AI-Powered Assessment Product
Document ingestion to scored indicators to a generated recommendation roadmap. Shipped end to end and packaged as a repeatable proof of concept for GCC real estate portfolios.
Overview
A sustainability assessment product for GCC real estate portfolios. A client uploads their documents, the system scores a set of indicators, and it generates a recommendation roadmap they can actually act on. Built solo in Dubai from February 2026, discovery through to deployment.
The problem
Portfolio owners in the region are being asked for sustainability positions they have no structured way to produce. The data exists, scattered across utility bills, building specifications and consultant reports. Turning that into a defensible position is a consulting engagement that takes weeks and produces a document nobody can audit afterwards.
How it works
Deterministic Python computes every score. The language model does two jobs only: it narrates the result, and it cross-checks each figure it cites back against the source. Nothing in the output is untraceable.
This is the second time I have built it this way, independently of the first. The reporting pipeline at NSAI uses the same separation. Having reached for it twice without planning to, I think it is the right default for anything where a reader might reasonably ask where a number came from.
What shipped
- Document ingestion and extraction.
- Scored indicators with a defined methodology behind each one.
- A generated recommendation roadmap.
- Client-ready PDF export.
- Packaged as a repeatable proof of concept rather than a one-off engagement.
Tech stack
What is honest about this
It is deployed and it has produced client-ready output. It has not been through a real adoption cycle, so I cannot tell you a retention number or an activation rate. That gap is deliberate to name rather than paper over, and closing it is what I am working on next.