How I Created a Business Plan for a Real Estate Company with the Help of Claude

I didn’t hire a consulting firm to draft the business plan, nor did I spend weeks in Excel. Using a precisely written prompt and Claude, I put together a complete, auditor-level 5-year business plan for a real estate company—including a balance sheet, income statement, and cash flow statement—and then verified every number with a separate audit prompt.

Below, I will show you exactly how. Not the theory, but the actual process – including the prompts, in their entirety.

What was the assignment—a business plan using artificial intelligence?

A business plan needed to be prepared for a real estate company. The company acquires industrial properties – halls, workshops, warehouses – leases them out, and then, after a holding period of 2-3 years and subsequent appreciation, sells them, repays the bank loan, and embarks on new investments. Financing is split 30-70 between equity and bank funding.

The goal was for the plan to be not just a narrative, but a truly closed financial model: balance sheet, income statement, cash flow – all in accordance with Hungarian accounting law. Furthermore, it also needed to support the valuation background for the sale and purchase of a business share.

AI does not replace financial thinking. However, for those who can think precisely and write accurately, it multiplies their speed.

The Process – Step by Step

Step 0: I Wrote the Prompt with a Prompt

Before generating anything, a question arose: how do I write a prompt that truly yields a good business plan? The answer was simple — I did this with Claude as well.

I described the situation in a few sentences and asked it to write a precise, professional Claude prompt from it. This was the 'meta-prompt' — the prompt that wrote the prompt:

From this brief description, Claude wrote the following detailed specification. I simply approved it, and that's how the whole process began.

Step 1: The Prepared Prompt

The meta-prompt resulted in a precise, structured specification — with role definition, mandatory elements, assumptions, and quality requirements. This was the prompt I used to generate the actual business plan:

The key is: I didn't draft this specification manually. I told Claude what I wanted to achieve — and it wrote the prompt. I merely reviewed and approved it.

Step 2: Plan preparation

Claude delivered the complete plan in a single run: a portfolio development table, transaction schedule, five-year income statement, balance sheet, cash flow statement, and valuation metrics. The output was provided in Markdown tables, with built-in verification rows – it even indicated if a row's sum matched.

The first run took approximately 90 seconds. With a consulting team, this would have taken weeks.

This is what the generated material looked like – with a blurred filter over the numbers, as the specific data is confidential:

Step 3: The Auditor's Audit Prompt

A business plan is only as good as its numbers are accurate. That's why I didn't rely solely on it for this – I wrote a second, independent prompt whose sole task was a complete numerical audit:

This prompt was simple, but the context – the entire plan – was already present in the conversation. Claude then traced back all the numbers, line by line, step by step, from the initial assumptions to the results. What exactly did it check? Everything:

• Balance Sheet Consistency: Each year, total assets equal total liabilities and equity, with zero discrepancy.
• Depreciation Derecognition: Upon sale, the accumulated depreciation of sold properties is accurately deducted.
• Corporate Income Tax Base Calculation: Verification of the correct utilization of deferred losses from the tax base.
• Cash Flow Consistency: The closing cash balance each year precisely matches the cash line item on the balance sheet.
• Retained Earnings Roll-forward: The previous year's after-tax profit is accurately carried forward to the next year's retained earnings.

The result: zero discrepancy for all five years and across all three financial statements. The only note was a rounding difference of 31 thousand forints in a tax calculation – which is perfectly acceptable for a five-year business plan.

The quality of the prompt determines the quality of the output. This is not a metaphor – it's an engineering fact.

What AI capabilities did Claude use in this business plan?

• Financial Modeling: Building a closed calculation chain with internal consistency for five years, ensuring cross-consistency across three statements.
• Application of Hungarian Accounting Regulations: Adhering to the total cost procedure as per Act C of 2000, and treating real estate as fixed assets.
• Structured Verification: Auditing its own work from an external perspective, tracing back every number step by step.
• Combination of Valuation Methodologies: Harmonizing NAV-based, EBITDA multiple, and book value-based approaches.

What did I learn from this?

First: prompts can be generated by prompts. If you don't know exactly how to phrase your request — ask the AI to help you write the prompt. This meta-prompt technique is one of the most useful things I've learned when working with AI.

Second: role assignment is not optional. “You are an experienced financial planner” – this sets the model's behavioral scope, terminology, and quality threshold.

Third: quality requirements must be stated explicitly. If you want the balance sheet to balance, include it. The model will not know on its own that this is non-negotiable.

Fourth: verification is also a prompt. Don't manually review the numbers – have them checked by an audit role. It's more reliable and auditable.

Summary

A meta-prompt, a generated specification, a business plan execution, and a dedicated audit prompt. This was the technical infrastructure. The output: a complete, closed, five-year financial model, compliant with Hungarian accounting standards, with zero balance sheet discrepancy – all in a few minutes.

This is not about AI knowing everything. It's about AI multiplying the efficiency of those who can think precisely and instruct accurately. The financial knowledge, the development of assumptions, and the valuation logic – these remained with me. The calculations, the structure, and the verification – these were performed by the model.

Frequently Asked Questions (FAQ)

Is it possible to create a business plan using artificial intelligence?

Yes—Claude AI or ChatGPT can generate a structured business plan that includes financial projections, market analysis, and risk assessment. The key is a good prompt: the more context you provide, the better the result.

Which AI is better for business purposes: ChatGPT or Claude?

ChatGPT is faster and more creative, while Claude AI is more precise and better at handling long documents. Claude is stronger when it comes to business plans, contract analysis, and data analysis. ChatGPT is better for marketing copy and brainstorming.

How reliable are AI-generated financial forecasts?

It’s a great starting point, but every figure needs to be validated. The AI provides the structure and logic—you contribute the market knowledge and empirical data.

How do Hungarian business leaders use AI in practice?

Business planning, meeting preparation, email automation, document generation, data analysis—at Gloster, 23 AI skills are used in daily operations, saving 15–20 hours per week.

Related articles

• How I Prepared for a PhD Interview in 15 Minutes — Using AI — Another Case Study: Analyzing 40+ Sources, a 5-Iteration Prompting Process
• We no longer build chatbots—we build a digital workforce —The strategic framework: why “agent” and not “chatbot”