AI real estate analysis and feasibility system interpreting property listings, images, floor plans, and location data to assess renovation potential and investment viability.
This system was developed for a property developer operating in the Portuguese real estate market.
The client’s typical workflow relies on reviewing online listing platforms, maintaining spreadsheets, and using informal input from architects or local contacts to form an initial view on a property. Each opportunity requires going through the same process, reviewing the listing, checking available drawings and photos, looking into planning constraints for the specific location, and forming a judgement before deciding whether to involve external consultants or proceed further.
They were looking for a way to approach this stage without repeating the same process manually or relying heavily on external input at an early stage.
The way information is received varies from one property to another. Some listings include floor plans and detailed descriptions, others are limited to a few images or basic data, and in some cases information comes through conversations or emails rather than structured listings.
Each time, the client has to piece this together manually, cross-check details against planning rules and local constraints, and rely on prior knowledge or external advice to form an initial view. This process makes it difficult to review multiple opportunities in a consistent way or move quickly without additional cost.
In the Portuguese real estate market, this is further complicated by planning frameworks, heritage classifications, and local precedents that require interpretation before any design or financial decision can be made.
The approach was to break down each listing into its different parts and analyse them using the type of knowledge most relevant to each one.
Written descriptions, images, floor plans, and location data were treated as separate inputs, allowing each to be interpreted in a more direct way, for example checking planning constraints against location, or using visual information to assess condition and likely scope of work. These observations were then brought together into a single view, so that each part could inform the others rather than being considered in isolation.
The goal was to build a process where information is handled in a consistent structure, while still allowing all available inputs to contribute to the final assessment.
Each part of the process is handled by a separate component:
The system is set up as an automated workflow that takes property information from an online listing or manual input and processes it through a series of AI-driven analysis steps.
Each input type is interpreted using AI models configured for that specific kind of information, supported by a defined set of reference material, including Portuguese planning rules, typical development approaches, and cost assumptions. The results of each analysis are then brought together into a single output, allowing the property to be assessed as a whole rather than through separate checks.
The current workflow was designed to support the review of individual properties at an early stage. Through use, it became clear that the next useful step would be the ability to assess multiple opportunities side by side using the same structure.
This is now being explored alongside a broader product version of the system, where the same workflow could be used to assess and compare multiple opportunities, and adapted into a tool that can be used more widely within the Portuguese real estate market.