We didn't bolt a chatbot onto a finished system — we built the system around an engine that executes: 157 operational skills, real understanding of everyday Arabic, and one unbreakable rule: nothing posts without your confirmation.
Every accounting system the Arab merchant has ever used asked the same thing of him: learn the system's language before you start working. Know where the right screen lives, what this field means, why the form refuses to save. We built Sinad on the opposite conviction: the system should learn your language — its dialect, its shorthand, its way of naming things.
That is why we didn't bolt a chatbot onto a finished product. We built the ERP itself around an AI engine that understands the everyday Arabic request and executes it: 157 operational skills covering sales, purchasing, inventory, treasury, journal entries and reports — all passing through one unbreakable rule: nothing is posted and nothing is sent without your explicit confirmation.
Your language is the real interface
Nobody in a market says “create a sales document of type tax invoice”. A merchant says: “record an invoice for Abu Khalil — twenty cartons of milk at the wholesale price.” The first sentence is systems language; the second is business language. Every minute an owner spends translating his own words into the system's words is a daily tax on using software — one everyone has paid for so long they think it's natural.
When Sinad receives that sentence it doesn't hunt for keywords. It extracts the full structure of the operation: which customer? which item, in which unit? what quantity and price? is the price tax-inclusive? in which currency? Then it shows you a complete draft invoice — before anything touches the books.
What “understands your dialect” actually means
“Dialect” is an easy word in marketing and a hard one in engineering. Real understanding takes three layers working at the same moment:
Intent extraction
“Record”, “make me”, “put down”, “open one up” — different verbs for the same request. The right skill out of 157 has to be chosen from meaning, not phrasing, whether the request arrives with a Levantine, Gulf or Egyptian accent.
Entity extraction
Customers by their kunya (“Abu Ahmad”), items by their street names, quantities in market units (carton, sack, dozen), and prices in more than one currency inside the same sentence. All of it becomes real records in your system — not free text that goes nowhere.
Accounting context
The same sentence carries accounting decisions: 16% VAT where it applies, the correct customer account, the inventory effect, the exchange rate if the currency is foreign. Understanding that stops before the journal entry isn't understanding — it's a pretty interface over manual work.
AI that never reaches a balanced entry is just pleasant conversation.
Intelligence without recklessness: the confirmation gate
The merchant's biggest fear about “automation” is entirely legitimate: what if the system gets a number wrong? So every skill in Sinad is designed on the principle of preview before effect: the assistant prepares the document or entry in full, shows it to you with every line and figure, and waits. The confirm button is yours alone — nothing posts and nothing sends before you press it.
That isn't a limit on the AI; it's the correct definition of it. The system saves you the typing, the navigating and the searching — and leaves you the decision. When you confirm, the document posts through the same double-entry engine that governs everything in Sinad, so your trial balance stays closed at zero.
Why 157 skills and not one chatbot?
The difference between an assistant that answers and an assistant that works is wiring. Each Sinad skill is wired to a real operation in the system: creating a customer with full details, recording a receipt voucher, counting stock, drafting a journal entry, running a sales report in your base currency. Not generic accounting answers — execution inside your books, under your permissions.
And because they are separate skills, each has clear boundaries: what reads doesn't write; what writes previews before it posts; what sends — like WhatsApp documents — asks first. Trust is built from these small details, not from slogans.
We don't believe the future of business software is “prettier screens”. The future is running your company with the sentence you would have said to your clerk — while underneath it all, your ledger remains a ledger that honors the rules of accounting to the letter.
