PH · FINTECH · 4 × 21-DAY CYCLES

AI loan origination rebuilt for a Philippines digital lender.

A ~$60M loan-book PH digital lender replaced a six-month-stalled overhaul with four sequenced Software Factory cycles. Live in 12 weeks.

26h → 8m
Application to decision
+34%
Monthly origination volume
BSP-aligned
Default rate held in band
PH FinTech AI loan origination — Xamun Software Factory case study

Photo: Unsplash. Client name anonymised per NDA.

THE PROBLEM

Six months in, the in-house overhaul had shipped nothing.

The client — a Manila-based digital lender focused on consumer and SME unsecured loans — had spent six months and roughly ₱18M on an in-house origination overhaul. The build was stalled: scope kept expanding, a CTO had left mid-project, and the existing system was bleeding applicants to fintech competitors who decided in minutes, not hours.

The CEO wanted three things from the next attempt: a fixed timeline, a fixed scope, and ship-then-iterate cycles instead of a 12-month big-bang. Most AI companies in the Philippines they had spoken to bid the work as a 9-month T&M engagement. Xamun bid it as four 21-day cycles with sign-off gates between each.

WHY XAMUN

A Philippines AI software development company that ships in 21 days, not 21 months.

The Manila Software Factory is structurally different from a typical PH IT services bench. Each cycle has a fixed scope, a fixed delivery date and a working software output. Cycles compose into a larger build; nothing gets bundled into a vague "Phase 2 — TBD". For a CEO who had just lost six months to scope creep, this was the deciding factor.

The second deciding factor: the London-side XI team produced a one-page business case before scope sign-off, anchored to a specific origination-volume KPI. The CEO had a clear number to commit to internally — not "AI will help us grow".

THE APPROACH

Four 21-day cycles. Manila-led build.

CYCLE 1 · APPLICANT INTAKE

Mobile-first application capture with document OCR (IDs, payslips, ITRs). LLM-assisted document validation. Replaced the existing 14-field manual data-entry workflow.

CYCLE 2 · DECISIONING

Affordability + risk model + BSP-aligned audit trail. AI-driven credit decisioning with explainable reasoning. Underwriter override dashboard for edge cases.

CYCLE 3 · DISBURSEMENT

Integration with the lender's existing core banking + GCash/InstaPay disbursement rails. AML/KYC sweep on every approved application. Audit log per transaction.

CYCLE 4 · OBSERVABILITY + LIVE

PlayStudio dashboards for the CEO, COO and Compliance Officer. Live KPI tracking against the Cycle 0 business case. Shadow run for 2 weeks, then 100% cutover.

TECH STACK
Software Factory PlayStudio AWS ap-southeast-1 React Native FastAPI + Python XGBoost AWS Textract InstaPay / GCash APIs BSP audit logging
THE OUTCOMES

What actually moved.

DECISION TIME
26 hours → 8 minutes

Median application-to-decision. Approved-loan-to-disbursement: same-day for ~92% of applicants.

ORIGINATION VOLUME
+34% monthly

Month-on-month new originations Q1 post-launch. CEO's Cycle-0 business-case target was +22%.

RISK
In-band default rate

90-day default held within the lender's BSP-aligned tolerance band. No risk drift in the first 6 months.

UNIT ECONOMICS
Cost-per-decision ↓ 71%

Underwriter team reduced touch on the 78% deterministic applications; capacity refocused on disputed cases.

WHAT THIS COST

~$95k total. Vs $400k+ for the stalled in-house build.

Four 21-day cycles, fixed-priced. Total contracted value: ~$95k (~₱5.4M). The stalled in-house overhaul had already cost the lender roughly $310k with nothing shipped. Xamun shipped a working production system for less than a third of that, in 12 weeks.

In-house build stalled? Talk to Manila.

The Software Factory specialises in restart engagements where a previous build has not shipped. Half-day Discovery, fixed-priced cycles, shipping software from day 22.

Book a Manila Discovery → Talk to the team →