AI can format a report. But first you need to know what to calculate

Recently, I was asked to analyse two linked investment projects near Valencia, Spain.

The owner valued the properties at approximately €650,000. The expected investment across both projects was around €970,000–1,200,000. The anticipated total revenue after completion was more than €2.5 million.

On paper, the figures look attractive. But my task was not to confirm them. It was to test how realistic they actually were.

Both properties already had completed designs, permits and licences. The next step was to examine the investment model itself.

What I assessed

How justified the €650,000 price was for the two properties; the likely construction cost in that area; how the cost would change with a pool, solarium, parking and other elements; the likely cost of demolishing the existing building; whether existing utilities could be used and which additional connections would be required.

I also assessed the efficiency of the layouts; whether a boutique hotel, individual apartments or the sale of one entire property would be more profitable; the current price per finished square metre in the area; the realistic selling price of the completed product; the implementation period; and what positive and negative scenarios would look like.

Only after this can you talk about returns. In development, the price of the property is only the beginning. Construction, timing, utility connections, approvals, demolition, design changes, possible delays and exit costs follow.

If a project already has the land, a completed design and a licence, and construction can start, that is one scenario. If documents, layouts, use or utility issues must be changed or resolved, it is another. The difference in timing can be significant.

In one case, a project may be completed in roughly 1.5–2 years. In another, 2–3 years. Sometimes four years. Under difficult conditions, up to six years. That is already a very different capital economy.

That is why I always look beyond potential profit. I consider how long the money will remain in the project, what margin of safety exists in the model and what happens if the market changes.

Over more than 35 years in real estate, I have seen several situations where the market looked strong, prices formally held, yet real sales virtually stopped. That is why, even in a rising market, I do not calculate only one scenario.

Where AI helps in this work

After completing the analysis, I sent the client the report. The first message I received was: “Oh, Claude 🙂”

Yes, I use artificial intelligence. It helps structure material, assemble tables, format PDFs and turn a large volume of information into a clear form. It saves time.

But the analysis itself is different. AI can find a figure. It does not automatically know whether that figure fits a particular plot near Valencia.

It can find construction costs. But without additional review, it cannot determine whether they apply to this property, this structure, these conditions and this specification.

It can find comparables. But who decides whether they are truly comparable? Who accounts for a discount in a real transaction? Who assesses whether a layout is convenient? Who separates an optimistic scenario from a realistic one?

That is where professional work begins. In my reports, artificial intelligence can help with presentation. But the numbers, logic, choice of comparables, market analysis, risk assessment and conclusions are my work.

For me, AI is a powerful tool today. But investment analysis still does not begin with a neural network. It begins with an understanding of the market, the property and the money.