a chat only talks, a real assistant also does. This page is for anyone non-technical still doing everything by hand, and it explains, in plain words, why the old loop quietly eats your hours, and the safe way out that breaks nothing.

At the end of a long day you ask the chat how to merge two tables, or reword a document, or clean up a whole folder. It gives a nice answer, and then what? Then you still do all the work yourself: open the file, copy, paste, try it, find it doesn't quite fit, go back and ask again, and around it goes. The chat talks, you execute. Without noticing, you became the tool's manual labor.

And it isn't because you're "not technical." It's a habit, left over from a time when tools really did only speak in a programmer's language. That habit eats your hours quietly, and you don't need it anymore.

A chat only talks. A real assistant also does.

The tool I'm talking about is a different generation entirely. It doesn't only talk, it opens your file itself, reads it, carries out the task end to end on a copy, and shows you every step. It's the difference between someone telling you in words how to assemble a cabinet, and someone who comes over and assembles it in front of you. It isn't a smarter chat. It's an assistant that does the work.

It's also why a plain chat once gave you wrong numbers: it never saw your file, it guessed from what sounded reasonable. This assistant works the other way around, it reads the real data and works on it, checks itself, and flags every mismatch, including the mistakes your eye misses at the end of a long day.

This assistant already exists, and for non-technical people too

It's Claude Code, Anthropic's official AI agent that reads and edits your real files and carries out whole tasks. Until now it was seen as a tool for programmers. Our free suite installs and opens it for you in one click, no code editor and no commands, so you get the real tool as well.

No more "I don't know what it's good for"

Here's exactly what to ask it. This is your way out of the loop.

Each of these runs on a copy, the original stays clean until you're happy. The examples come from the world of files and data, and the same idea works on documents, folders, and lists too.

Merge tables from different sources into one clean table

Files in different shapes, different headers, separate tabs. It all becomes one orderly table, by the rules you set.

Example request"Take the tabs from all the branches and merge them into one table, one row per customer."
Check data quality and produce a report

It goes over the table and flags what's missing, what doesn't match, what looks off, and hands back a color-coded check report instead of you hunting for mistakes by eye.

Example request"Go through the list and flag every row with no phone number, and write 'phone missing' in red."
Clean up and remove duplicates

Extra spaces, the same name in two forms, the same row twice. It tidies and merges by clear rules you approve up front.

Example request"Merge duplicates of the same customer, and keep the record with the most details."
Build a recurring report, and re-run it whenever you want

The monthly, weekly, quarterly report. You explain once how it looks, then get it ready, again and again, in exactly the same steps.

Example request"Build my monthly sales report, filtered by month, summed by branch, sorted newest to oldest."
Compare two versions of a file and show what changed

There's no "compare" button in a spreadsheet. It finds what changed between the old version and the new one, and shows you clearly.

Example request"Compare last month's file to this month's, and tell me which rows changed and what changed in them."
Calculate, summarize and sort, without formulas

Totals, groupings, sorts, filters. You say what you want to see, it builds the result, with no formula to remember and none to break by accident.

Example request"Total the orders per customer for me, and sort from highest to lowest."
Turn a recurring task into a permanent tool

Once something worked nicely one time, you save it as a ready-made tool. Next time you just say "do it again on the new file", and it remembers all the rules.

Example request"Save all these steps, so next month we do it in a single request."

The exact same task, before and after.

✗ Alone, by hand
  • Open ten tabs, copy and paste row by row.
  • Hunt for duplicates by eye, and miss half.
  • Rebuild the same report every month, from scratch.
  • Afraid to touch the original file, because there's no backup.
  • A formula breaks? An hour of detective work to find out why.
✓ With an assistant that does the work
  • Explain in words what's needed, and it merges into one clean table.
  • Duplicates are flagged and merged by rules you set.
  • The report is built once, then comes back in one click every month.
  • Everything runs on a copy, the original stays safe, with a backup before each change.
  • At the end you get a color-coded check report showing exactly what came out.
And the voices in your head? Let's take each one seriously

Every hesitation is reasonable. Let's see what's really behind it.

I won't laugh at a single one of these thoughts. Each protects you from something real, but along the way it locks you out of what's already possible today.

The confidence one
I'm not a technical person, it's probably too complicated for me

Notice what you actually have to do here: nothing technical. No code, no formula. You explain in plain words what you want to happen, exactly as you'd explain to a colleague sitting next to you: "merge these files for me", "flag the duplicates", "build my monthly report". The technical part is precisely the part the assistant does for you.

What's really going on here: "I'm not technical" is almost never a fact. It's a belief that took root back when tools spoke only to programmers. This one speaks your language, and the technical side is its headache, not yours.
The fear of breaking it
I'll break the file my livelihood depends on

Here's exactly how it works, so that fear can settle: the assistant never touches the original, it works on a copy. The original stays closed and safe to the side. Even on the copy, every step that changes something stops and waits for your approval, and you see what's about to happen before it happens. A backup is saved before each change, so there's always somewhere to go back to.

What's really going on here: that fear is healthy, and it's exactly what kept you safe until now. The difference is that now there's a real safety net: a copy, approval on every step, and a backup. The original is never at risk, at any point.
The time one
I don't have time to learn another tool right now

Let's do the math quietly. That manual work, the merging, the cleaning, the sorting, the re-typing, the recurring task, takes you somewhere between one and two hours almost every day. Call it one hour, to be conservative. That's five hours a week, more than twenty a month, gone quietly. And the cost of learning the new way? One afternoon. Not a course, not months. One afternoon, after which you start winning those hours back.

