FAQ
Straight answers
The questions partners, reviewers and IT teams ask us most.
Is the AI making accounting judgements?
No. The AI reads documents, drafts journals and writes narratives — always labelled as suggestions. Tax, balances, ZATCA checks and IFRS figures come from a deterministic, versioned rules engine, and nothing is final until a person at your firm reviews and signs it. Judgement and accountability stay human — by design.
Where does our data live? Does it comply with PDPL?
Structured business data — ledgers, documents and the audit trail — stays in the Kingdom, in Riyadh. During the pilot, limited text may be sent to AI models through a cross-border gate that classifies the data and strips or pseudonymises personal information first; secrets like ZATCA certificates never leave. This is aligned with PDPL and SDAIA cross-border rules, and the terms are written into the agreement.
What does the pilot cost?
Nothing — founding firms use the platform free during the pilot. Pricing comes later and is finalised together with founding firms, who lock in preferential terms and help shape it.
Is the ZATCA integration real or simulated?
During the pilot it is a sandbox: invoices are rules-checked before submission with the same deterministic rules and error codes as ZATCA, but nothing is actually submitted. You get the full error-catching value with zero risk while you evaluate. Real submission is a later, gated step.
What happens to our data if we stop?
You can export your data at any time — there is no lock-in. If the pilot doesn't deliver, you leave with your data and no obligations. Deletion on request, in line with PDPL.
Will our data be used to train models, or shared with other firms?
No. Every firm is fully isolated and data is never shared across tenants; the AI reaches data only through controlled, permission-aware tools. The agreement spells out the data-use terms — the default position is that your client data belongs to you.
Do we have to replace our current software?
No — the approach is integrate, not replace. The pilot runs on one client account with Excel, PDF or XML files loaded directly: no connectors, no IT project, alongside Wafeq, Qoyod, Zoho, Excel or whatever you use today.
What if the AI gets something wrong?
The platform treats the AI as fallible by design. Every extracted field carries a confidence score, and anything low-confidence routes to a human review queue instead of posting. The consequence of an AI mistake is an item waiting for review — not a wrong number in the ledger. Accuracy targets are agreed up front and measured during the pilot.
How long until we see value, and how much effort from us?
About four weeks to a signed-off result on one client. Your side stays light: pick the client and period, provide the files, and have your reviewer and partner do at the gates what they already do today.
Is the platform in Arabic?
Arabic-first, with full right-to-left interfaces and English alongside. Document intake reads Arabic and English documents alike, IFRS export is bilingual, and your clients' portal is bilingual too.