Al-Shirshara Platform
A team that never sleeps: five assistants answer customers on Facebook and Instagram, turn their messages into orders, and hand off to a human when it matters.
A growing olive-oil brand was buried in Facebook and Instagram comments and DMs: questions, orders, and complaints at all hours, faster than staff could answer. Leads slipped through.
Every comment answered, in the brand's voice
Assistants reply to Facebook and Instagram comments and DMs around the clock, each reply tagged with the agent that handled it. The brand's tone is part of the system, not a hope.
Chats become orders, automatically
Structured orders are extracted from free-form Arabic chat: name, product, quantity, phone, city. Each one arrives with approve / reject actions, so a person signs off before money moves.
Five agents, one control panel
Gemini and OpenAI keys, the embedding model, and per-agent behaviour, configured in one place. Semantic search runs on pgvector over the brand's own content, so answers come from what the company actually sells.
Publishing, with products attached
Posts go out with products attached, so every campaign stays connected to the catalogue the assistants sell from, and every comment under it lands back in the same inbox.
Every action on the record
A full audit trail of every assistant and human action. When an agent's confidence drops, the thread hands off to a person, and that handoff is logged too.
Five specialised agents behind one inbox, with a confidence-gated human handoff: the moment an agent is unsure, a person takes over. Structured orders are extracted from free-form Arabic chat, semantic search runs on pgvector over the brand’s own content, and every update reaches the dashboard over WebSockets (Laravel Reverb), not polling.
Customers get answered instantly, around the clock.
Orders stop falling through the cracks.
Staff step in only when they're genuinely needed.
Live today.