What's really going on here: "I have no time to learn" translates, in practice, to "I have no time to stop wasting time." The free moment won't arrive on its own, because the thing eating your time is exactly the thing you haven't learned to replace yet.
The privacy one
I don't give AI access to my data

The right question to ask, and no overblown promises. To do the work, some of the information the tool reads may go to Anthropic's servers, the company behind it, so you decide up front how sensitive the data is and handle it accordingly, by your organization's policy and the law. You decide what the tool even sees: most day-to-day work is ordinary business data, and for especially sensitive information you can hide names and ID numbers and let it work on the structure and the numbers only.

What's really going on here: it isn't "give AI the data" or not, it's how much, how, and at what sensitivity. You decide what goes in, and you work within your organization's rules and the law. The decision is in your hands.
The way in, safely

You don't jump in. You climb down a ladder. One week.

Real change doesn't work with "everything's different from today." It works with small steps, where each one is completely safe and you can step back at any moment. Here's how it looks, with nothing technical:

Day 1 · Just look, change nothingThe first step, zero risk

Open the tool on a copy of a messy file, and make one simple request: "Look at this file and describe in words what's in it, where the duplicates and gaps are. Don't change anything." Within fifteen minutes you have an orderly description of the mess, without touching a thing.

Days 2 to 4 · One small, real task, on a copySmall, contained, controlled

Give it a small, contained task on a copy, something like "flag in red every row with no phone number", or "merge these two lists for me". It works, and stops before each step for your approval. You look, approve, or say "no, like this". Still on a copy, still fully reversible.

Days 5 to 7 · Turn the recurring task into one clickWhat took a day, comes back on request

Take that thing you rebuild from scratch every month, and explain it to it once: what to filter, how to summarize, how to sort. Once it works one time, next month you just say "do it again on the new file". What took you a day, comes back in one request.

The whole way, the safety net is in place: everything runs on a copy, the original is never touched. Every change is reversible, there's always somewhere to go back to. Every step stops and waits for your approval, and you're the one who decides what the tool even sees. If on day two you decide it's not for you, you close the copy and go home. You lost an afternoon. And in the other direction, all the hours you'll stop wasting are waiting for you.
A real case

Three days of work that became hours.

A client received a huge file assembled from different sources: dozens of tabs, each built a little differently. Sometimes the header had spaces and sometimes not, sometimes the column was called "phone" and sometimes "extra phone", and sometimes a number hid inside a notes column. Out of all that mess they needed one orderly table, one row per record, ready to load into a system.

Doing it by hand is days of work, and even then full of mistakes. Instead, we explained the rules in words: one row per record, names merged with no duplicates, phone numbers gathered from the notes columns too but only where there really was a number, and where a record had no phone at all, write "phone missing" in red.

The result: one orderly table, and beside it a check report that color-codes every mismatch. And here's the important part: the tool caught exactly the mistakes the eye misses, a field placed in the wrong column, a hidden duplicate, a missing value. What would have taken days was done in hours, without touching the original file.

Dozensof tabs merged into one table
Days → hourstime spent on the project
In colora check report catching every mismatch
Someone who was exactly where you are

Meirav fought this harder than you will.

Twenty years in Excel, finance manager at a mid-size company, one of those people who hold a whole organization together on a single spreadsheet. Every month-end was the same nightmare: files from every branch, each in a slightly different shape, and out of the whole mess she had to produce one clean, merged report the management makes decisions from. She did it by hand, always by hand, three days every month. And when someone mentioned "AI", she laughed: "It's a toy, and besides, I'm not a technical person."

Then someone showed her the safe way, the exact one you just read: on a copy only, one small first request, every step stopping for her approval, everything reversible. She learned nothing technical, she just explained in words what needed to come out. A month later, the three-day merge was done in hours, with a check report beside it that caught two mistakes she herself would have missed.

"I kept thinking I'm just not a technical person.
It never occurred to me that I never had to be." And she didn't become any less of a finance expert. For the first time in years, she does only the part that actually needs her.
For decision-makers

For managers: why it pays off.

Most management teams never see it. The team burns hours every week on manual work that shows up in no report, merging tables, cleaning, re-typing, assembling reports. It's expensive time that vanishes quietly, and it's why the reflex is to hire more people instead of speeding up the ones you already have. This tool buys that time back, and it's free and open source.

⏱️ Time that comes back, every month

An employee wasting a day or two a month on merging, cleaning and preparing reports frees those hours for work that actually needs them.

✅ Fewer mistakes, less risk

Manual spreadsheet work is a source of costly errors in the reports decisions are made from. A process that checks and verifies itself lowers that risk.

🚀 Reports in hours, not days

"Get me the report" goes from days to hours. Decisions get made faster, and deadlines stop being a nightmare.

👥 The same team, several times over

When the load grows, the old way dictates another hire. Instead, the same people who already know your data deliver several times as much, with no expensive software project.

🛡️ Low risk to try

It runs on your machine, on a copy, and everything is reversible. IT controls permissions with a recommended settings file that makes sure nothing runs without approval, and you decide what information the tool sees.

💡 A capability that stays with you

Once the team learns to explain a task once, the know-how stays in-house and comes back every month, with no dependence on an outside party.

The future isn't coming to replace you. It's waiting for you to stop doing the grunt work by hand.

The suite is free and open source, and installs the full Claude Code agent in one click. It works on Windows, macOS, and Linux.