The first AI mortgage agent that does the math
Every lender has a broken front door. We fix them, one door at a time.
Other AI talks. AFDAL's agent computes: it applies every bank's rules, prices every feasible loan, benchmarks each offer against the market, cross-matches them and proposes better alternatives, then prepares a branded pre-approval in under 30 seconds. Try it now.
Guided demo on illustrative Moroccan grids. The production agent runs live, in voice and avatar, as Fadila at Bank of Africa and Amal at Bank Assafa.
Only AFDAL
The first AI agent that does the math, not just the talking.
Chatbots and voice bots explain. Rate calculators price one scenario at one bank. AFDAL's agent calls a combinatorial engine as a tool, so every figure it says is computed, compared and defensible.
| Capability | Rate calculator | Chatbot / voice bot | AFDAL agent |
|---|---|---|---|
| Computes exact payments, APR and total cost | Partly | — | ✓ |
| Applies each bank's eligibility rules | — | — | ✓ |
| Explores every feasible combination | — | — | ✓ |
| Benchmarks each offer against the market | — | — | ✓ |
| Cross-matches offers by payment, cost and APR | — | — | ✓ |
| Proposes better alternatives | — | Partly | ✓ |
| Explains in natural language, voice and avatar | — | ✓ | ✓ |
| Branded pre-approval in under 30 seconds | Partly | — | ✓ |
Watch it work
Voice, avatar and the engine behind them, on camera.
This preview can't play embedded video.
Watch on YouTube ↗Latest agentic AI demo: a full mortgage simulation in English.
This preview can't play embedded video.
Watch on YouTube ↗A world first: the first phone call placed by an AI mortgage advisor
This preview can't play embedded video.
Watch on YouTube ↗GITEX Africa, Marrakech 2026
The problem
Three broken front doors. Every lender has them.
Buying a home shouldn't feel like a bureaucracy. Yet borrowers meet a paper form ported to the web, and loan officers price by hand. The mortgage market isn't limited by capital; it's limited by coordination.
of online mortgage applications are abandoned
MortgagePoint, 2025of frustrated applicants leave for a competitor
MortgagePoint, 2025calculation reliability in branches without AFDAL
AFDAL study · 135 branches, 11 banksconversation that qualifies, computes, pre-approves and hands a qualified file to a human
Replaces all three doorsForward arithmetic
Loan × Rate × Term → Payment
One spreadsheet cell. It assumes the borrower already knows which loan to ask for.
Inverse optimization
Profile → every eligible loan, ranked
Max capacity, down payment derived from the profile, minimum and maximum eligible term, up to 15 ranked options with full amortization.
The one-question test: ask any rate tool for the minimum term a borrower qualifies for. None can answer, because none solve the problem in that direction.
The platform
Three layers. One brain.
Most mortgage tech stops at the chat window. AFDAL goes down to the math: every quote the agent speaks comes from a full combinatorial computation.
Turnkey white-label SaaS
Multi-tenant. No-code back office where product, risk and credit teams configure rate grids, conventions, aids and exceptions without engineering tickets. iFrame, WebView, REST API.
Voice + avatar agentic AI
Conversational layer that calls the engine as a tool. Five languages live, any language addable on demand. Chat, inbound and outbound voice, WhatsApp.
Combinatorial pricing engine
Enumerates every legal financing scenario (offers × insurance × terms × LTV × options) and ranks them by total cost. Seven years and ~10,000 person-days of edge cases encoded.
The data flywheel
Our marketplace is also our sensor network.
