The least AI-native problem on my desk

The packaging business has outgrown its current warehouse. I need 400 to 800 square meters somewhere in greater Seville. Budget: €600,000 to €700,000. Most of it financed.

This is the least AI-native problem I’m working on right now. There is no MCP server for industrial real estate. No agent to call. The data lives in WhatsApp groups of local brokers, in PDFs sent over email, in second-degree contacts who know someone who knows someone. The process is exactly what it was in 1995, with WhatsApp replacing the fax.

The interesting part is not finding the building. It’s the financing structure I’m trying to stack:

• A standard mortgage from a Spanish bank for roughly 60% of the price.
• An ICO operative loan for the working-capital impact of moving.
• A grant from Andalucía TRADE that must be requested before signing the purchase, not after.

That last constraint is the one most owner-operators get wrong.

The grant is non-trivial — up to 30% of eligible capex in some tiers — but it’s gone the moment you sign without applying. The math forces a different decision sequence than the natural one: identify the property, lock the conditional grant approval, then close.

I write this down because the day I close on the building I’ll forget how much sequencing mattered. And someone else trying to do the same thing in Spain might save themselves a month of going in the wrong order.

I’ll write a follow-up when it’s signed.