The credit-decision engine that gets sharper with every loan.

No. The AI is the core decision engine — it underwrites, prices, and services within your rules — while 'B' is only the interface on top of it, so it's a multi-step system of action, not a single model call. It connects to your existing core and loan-management system with no rip-and-replace: Bolehlah becomes your decision-of-record while your ledger of record stays yours. We orchestrate and constrain a sophisticated AI model — we don't build foundation models — and the defensibility is in the vertical workflow, not the base model.

The engine is designed to surface specific reason codes for every recommendation — the same model that produces the decision surfaces its drivers — so it's interpretable by design rather than a black box. That lets your team generate the borrower-facing rationale needed for adverse-action and decline notices, all backed by a tamper-evident audit trail. Bolehlah surfaces the reasoning; your institution remains responsible for the final notice and its regulatory sufficiency.

Bolehlah never lends and never decides unilaterally. Your institution sets policy, keeps final authority, and remains the regulated entity; the AI recommends, prices, and drafts strictly within your rules, with human sign-off required before any consequential action. It's a decision-support layer with approval gates, not an autonomous lender.

Two layers. Today the moat is the vertical workflow — the regulatory-fit AKAD/PDPA loan journey, per-lender policy calibration, human-sign-off gating, and the audit layer that makes Bolehlah your system of record — switching-cost defensibility a horizontal model can't replicate. Over time, Bolehlah is architected as a data flywheel, built so each loan can sharpen the next decision on regulated koperasi and licensed-money-lender data that's hard to reach at scale. We're early and pre-revenue, so we describe that compounding as the designed direction, not an advantage we've already accumulated.

Reliability is engineered into the system around the model, not assumed from the model. We ground the AI on your own policy and data, enforce deterministic (non-AI) guardrails on money-critical fields like rates, amounts, and fees, and put human approval gates before any consequential action. The AI can't assert a rate or a rule it isn't grounded in — the surrounding architecture is what makes it dependable.

Yes — it's useful from the first loan. On day one the engine runs on your existing policy rules and underwriting logic, grounded in bureau, CTOS, and eSGPA data — no accumulated history required. The data flywheel then improves calibration on top of that working baseline over time, so the compounding is upside, not a precondition.

The engine is built for fairness monitoring and feature governance — controlling what the model may and may not use, so protected-attribute proxies are governed rather than slipped in. Your institution remains the regulated entity that sets acceptable criteria; Bolehlah provides the monitoring, documentation, and reviewable audit trail to evidence it. It's designed to align with Malaysia's AI-governance direction and global frameworks such as the EU AI Act's bias-monitoring expectations.

We run a sophisticated AI model that we orchestrate and constrain; we don't disclose a specific vendor and we don't build our own foundation models. Your lender and borrower data are contractually excluded from training the model — no data is used to improve a shared or base model. Raw inputs are processed in-region and are not retained for model improvement; where records must persist for your own audit and regulatory obligations they stay under your control, and only anonymized, aggregated decision-intelligence compounds across the platform — never identifiable records.

Yes, by design. We route each task to a right-sized model, use grounding and caching to avoid redundant calls, and let deterministic guardrails — not the model — handle money-critical arithmetic like rates and fees, so you don't pay a model to do math. It's built to keep AI cost a small, controlled fraction of the value of each decision rather than a per-query drag on your margin.

Encrypted at rest, hosted in-region, with per-lender isolation at the database layer and a hash-chained, tamper-evident audit trail. AI processing runs in-region under contractual no-training terms, and raw inputs are not retained for model improvement. We never sell data.

Under PDPA 2010, Lunar Flame Sdn Bhd acts as the data processor and your institution is the data controller, governed by a written Data Processing Agreement and a tamper-evident audit chain. Borrowers give consent through your onboarding flow. Bolehlah is the layer that helps you meet these obligations, not one that adds new ones.

No — Bolehlah is a SaaS tool, not a regulated lender, so your licence carries the BNM/KPDN/AKPK obligations and Bolehlah sits alongside it. The platform is designed to align with Bank Negara Malaysia's direction on AI in financial services and Malaysia's national AI-governance guidelines — human oversight, model-risk governance, transparency, and audit trails — so it helps you meet those expectations rather than creating exposure. We describe this as designed to align; we don't claim certification or endorsement we don't hold.

Yes. The loan journey is AKAD-aware: itemised cost disclosure, profit/principal separation, aqad e-signature, and a Shariah-board-reviewable audit trail. Conventional and Shariah books run side by side.

Setup is designed to take a few business days from BD sign-off — mostly connecting data sources and calibrating your policy. B runs across one role-aware app (lender staff, borrowers, investors — live on Google Play and the iOS App Store, HUAWEI in review), a web portal, Telegram, and an HTTPS API, with one conversation history per borrower. Use Telegram for quick check-ins, while sensitive flows — eKYC, statements, AKAD signing, approvals — run in the more-secured app.

Pricing starts free, then scales by loan-book size across Free, Plus, Pro, and Max, with a by-contract Enterprise track and individual Officers available à la carte. Exact amounts are published at launch.