Afdal.ma is Morocco's consumer mortgage marketplace, with our avatar AI advisor built in. Every borrower simulation reveals what lenders offer and what borrowers want. That signal powers the intelligence modules we sell to lenders. We can replicate the same framework in any market.
compare free on the marketplace, guided by the avatar advisor
afdal.maoffers × demand × eligibility, per simulation
1.2M+ sims · 6 yrsScore · Copilot · Pulse · Forecast · Spark+
12 banks benchmarkedsharper pricing, better offers listed on the marketplace
+ leads, + conversion↺ Better offers attract more borrowers: the loop compounds.
| Signal captured on the marketplace | Powers | What the lender gets |
|---|---|---|
| Every lender's offers and rate grids, refreshed continuously | Score · Copilot | Real-time competitiveness on each offer; best-rate guarantee |
| Borrower demand: profession, income, city, term, product | Pulse · Forecast | Demand portraits, what-if pricing, 9–12 month production forecast |
| Eligibility vs. most-competitive offer, per simulation | Pulse | Competitiveness, Preference and Eligibility indices |
| Simulation sessions from borrowers in shopping mode | Spark+ | Its own customers identified while comparing, via secure API, and re-engaged in its app |
Launch
Stand up a local marketplace on the AFDAL framework, alone or co-branded with a property portal.
Localize
Load local lenders' grids, insurance, aid programs; switch the avatar advisor to the local language.
Accumulate
Borrower traffic builds the market's pricing and demand dataset: afdal.ma holds 6 years of it.
Monetize
Local lenders buy leads, then Score, Copilot, Pulse and Spark+, powered by data only we hold.
The full lending journey
A module for every stage, from pricing design to funding.
Activate one module or the full suite. They share one engine, one data layer and one AI orchestrator.
Offer design
Position pricing grids on live market data before rollout.
Capture
Reach borrowers before they compare, at a viable cost per lead.
Simulation & pre-approval
Every combination computed, the best one explained, pre-approved in <30s.
File assembly
Documents, eligibility and the property itself, in one flow.
Underwriting to funding
Real-time status, e-signature, notary coordination, AI follow-ups.
Measured in production
The numbers banks audit us on.
mortgage origination
CIH Banktime to approval
CIH Bankfull journey, simulation to funding
Bank of Africafinancing requests via web + app
Banque PopulaireCalculation reliability · mystery-shopper study, Feb 2026
135 branches · 10 banks + 1 broker · AFDAL mystery-shopper study. Bank of Africa network moved from 20% to 95% after deployment.
Agentic AI advisor
Talks like your best loan officer. Calculates like your pricing desk.
Voice, text and avatar in the borrower's language, with mid-sentence switching. The agent never invents a number: every figure comes from the engine.
- Algorithmic pre-approval in under 30 seconds, branded PDF
- 15 borrower profiles detected from language, never named to the customer
- Warm handoff to a human advisor with the full file
SESSION 8f3a-… · WEB VOICE · DARIJA ↔ FR
tool: full_simulation · 1.9M calcs · 0.7s
Illustrative transcript
Already on duty
Your advisor, your name. Two are live today.
Fadila
Bank of Africa
Bank of Africa's AI mortgage advisor, running on the engine behind creditbankofafrica.ma.
● In productionAmal
Bank Assafa · Attijariwafa bank group
Bank Assafa's Murabaha advisor, live since 6 April 2026: the first agentic-AI mortgage advisor in production in Africa.
● In productionSame AFDAL engine and agent underneath; each lender chooses its advisor's name, voice, avatar and brand.
Mode 01 · Borrower-facing
The front door, unbroken.
Replaces the static form on the website, app and branch tablet with one agent (voice, avatar or chat) running live on the engine. 24/7, responds in under a minute.
Mode 02 · Internal AI loan officer
The loan-officer bench, 24/7.
The same agent works as an AI employee: outbound calls, stalled-lead follow-up, document collection, custom simulations. Loan officers keep judgment and the relationship; the lender keeps the credit decision.
We are not here to replace people. We are here to return people to the work only people can do.
Who we serve
Built for every institution that originates mortgages.
Give every member a mortgage advisor, 24/7, without adding loan officers.
Activate AFDAL for your institution.
Choose your modules, get sandbox keys and an embed snippet in minutes. A deployment lead joins you for go-live.