AI-Assisted Fatwa
Mujīb Research Draft
Validated
Musahhih Policy Reference
Audited
CASE STUDY: RETROSPECTIVE
AI-Assisted Fatwa
Operations Platform
A case study showing how our team built a human-controlled RAG system around the Dar-ul-Ifta review chain — Mujīb → Tashīh → Tasdīq — to reduce repeated search and blank-page drafting while keeping scholarly authority intact.
minimum wait reduced
0
+ days
Human-approved
AI never auto-publishes
Reference-backed
Citation-first drafting
Trusted by 1000+ Brands
Dar-ul-Ifta Overview
What The Dar-ul-Ifta Does — Islamic Jurisprudence Authority
The Dar-ul-Ifta is an Islamic jurisprudence institution that receives shar’i questions from the public and issues formal religious rulings (fatwas). Each question passes through a scholarly chain: the Mustafti asks, the Mujīb researches and drafts, the Musahhih checks ruling and references, and the Musaddiq gives final approval before publication.
Scholarly Review Chain
Mujīb, Musahhih & Musaddiq
Multi-stage review process ensuring each fatwa passes through qualified Mujīb drafting, Musahhih validation, and Musaddiq approval before public issuance.
Knowledge Repository
Qur'an, Hadith & Fiqh
Vast archive of published fatwas, Qur’an/Hadith references, fiqh books, and institutional rulings spanning decades of scholarly work.
The Challenge
Where The Operation Was Slowing Down
The real bottleneck was the full preparation cycle: search → understand → reference → draft → recheck. Highly qualified scholars spent too much time on repeated discovery and formatting.
01
Long Response Times
Users waiting at least 15 days; complex religious questions could remain in queue for much longer due to manual processing.
02
Manual Reference Search
Mujīb must manually search across old fatwas, Qur’an/Hadith references, fiqh books, and previous responses for each question.
03
Duplicate Submissions
Similar answers already exist, but users submit again because semantic discovery across the fatwa archive is weak.
04
Complex Classification
Inheritance, divorce, finance, oaths, and modern business issues each require careful triage and domain-specific knowledge.
05
Limited Visibility
Users and staff lack transparent status tracking from question submission through review to final publication.
Operational Impact: Qualified scholars spend too much time on repeated discovery and formatting instead of applying high-value scholarly judgment to complex questions.
The Solution
What We Built For The Dar-ul-Ifta
A human-controlled RAG system that prepares an evidence pack for each question: similar fatwas, citations, draft answer, validation flags, and workflow status.
STAGE 01
Ingests
Receives and normalizes Urdu/Roman Urdu/English questions.
STAGE 02
Retrieves
Searches fatwa archive, books, and citations via semantic RAG.
STAGE 03
Drafts
Generates evidence-backed answer draft with source references.
STAGE 04
Validates
Checks citation gaps, contradictions, and confidence levels.
STAGE 05
Routes
Assigns to Mujib queue with risk flags and priority.
DASHBOARD
Dashboard Features
Queue health, assignment tracking, SLA monitoring, and audit trail.
AI Evidence Pack
For each question: normalized query, retrieved sources, draft answer, validation notes, missing-information questions, and complete audit trail.
Scholarly Review Queue
Mujib reviews draft and references, Musahhih checks ruling and wording, Musaddiq gives final approval.
Duplicate Detection Engine
Finds similar published fatwas and shows fatwa numbers before a duplicate question is submitted to the review queue.
Multi-Agent Query Handling
Splits complex questions into parallel processes: inheritance calculation, divorce ruling lookup, and financial guidance retrieval.
Audit-Proof Document Trail
Every draft, source, correction, approval, and publication action is logged with role, timestamp, and reason.
Workflow Integration
How A Ticket Becomes A Published Fatwa
Streamlined operating sequence showing how AI assists the scholarly chain while keeping human authority at every decision point.
Control Rule: AI never issues the fatwa. It prepares a review packet for the scholarly workflow.
Technical Stack
A Modular Platform, Not A Fragile Script
Designed for security, reliability, and enterprise scale, the platform architecture segregates processing logic from data schemas.
User Interface
- Search portal
- Question submitter
- SLA tracker
- Scholarly web workspace
Processing
- Urdu normalization
- FastAPI endpoints
- Intent classification
- Parallel routing
RAG Retrieval
- Hybrid search
- Semantic ranker
- Kitab-Bab directory
- Fatwa registry index
LLM & Validation
- Draft generation
- Citation matches
- Hallucination shields
- Verification checkers
Workflow Engine
- Role assignment
- SLA monitors
- Change logs
- Correction feedback
Data Stores & Adherence Standards
Data stores: SQL archive, Vector database, Object store for PDFs/books, Audit log, Analytics warehouse. All platform pipelines and logistics are fully compliant with TLS 1.3 encryption in-transit and AES-256 at-rest.
Specialist Network
Specialized Agents Working As One Team
Instead of a single unstructured prompt, our platform deploys coordinated micro-agents, each highly trained in specific business roles.
- Triage Agent
Interprets intent, parses urgency, category, and language. Dispatches sub-tasks to specialized domain agents.
- Archive Search Agent
Searches published fatwa database by semantic similarity, fatwa number, category, and historical references.
- Book/Citation Agent
Retrieves relevant passages from digitized fiqh books, Quran/Hadith references, and classical commentaries.
- Draft Composer
Generates structured first-draft answers using retrieved evidence and institutional formatting templates.
- Validation Agent
Checks citation validity, contradiction scan, missing facts, and confidence scoring before human review.
- Policy/Routing Agent
Routes high-risk categories to manual-first review paths based on institutional guidelines.
- Learning Agent
When Mujib corrects a draft, the delta is analyzed, validated, and loaded into vector memories weekly.
Continuous Training Engine
Mujib corrections improve search quality, prompt accuracy, reranking relevance, and validation rules over time to dynamically align with current consensus.
Outcomes
What Changes After Platform Implementation
Empirical operational metrics showing how the institution redefined question processing and scholarly review capabilities.
Faster
Draft preparation time
AI prepares structured first draft and evidence pack instantly
Reduced
Duplicate submissions
Existing fatwas surfaced before new question reaches the queue
Higher
Reference reuse rate
Relevant old fatwas, books, citations attached to every draft
Better
Review consistency
Standard templates and validation notes reduce variation
Before Implementation
- 15+ day average wait time for public inquiries.
- Manual search across unindexed historical fatwa books.
- No active duplicate detection on question queue.
- Limited status visibility for waiting mustaftis.
With MT Pixels Platform
- AI-assisted structured draft prepared in seconds.
- Semantic RAG retrieval with precise source citations.
- Automatic duplicate surfacing during public submission.
- Full transparency, auditable path, and progress search.
Governance
Conservative By Design
A religious AI workflow must be auditable, role-aware, and explicitly human-controlled.
01
Human Approval Only
AI can recommend or draft templates; only qualified scholars make the final decisions.
02
Citation-First Drafting
Drafts display original source evidence and confidence state before being reviewed by a scholar.
03
Sensitive-Case Routing
High-risk questions are routed to manual-first review paths or prompt the user for more facts.
04
Versioned Edits
All corrections, rejected drafts, source changes, and approval actions are securely logged.
05
Role-Based Access
Private questions, classical libraries, reviewer notes, and publishing rights are tightly permissioned.
06
No Auto-Publishing
The system cannot release a fatwa publicly without final verification from the scholarly workflow.
Final principle: AI reduces search and drafting efforts; religious responsibility remains entirely with qualified scholars.