Prototyping with AI

You have just made your computer do things in seconds that used to take hours. Now we go one layer higher: instead of typing commands, you describe what you want in plain English — and AI builds it.

This is prototyping with AI tools: the fastest way from an idea to something you can show, test, and think with. No installation. No code yet. Just a conversation.

What Is a Prototype?

A prototype is a rough, working version of an idea — built quickly to answer the question "does this actually make sense?" before you invest real time.

In a CA practice, a prototype might be:

• A draft email template for client reminders
• A sample financial summary your client could read
• A layout for a simple compliance tracker
• A script that describes the logic of a calculation

The goal is not perfection. The goal is clarity — making an invisible idea visible fast.

The Three Tools — and the Difference

All three tools (ChatGPT, Gemini, Claude) can prototype in plain English. They are not identical:

For building software — the focus of this programme — Claude is where you will spend most of your time, for reasons we will cover in the next session. But understanding all three makes you a sharper user of each.

Exercise 1 — Draft a Client Letter in All Three

Open each tool in a browser tab and paste the same prompt:

"Write a professional email from a CA firm to a client who has not filed their GST return for the last two months. The tone should be firm but polite. Sign off as 'CA Subhash Chandra, Sahinov & Associates'. Keep it under 150 words."

Run it in ChatGPT, Gemini, and Claude. Observe:

• Which response feels most professional?
• Which one changed more with a follow-up ("make it shorter")?
• Which one was easiest to edit by continuing the conversation?

Exercise 2 — Build a Compliance Tracker Layout

Now try something structural. Paste this into Claude:

"Design a simple compliance tracker for a CA firm. It should track the following for each client: PAN, GST number, ITR due date, GSTR-1 due date, GSTR-3B due date, and filing status (Filed / Pending / Overdue). Show me this as a structured table with 3 sample client rows."

Claude will produce a clean, readable table with sample data. This is your first prototype of the kind of tool you will build as a real application in this programme.

Now push further: Tell Claude to add a column for "Days Until Deadline" and sort by urgency. You will see the prototype evolve, live, in the conversation.

Exercise 3 — Describe and Build a Screen

Try this in any of the three tools:

"Design the layout for the home screen of a simple school fee management app. The app shows: total fees collected this month, list of students with pending fees (name, class, amount due), and a button to mark a payment as received. Describe the layout clearly."

The AI will describe a screen — sections, buttons, information hierarchy. This is a wireframe in words: a prototype of the structure before any code is written.

The Feeling at the End of This Exercise

At this point you have something — a letter draft, a table, a layout. It exists. You made it in minutes.

But here is the gap that the next session reveals:

It only lives inside that chat window. You cannot send it to a client as a real working app. You cannot let someone else log in and use it. You cannot open it on your phone. You cannot update it from your office.

What you built is a sketch on paper — not a published product. To go from sketch to live on the internet, you need one more idea:

The cloud.