Building a Credit Engine That Can Defend Itself

A handful of credit applications a month, assessed by hand, is survivable. Fifty a month is not — and more importantly, manual review is never fully consistent. Under South Africa's National Credit Act, inconsistency isn't a bad debt, it's a regulatory finding: every application requires a full affordability assessment, every decline requires written reasons. So when we began working with Kusanda Capital, a registered South African credit provider, we built their credit assessment engine from the ground up, specifically so manual review would never become the operating model, letting them start operationally efficient from day one rather than retrofit it later.

This credit engine reads payslips, bank statements and bureau reports, assesses affordability, surfaces red flags early and hands the human assessor a structured, comprehensive analysis to work from rather than a stack of raw documents. Every number the final decision rests on — affordability, ratios, thresholds — is computed deterministically and handed to the reasoning layer as a fact it must use, never one it calculates; and bureau findings are scored against a written rulebook we can print and defend, not a model's own impression. From there, the engine drafts and digitises the entire agreement process, and a decline generates its statutory reasons automatically. One deliberate boundary: the credit bureau check itself, the debit order mandate and payment collection run through a separate, accredited provider built for exactly that — our engine gates disbursement on their confirmation and logs the outcome, rather than duplicating regulated infrastructure that already exists. Short of those three pieces, our credit system is end-to-end: from the point of application to the point of recording final payment collection; with one rule we have never relaxed, which is that the system only advises and a person decides.

Once the system was live, production surfaced early, crucial lessons that a testing environment simply could not have shown us — the kind that only appear once real applications, real documents and real money are actually moving through a system. We took each one seriously and moved quickly, producing an updated version of the system that folded those lessons in rather than letting small, but significant gaps sit unaddressed. That discipline of rebuilding rather than patch when the structure itself can be significantly improved, and treating what production reveals as more authoritative than what testing suggested, is what we bring into every engagement.

This credit system is a natural fit for lenders and other financial services providers most directly, but the same discipline serves any founder-led business making decisions that need to hold up months later, not just feel right on the day. If that describes a process in your business, we would welcome the conversation.

Kusanda Capital is a registered credit provider in South Africa: kusandacapital.co.za. Lending is subject to Kusanda's own assessment and approval; this article is a systems case study, not a credit offer. Fundamentals Firm's founder and Principal Lead, Mhlengi Kunene, is also a director and shareholder of Kusanda Capital.

Next
Next

An AI Assistant, Built In-House