Financial Inclusion
Most economic life never touches a bank
In the markets this fund serves, the branch was built for a customer who barely exists: salaried, documented, collateralised, fixed at one address. Economic life carried on regardless. Trade credit moves through relationships, and working capital sits in stock and cash. Income is real and unrecorded, and the formal system has no instrument that can see it. Credit, payments and identity for people that system has never priced: that is the sector. What follows is the argument for why the pricing failure is structural, and what we want from the founders building past it.
Exclusion is a cost structure
Banks did not overlook these customers. They priced them, correctly, as unservable under a branch cost model. Underwriting needs evidence, the branch accepts only the documented kind, and producing documents is exactly what informal economic life does not do. Assessing a borrower with no file takes human judgement, and human judgement at branch cost exceeds the value of a small loan. So the system rations by paperwork.
The exclusion that results looks like a social failure, and it is usually described as one. It is an accounting outcome. Nobody inside the model acted irrationally, which is precisely why the model cannot fix it from within. The constraint is the opportunity.
What AI changes
The binding cost in consumer and small-business finance is the cost of establishing trust. AI collapses it. A model can read the evidence informal economies actually produce: the rhythm of a stall's stock cycle and repayment behaviour inside a trading network. None of that fits a loan file. All of it is signal. When the marginal cost of assessment falls towards the marginal cost of compute, the customer who was uneconomic to know becomes serviceable, and the question stops being whether these markets can be served. It becomes who builds the systems that serve them.
Who builds them matters because of where models come from. A credit model trained on salaried borrowers in the old centres misreads a Karachi wholesaler in both directions at once. Underwriting is sovereignty applied to money: whoever owns the model decides whose economic life counts as evidence. This is the fund's thesis, Sovereign AI for the Five Billion, applied to finance.
What we look for
The same things in every market, stated plainly.
Founders native to the condition, building from how these economies actually work rather than porting a playbook written for markets where everyone has a file.
Underwriting proven on the evidence these markets produce, in the languages their customers use. Show us the model reading reality, not a slide describing it.
Distribution that meets economic life where it already happens. Our reading is that behaviour change is the most expensive product feature there is, and that the best products here need none.
Regulatory posture treated as architecture. The fund is managed by a regulated manager, and we back founders who design for the licence from the first commit. In finance the licence is a load-bearing wall.
A founder who clears these has usually built something the formal system cannot copy, because copying it would mean abandoning the cost model the incumbent is made of. That asymmetry is the moat.
Why these markets
The fund's geographic focus is the Middle East and North Africa, South Asia, Africa and Southeast Asia. As we read these markets, the distance between economic activity and financial infrastructure is widest here, which makes them the markets where AI-native finance gets built first rather than retrofitted. In the fund's reading, there is no core banking estate to integrate with and no legacy margin to defend. The fund expects Lagos, Dhaka, Jakarta and Cairo to skip the banking model the old centres grew into rather than converge on it.
Operating from the DIFC keeps the fund within reach of the capital behind this work and of the markets it is built for. Distance is a cost structure too.