Hisabi.ai · A Codenovai product
Hisabi.ai: Building a GCC-Native AI Finance Platform from Zero to Production
Codenovai's own product. AI bookkeeping, cash-flow forecasting, and invoice automation built natively for GCC SMEs — Arabic + English, VAT-compliant, sovereign-deployable. The proof point that we operate in production, not just advise.
Outcomes
100%
GCC-native
Built for UAE/Saudi VAT, Arabic+English UI, AED/SAR-first calculations
<2s
Median LLM response
Hybrid routing: cheap classifier on Llama, complex reasoning on Claude
AED-first
Pricing & VAT
Built for the GCC tax framework, not localised from a US product
Sovereign
Deployment option
Customer data stays in me-south-1 (Bahrain) by default, on-prem available
The Problem
GCC SMEs were stuck between two bad choices for finance software. Western SaaS (Xero, QuickBooks) wasn't built for AED/SAR primary, didn't speak Arabic, didn't understand UAE VAT, and stored data in foreign jurisdictions. Local players had the compliance but lacked the AI capability that finance teams started expecting after ChatGPT shipped. The market needed a native solution that combined GCC compliance fluency with frontier AI capability — and nobody was building it.
What We Built
GCC-native foundation
Built from day one for AED/SAR, UAE VAT, Saudi VAT, and Arabic + English UI. Not a translation layer over a Western product — the data model, the chart of accounts, and the tax engine are GCC-first. Bilingual users switch language per-document, not per-account.
Hybrid AI routing
Designed the AI layer with cost in mind from the start. Classification tasks (categorise this transaction, parse this receipt) run on Llama 3.3 70B at near-zero cost. Reasoning tasks (forecast this cash position, draft this VAT filing summary) route to Claude. The router is invisible to the user but cuts AI costs ~70% versus calling Claude for everything.
ZOHO Books deep integration
Most GCC accountants already use ZOHO Books. Hisabi.ai sits on top with bidirectional sync — read from ZOHO, AI-augment, write back. No data lock-in, no migration friction, no fight with the existing accountant.
Sovereign deployment path
Default cloud deployment runs on AWS me-south-1 (Bahrain). For finance teams under stricter mandates (family offices, regulated entities) we offer on-premise deployment via the Private AI infrastructure pattern. Same product, same UI, deployed to their hardware.
Production-grade observability
Built with the same eval harness we ship to enterprise clients. Every AI output is sampled, scored, and cost-tracked. We caught two model regressions before customers noticed — the second model swap took 90 minutes from incident detection to production fix.
Stack
Codenovai Services
The Outcome
Hisabi.ai is in continuous production with paying GCC SME customers. More importantly for the agency side: it's the operational training ground for every Private AI and AI Automation engagement we sell. Every architectural decision Codenovai recommends to a client has been pressure-tested on Hisabi.ai first. When prospects ask 'have you actually shipped this?' — Hisabi.ai is the answer.