Historical source figures retained as provided (August 2026); original institution withheld. Not MMT figures.

Inclusion Pulse

65,480,112 people save with MMT Partner Bank.
7,004,206 of them borrow.

MMT Banking AI is built for the difference: using AI to turn savings, wallet and agent behaviour into safe, explainable credit for the people Pakistan's banks have never been able to score.

Active borrowers: 10.7% of depositorsSavers without credit: ≈58.5 million
PKR 280.4BTotal deposits
PKR 131.4BGross loan portfolio
324,934Branchless agents
1,593,298Women active borrowers

What the numbers say

  • 46.8%

    Loan-to-deposit ratio. Nearly PKR 149B of deposits is not yet working as microcredit — headroom for AI-underwritten lending.

  • 18,758

    Average loan per active borrower (PKR). This is nano-lending at scale, where manual underwriting cannot pay for itself. Only automated, AI-driven decisions can.

  • 22.7%

    Share of active borrowers who are women. Women’s Finance Program and Women’s Financial Services can grow this with gender-aware product matching.

  • 2.2%

    Share of active borrowers classified rural (157,076). Agri, livestock and Agriculture Finance Program lending has room to scale, but only with climate-aware risk models.

  • 21.6

    Active borrowers per agent. Each agent is a sales, service and fraud-risk point that AI can monitor in real time.

Borrower mix

Customer profiles and wallet signals are synthetic; product names are from the demonstration bank's live catalogue

Voice Loan Officer

A customer describes her business in her own words — Urdu, Roman Urdu or English, typed or spoken. AI extracts the facts, scores her from Digital Wallet and agent behaviour instead of a credit file, matches a microfinance credit product, and explains the decision in both languages.

Underwriting view

Real GLP PKR 131.4B · regional mix and sensitivities are illustrative

Climate & Portfolio Risk Radar

Pakistan's microfinance book lives or dies with the monsoon and the wheat price. Move the scenario, and the model re-prices portfolio-at-risk by region and product before the first missed instalment.

Start from a scenario
AI-recommended actions
    Projected PAR 304.2%▼ 0.0 pts vs normal year
    Portfolio at riskPKR 5.5Bof PKR 131.4B GLP
    Borrowers needing outreach382,430of 7,004,206
    Simulated live stream across the 324,934-agent network

    Agent Network Guardian

    Every branchless transaction is scored in milliseconds for SIM-swap takeovers, mule accounts, split cash-outs and agent collusion. Tap any flagged row to see why the model stopped it.

    Transactions scored14,820
    Held for review42
    PKR protectedPKR 3.8M
    TimeAgentCityTypePKRRisk

    Case file

    Tap a highlighted transaction to open its case. The first high-risk one will open automatically.

    Real depositor and borrower totals · segment sizes and conversion rates are assumptions you can change

    Savings-to-Credit Bridge

    About 58.5 million people already trust a demo bank with their savings but have no loan. The model finds the ones whose behaviour already looks like a good borrower, and designs the journey that brings them in.

    Conversion Funnel & Portfolio Lift

    PKR 8.2B

      Ask MMT Banking AI

      A leadership copilot grounded in the bank's own numbers and the scenarios on this page. In production it would sit on the bank's warehouse, with every answer traceable to a query.

      How it would run inside MMT Partner Bank

      DataCore banking, mobile banking app events, Digital Wallet signals, agent POS, CRM and complaints — into a governed lakehouse on the bank's own cloud.
      ModelsAlternate-data credit scoring, climate-exposure risk, real-time fraud graphs and next-best-product propensity, each with SBP-ready explainability.
      AI agentsUrdu voice loan intake, collections nudges, agent coaching and a leadership copilot, with human approval on every credit decision.
      GovernanceModel risk register, bias checks by gender and region, audit trails and ISO 27001-aligned access controls.