FactCheck-AI Recommender is a Generative AI-powered task assignment system for AFCN, using Amazon Bedrock to match fact-checkers with verification jobs in seconds, improving efficiency, accuracy, and scalability for 40+ organizations across the Arab region.
About AFCN
The Arab Network for Fact-Checkers (AFCN) from ARIJ is a network that works to enhance transparency and neutrality in fact-checking in Arabic content and in the Arab region.
The network comes as a result of the rise in false and misleading information amidst unprecedented crises following COVID-19 worldwide. AFCN provides support to 40 fact-checking organisations/initiatives and over 250+ fact-checkers from all over the Arab world, through capacity building, protection, networking and innovation.
The network cares for news and rumors fact-checking, post-publication fact-checking and editorial fact-checking (pre-publication) which is scarce within newsrooms and platforms in the Arab region. It is a laborious process that requires professional, experienced, tech-savvy and nonpartisan personnel as well as the financing to evolve. ARIJ, as one of few Arab networks that carry out solid editorial fact-checking, stands by its firm belief that a deep-rooted fact-checking culture within the Arab media field will influence a sustainable future of accountability and transparency.
Challenges
The primary business challenge centered on the manual processing and verification of multiple loan application documents, including pay slips, bank statements, and salary certificates. The complexity of document verification presented a significant obstacle, with each loan request requiring the processing and validation of several documents to ensure authenticity and consistency.
The financial implications were substantial, with the manual verification process being both time-consuming and resource-intensive. The institution faced increased operational costs due to the labor-intensive nature of document checks, potential risks of human error leading to fraudulent loan approvals, and delays in loan processing times affecting customer satisfaction.
The Solution
To address AFCN’s operational bottlenecks, FactCheck-AI Recommender was developed — we implemented an intelligent, GenAI-powered system that automates the assignment of verification tasks to the most suitable fact-checkers based on content type, semantic context, language, and domain expertise.
Key Components:
Amazon Bedrock (Claude Models): Core generative AI engine providing semantic analysis, contextual understanding, and intelligent task-to-fact-checker matching through dynamic prompt engineering and few-shot learning.
AWS Lambda: Manages workflow orchestration, API integrations, and real-time event handling.
Amazon S3: Stores incoming job metadata, content, fact-checker profiles, and GenAI processing outputs.
Event-Driven, Serverless Architecture: Ensures scalability, low operational overhead, and near-instantaneous recommendations with minimal infrastructure management.
Integrated Performance Metrics Dashboard: Tracks matching accuracy, turnaround time, fact-checker workload distribution, and content trends.
Key Benefits:
Reduction in task assignment time
weekly labor hours saved, freeing up expert resources for core verification work.
Reduced reassignment rates, improving operational efficiency and minimizing redundant effort.
3x increase in verification task capacity without requiring additional staff.
Improved fact-checking precision by matching tasks with domain-relevant experts using AI-driven semantic and contextual analysis.
