How to Build Internal Firm Tools with AI in Under 30 Minutes

A practical guide from Growth Club:

Artificial intelligence is no longer something firms are experimenting with on the side. It is quickly becoming the operational backbone for modern accounting practices. Across advisory, compliance, workflow, quoting, onboarding, and simple internal tools, AI is letting firms automate tasks that used to be time consuming and expensive to build.

In a recent Growth Club session, Canadian accountants Brian Clare from Blueprint Accounting and Anderson Peter George from Quanto demonstrated just how far this has come. They built a working pricing calculator in under 30 minutes without writing any code. No engineering team, no custom development project, and no complicated tech stack. Just ChatGPT and a no code tool called Lovable.

This guide walks you through how they did it and how you can follow the same process inside your own firm. If you want to start building internal tools that cut admin, speed up decision making, and create consistency across your practice, this is a simple way to get started.

1. Start by defining the problem clearly

Most firms fall over before they even start because they try to build something too big. The power of AI comes from narrowing the scope and being crystal clear on the problem you want to solve.

Brian and Anderson didn’t set out to build a full quoting system with integrations, tax logic, role permissions, and an approval hierarchy. They picked one high value, repetitive process that every firm struggles with: consistent pricing.

Before you open ChatGPT, take five minutes and map out the manual tasks or spreadsheets slowing your team down. Look for anything that:

• Requires judgement but is repeatable

• Lives in someone’s head rather than a shared tool

• Causes your team to quote inconsistently

• Leads to rework or back and forth

• You keep meaning to fix but never have time for

Pick one pain point and tackle that first. You can build more later. The goal is momentum, not perfection.

2. Write a strong, structured prompt for ChatGPT

The quality of your build comes down to the quality of your initial prompt. ChatGPT works best when you give it a clear role and context.

Here is a simple structure you can follow:

1. Give it a persona.

“You are a specialist prompt engineer with deep experience in accounting software, quoting systems, and workflow tools.”

2. Describe the end goal.

“I need a web based pricing calculator that supports fixed fee and value based pricing, includes service tiers, calculates markups, and routes quotes for approval.”

3. List constraints or preferences.

“We are an accounting firm. We quote in AUD. We need a layout that is simple for admins and junior staff.”

4. Finish with a question.

“How would you improve this brief? What questions do you need answered before you generate the build instructions”

Ending with a question is important. It forces ChatGPT to surface gaps in your thinking and run through all the logic you may not consider. It will ask about approval flows, pricing logic, currencies, discounting, industry variations, user roles, optional extras, and edge cases.

Let the model interview you. The more you answer, the better the tool becomes.

3. Take the refined prompt and generate the app in Lovable

Once ChatGPT has helped you tighten the brief, take the final version and paste it into Lovable. Lovable turns natural language into a working app, complete with:

• A database schema

• A basic user interface

• Pricing logic

• Page structure

• Form fields

• Validation rules

• Basic styling

If the build stops or errors out, hit “Try to fix”. Lovable will troubleshoot its own code and continue the build.

Within a few minutes you will have a working application. It might be rough. It might need polish. But the foundations will be solid, and you will have saved weeks of developer time.

4. Iterate using plain English, not code

This is where the tool gets powerful. You can update and improve the app by talking to it as if it were a developer sitting next to you.

Ask for changes like:

“Add a CSV export button.”

“Insert a drop down for industry type and apply a 15 percent markup to ecommerce.”

“Add a toggle for monthly vs annual pricing.”

“Change all headings to our brand font and colour #0050B3.”

“Add a margin calculator so partners can stress test pricing.”

Each time you ask, Lovable updates the app and shows you the results live. You never touch HTML, CSS, SQL, or Javascript.

This is where internal tools get fun. You go from idea to working version instantly.

5. Connect your real data for scale

Once the structure is in place, you can connect the tool to live data. Lovable integrates directly with Supabase, a Postgres based backend that is perfect for accounting firms who want something reliable, secure, and scalable.

Supabase stores your data.

GitHub stores your code and version history.

If you ever want to bring in an external developer, all the documentation and structure is already in place. No messy handover. No guessing where things live. Your groundwork is solid.

6. Test it with your team

A tool is only useful if your team can actually use it. Once the first version is working, share it with two or three team members who regularly quote work or build proposals.

Ask simple questions:

• Does the logic match how we price

• Are the markups accurate

• Are the services clear

• Does the approval flow make sense

• Where did you get stuck

• What feels clunky

Treat the feedback session like a product sprint. Each piece of feedback becomes a new prompt for Lovable. The app gets stronger each day.

Once it feels intuitive for everyone from juniors to partners, you can embed it in your proposal software, connect it to your CRM, or even build a client facing version.

7. Practical prompting tips

Here are a few rules that make your builds stronger:

Always end your initial brief with “Please ask questions.”

This helps ChatGPT find gaps and assumptions.

Re use prompt structures.

Once you build one tool, the next one becomes much faster. You can clone whole components.

Keep your language simple and direct.

Technical jargon does not help. Plain English is easier for the model to follow.

Think in small building blocks.

Your first version should be tiny. Add complexity only as needed.

8. What you can build next

Once you have built a pricing calculator, it becomes obvious how many internal tools you can create with the same approach. Common examples include:

• Client onboarding forms

• Engagement letter generators

• Service scoping checklists

• Internal workflow triggers

• Leave request tools for staff

• Small data dashboards

• Debtors review checklists

• Document request lists

These used to require a developer or off the shelf software. Now you can build your own tailored versions that match your firm exactly.

9. What you can do this week

If you want to get started, here is a simple plan for the next seven days:

• Pick one repetitive process your team deals with

• Write your ChatGPT brief using the format above

• Let the model interview you until the brief is tight

• Build the first version in Lovable

• Share it with one colleague and gather feedback

• Treat every change request as a new prompt

• Track how much time you save

Most internal tools get built in under an hour. Some take a few days of iteration. But all of them improve productivity without the cost of a developer.