Fiji's Finance Ministry backs AI to expand MSME access to credit
The Ministry of Finance is urging banks and lenders to put Artificial Intelligence to work for micro, small, and medium enterprises. Minister Esrom Immanuel welcomed the Fiji Development Bank's AI-assisted credit assessment tool, noting its potential to help small business owners secure loans, access finance, and reach customers without extra friction.
He acknowledged that many people are unfamiliar with these tools, and confirmed that guidance and support will be provided so customers can use them effectively. Immanuel stressed that embracing AI is a practical step in modernising the financial system and creating opportunities for businesses across Fiji.
"For instance, FDB, that's their task. They have to bring the service down to that level… The customers or potential customers do not need to come to the office. So that is for FDB and also other financial institutions to go down to that level." He also assured the public that these advances are safe and supported by stronger IT infrastructure across banks.
FDB Executive Filimone Waqabaca underscored the economic upside: Fiji should use technology to build a stronger economy, and the new AI credit tool supports national plans. "A new era of artificial intelligence has dawned… with applications and tools that have been made publicly accessible." With this foundation, more people can start and grow businesses knowing practical support is available.
What this means for finance leaders
- Serve thin-file borrowers at scale: Use AI to assess MSMEs with limited collateral or formal history by analyzing transactional patterns, supplier payments, and verified alternative data-while staying aligned with AML/KYC requirements.
- Lower distribution costs: Remote origination and on-site field onboarding reduce branch dependency and shorten turnaround times, especially for rural customers.
- Improve unit economics: Faster, more precise decisions can lift approval rates without sacrificing risk discipline, improving portfolio yield and customer satisfaction.
- Build trust with guardrails: Keep humans in the loop for edge cases, set policy limits, and prioritize explainable features and clear adverse action notices.
- Operational readiness: Connect models to core systems via APIs, establish data pipelines, and instrument MLOps for monitoring, drift detection, and audit trails.
Guardrails to get right from day one
- Model risk management: Document assumptions, run backtests and challenger models, stress test for macro shifts, and align outcomes with credit policy.
- Fairness and inclusion: Test for bias across regions, genders, and sectors; use constraints and monitoring to keep outcomes consistent and defensible.
- Privacy and security: Protect PII end-to-end with encryption and strong access controls; vet third-party vendors and maintain clear data-retention rules.
- Customer protection: Provide plain-language consent, simple opt-outs, and staffed support channels-especially during onboarding and early repayments.
- Standards and guidance: Use frameworks such as the NIST AI Risk Management Framework and keep sight of supervisory expectations, including AML/CFT obligations.
How to execute in the next 90 days
- Pick two pilot products: e.g., MSME working capital and small equipment loans.
- Form a cross-functional squad: Credit, risk, data, IT, compliance, and frontline ops with a single owner and weekly decision cadence.
- Run a data and process audit: Identify usable internal data, key gaps, and quick integrations; define a minimal feature set for an explainable baseline scorecard.
- Layer in machine learning: Add ML on top of the baseline with clear stability thresholds and reason codes.
- Field-test remote origination: Equip officers with mobile tools; run with 50-100 applicants; compare against a holdout group.
- Set hard metrics: Approval rate lift, time-to-decision, early delinquency (30/60 DPD), cost per booked loan, and NPS/complaints.
- Train staff and customers: Short tutorials, helplines, and community sessions so borrowers know how the process works and where to get help.
Signals to watch
- Approval rate and loss rate changes for thin-file applicants.
- Average time to decision and first disbursement.
- Early-vintage delinquency curves and roll rates.
- Customer complaints, opt-outs, and fairness metrics by segment.
For broader context on supervisory themes, see the Financial Stability Board's overview of AI and machine learning in financial services.
If your team is mapping the tooling landscape, here's a practical starting point: AI tools for finance.
Your membership also unlocks:
