You regularly repeat the same small task: calculate an estimate, tidy up a brief or prepare notes before a meeting. You may already have a document template. But you still have to find it, fill it in and check that nobody has missed an important detail.
What if you tried turning it into a small interactive tool? Vibe coding offers a way to explore the idea through conversation with an AI. Start with a specific need and work in stages. Here is a method you can try on a simple example.
1 Choose a task you know well
Take a fictional example: a communications team receives requests for internal posts. Some arrive by message, others in conversation. The audience, deadline or objective is often missing.
A first tool could help colleagues prepare a complete brief. Keep its scope simple: ask a few questions and produce a summary sheet. Publishing, approval and sending remain with the team.
Choose a task whose quality you can assess. If you cannot describe a good result, start by agreeing on that with the people involved.
2 Write the user journey before discussing design
Describe the actions in order. The person opens the page, enters their objective, identifies their audience, provides a deadline and gets a sheet they can review.
Add the important rules. Which fields are required? Can someone go back and change an answer? What happens if they leave the deadline blank?
You could start with this request:
Create a prototype briefing form for a communications team. It has four fields: objective, intended audience, main message and deadline. The first three are required. A button displays a printable summary sheet. The person can return to the form without losing their answers. Use fictional data only. Do not include user accounts, sending or answer storage.
This brief gives you a first result to examine. Building through successive exchanges follows a loop: describe, generate, try and refine the request. Google Cloud describes this workflow.
3 Check the result in several situations
Start with a complete brief. Then try an empty field, a very long message and a deadline in the past. Also look at the sheet on a phone and in print preview.
Record what actually happens. “The form is badly designed” does little to help you fix it. “When I go back to the previous step, the main message disappears” describes an observable problem.
Do not rely solely on what the AI says it has done. If it says answers are not stored, have that checked before entering internal information.
4 Fix one difficulty at a time
If you request a new design, three features and a connection to your existing tools at the same time, it will be harder to pinpoint the cause of a problem.
Start with the main journey. Once it works, improve the labels, then the presentation. Keep a working version before making an important change so you can go back.
A correction request can be specific: “Keep the answers when I return to the form. Then check that the summary still displays all four pieces of information.”
5 Let its future user try it
Give a colleague a task: prepare a publication request using a fictional example. Let them use the tool, then ask what was missing.
Does the form help them avoid omissions? Can the communications team use the resulting sheet? Perhaps one field is unnecessary and another needs an explanation. Those observations will guide the next steps.
Before sharing the tool for real use, arrange a technical review, decide where it will be hosted and name someone responsible for updates. If the prototype adds little value, you can also stop there: the experiment will have helped you clarify the need.
Have a repetitive task in mind? Talk to eloRezo about the first tool you could prototype.