MurshidAI

Emirati entrepreneurs often don't know which government or private funding programs they actually qualify for, and navigating that landscape usually requires a human advisor who can ask the right questions and match a founder's situation to the right program.

• Designed a multi-agent architecture with four cooperating AI agents, each responsible for a distinct step of the advisory conversation.
• Built the primary interface over WhatsApp, so entrepreneurs could interact with the advisor through a channel they already use daily, with no new app to install.
• Implemented a profiling agent that gathers a founder's business details, then hands off to a recommendation agent that matches them to cited, real funding programs rather than generic advice.
• Added a confidence-based escalation path — the moment the system's confidence in its own recommendation drops, the conversation is handed off to a human advisor instead of guessing.

Built for a Khalifa Fund for Enterprise Development challenge, MurshidAI placed 4th at the 42AD Summer Program hackathon — demonstrating a working, safety-conscious AI advisor pattern that knows when to hand off to a human rather than overstate its own confidence.