The first time someone asked me how Fahroh uses AI, I gave the wrong answer. I said the words document analysis, valuation models, natural language search. They nodded politely and changed the subject. I'd described what we built and said nothing about what it does.
What AI does at Fahroh is the work no individual specialist would do at the price the median transaction can afford. That's the whole claim. It's also the only one that matters.
In a mature property market, the work of verifying a transaction is distributed across institutions that have been around long enough to amortize their fixed costs over millions of deals. A title insurer in Texas has priced risk on a hundred million houses. They have a database of fraud patterns no individual firm can replicate, and they run the analysis on each new transaction at marginal cost. A solicitor in London has spent decades looking at conveyancing documents and can flag a problem on a deed she has never seen before, because she's seen its cousin a thousand times. The system works because the institutions are old, and they're old because the markets gave them time.
In the markets Fahroh operates in, those institutions haven't had time. The countries are younger as legal systems. The registries are newer or absent. The conveyancing profession in many places is small or unregulated. A buyer who arrives in Dakar, Lagos, or Manila can't pay a hundred dollars to a title insurer and be done with it. The institution isn't there.
That's where AI becomes load-bearing.
A document analysis system trained on tens of thousands of authentic and fraudulent title documents from a specific West African jurisdiction can spot the formatting anomalies, registration number patterns, and metadata signatures that a local specialist with twenty years of experience would catch. The system isn't smarter than the specialist. It has compressed the experience of every previous case into a pattern it can apply in seconds, at no marginal cost. A real estate transaction in a market without a title insurer can now be screened with the same level of due diligence a transaction in a market with a title insurer has had for fifty years.
That's what the AI does. It stands in for an institution that hasn't yet had time to grow.
The other thing AI does, which gets less attention, is valuations. In the markets we operate in, there's no functioning multiple listing service. There's no comparable transaction database. There's no benchmark a buyer can use to know whether the price they've been quoted is reasonable. The buyer, especially the buyer at a distance, takes the seller's word.
A valuation model trained on the platform's own verified transaction data, weighted by satellite-derived neighborhood signals and macroeconomic context, produces the first reliable comparative pricing benchmark in markets that have never had one. In mature markets, that kind of asset is owned by Zillow or Knight Frank or the local equivalent. It isn't owned by anyone in most of the markets we serve. We're building it because, again, the institution isn't there.
I think sometimes about how I'd explain this to my dad. He'd understand it instantly, in a different language. He'd say what we're building is the kind of work an honest broker does. He'd be right.
The point of the AI isn't the AI. The point is to do, at scale and at price, the work an honest broker would do, in markets that don't have enough honest brokers.
— Princess Ugochukwu, co-founder, Fahroh
