Siraj: Building a National GenAI Learning Infrastructure in Jordan
Siraj is a generative AI learning assistant developed by the Government of Jordan in partnership with Replit to support students and teachers across the Kingdom. Built and piloted rapidly on Replit’s agentic software platform, Siraj provides an Arabic-language, conversational interface that draws on officially approved digital textbooks and curriculum resources. During its initial pilot, the system supported hundreds of thousands of interactions, enabling students and teachers to ask questions in natural language and receive verified, curriculum-aligned answers without any technical expertise.
Beyond a proof of concept, Siraj is positioned as a national learning infrastructure. The Ministry of Education plans to extend access to roughly 1.6 million students and 90,000 teachers, embedding AI-enabled self-directed learning and lesson support into everyday schooling. Framed as a strategic step in Jordan’s digital inclusion agenda, the initiative aims to improve learning outcomes, relieve pressure on overburdened teachers and parents, and familiarize a new generation with AI-mediated knowledge practices in Arabic.
Siraj can be read as a paradigmatic case of “algorithmic public service,” where a state uses generative AI to reconfigure educational relations, temporalities of study, and authority over knowledge. By tightly coupling the assistant to official textbooks, the system enacts a form of epistemic governance: it narrows the informational universe in ways that increase trust and alignment with national standards while potentially limiting exposure to alternative perspectives. The conversational interface normalizes AI as a co-present “para-teacher,” subtly shifting students’ expectations about where expertise resides and how help should be requested.
The project also highlights “platformized” state capacity. Jordan effectively outsources part of its digital modernization to Replit’s infrastructure and agentic tooling, accelerating innovation but entangling public education in a private, globally oriented stack. This raises long-term questions about data sovereignty, linguistic representation, and the reproduction of global North–South asymmetries in AI ecosystems. At the same time, Siraj’s Arabic-first design and clear curricular anchoring indicate a localized algorithmic culture, where global AI capacities are indigenized through language, policy, and pedagogy. The initiative thus operates simultaneously as an inclusion mechanism—broadening access to high-quality support—and as a site where new digital divides may emerge around access, literacy, and critical understanding of AI.
Practical Implications for Organizations
- Anchor AI assistants in verified, domain-specific content to build trust, reduce hallucinations, and align with regulatory or curricular frameworks.
- Treat generative AI deployments as sociotechnical infrastructures: plan for teacher/employee training, new work routines, and ongoing governance, not just technical rollout.
- Design in the local language(s) and cultural idioms to increase adoption and signal respect for local knowledge practices.
- Use pilots to gather granular usage data and narratives, then refine guardrails, UX flows, and escalation paths before scaling nationally or enterprise-wide.
- Negotiate platform partnerships with attention to data ownership, model improvement rights, and exit options to avoid long-term vendor lock-in.
- Pair AI tools with explicit critical literacy programmes so users learn when to rely on the assistant, when to doubt it, and how to cross-check responses.
Consumer tribes that may relate to this case study:
Eureka
Eureka
