How to save time by building an automation for yourself
Published: . Last updated:

Automation does not have to be big and expensive. Often it is a small helper that takes over one repetitive task: collecting data, sorting emails, preparing a report. Here is how to tell what is worth automating, where to start and when it is better to call a specialist
What personal automation is and why bother
It is a small helper for one task of yours: a script that moves data from emails into a spreadsheet, a bot that sends reminders, a formula that builds a report on its own. It has a real strength: you know your task better than any developer, and the AI turns your description into a working tool, so you do not have to write the code yourself.
It used to take a developer and a quote. Now you can build a trial tool in an evening and test it on real data straight away.
Why “one prompt and it all works” stays a myth
An AI does not know which of your emails turn out strange, what to do when a file has an empty row and who is to blame if the tool quietly deletes something important. All of this comes out only in practice, after long testing and patient refinement.
The better you understand your process and the more carefully you check the result, the more reliable the tool becomes. Experience matters here more than clever wording in the prompt.
What AI gives you and what a specialist gives you
AI wins on speed and price. You get a working version quickly, can try different approaches and do not have to explain the process to an outsider. For personal tasks this is an excellent route.
A specialist wins where reliability matters. They think through unusual cases in advance, handle data carefully, connect email, spreadsheets, payments and a CRM, set up backups and answer for it all working when you are not around.
Which tasks can be automated
People usually start with tasks like these:
- collecting data from several sources into one spreadsheet;
- sorting incoming emails and files into folders;
- preparing a weekly report from the same export;
- sending reminders and notifications;
- renaming, converting or processing a batch of documents.
A simple sign of a suitable task: you catch yourself thinking “not this again”. What fits is something that repeats, is done in the same steps and has a result that is easy to check.
Where to start: one task and a written list of steps
Do not automate everything at once. Pick one job that repeats, is done the same way each time and gives a clear result. If you are unsure, spend a week noting what your time goes on.
Before asking for anything, do the task by hand once and write down the steps: what you receive, what you do and what comes out. This record becomes a ready description for the AI and also a checklist against which you will later check the result.
A weak and a strong prompt
A weak prompt: “Automate my reports for me”.
A strong prompt (an illustrative example): “Every Friday I build a weekly summary spreadsheet from clients’ emails. Each email has an Excel table attached with the same columns. Here are my steps: (list). Here is a sample input file and a sample of the final table. Write a Google Sheets script that does this and explain in simple words what it does at each step. I am not a programmer. If a file of a different kind turns up, do not guess, show an error”.
The strong prompt has a task, steps, examples of input and result and a rule for unusual cases. So the model guesses less and you have less to correct.
What to watch for along the way
A few habits that noticeably improve the result:
- Pick the tool where you already work. If you live in spreadsheets, make it a spreadsheet script. If you live in email, make it an email rule. There is no need to learn a new environment.
- Order first, program second. If the task looks different every time, first put the process itself in order. Automating a poorly thought-out process gives the same mistakes, only faster.
- Test on more than one example. Take a typical case, an edge case (an empty field), a strange one (an extra column) and a case where nothing should happen. One successful run proves nothing.
- A copy before changing anything. If the tool changes or deletes files, it should save a copy first.
- No passwords in the chat. Do not paste passwords, access keys or clients’ personal data into an AI.
- Ask for explanations. Have the model say in plain words what each step does. That way you will notice if the tool does the wrong thing.
The overall plan
Put together, it looks like this:
- Pick one repetitive task.
- Do it by hand and write down the steps.
- Check whether a ready solution already exists in the services you use.
- Ask the AI to build a small tool from your record.
- Test it on five different examples: what you fed in, what you expected, what you got.
- For the first weeks, look through the results yourself.
- After a month compare two numbers: how much time it took before and how much now, counting the time spent on repairs.
Can you do it yourself
You can, and for personal tasks it is a good path. If you understand your work well and are used to checking everything, AI will be a real helper. Anyone can learn this, just like any other craft.
But it takes time and attention, often more than it seems. A tool breaks when the format of an email changes, a service updates or an unusual file arrives. If you are a specialist in another field, you will subtract those hours from your main work, and the attempt to save money will turn into a loss.
It is worth turning to a specialist if:
- the tool is used not only by you but by a team or clients;
- money, clients’ data or important documents pass through it;
- it has to run every day without your involvement;
- a mistake would be costly.
In such cases you need a reliable system: thought through, tested, with backups and a clear way to repair it. I use AI as a tool for speed, but I answer for the result myself.
Got a repetitive task that eats your time?
Tell me what you do now and how. I will reply with what can be automated and how long it takes

