Would your AI survive production?
A short questionnaire, an instant score, and a straight answer about where AI would break first in your organisation.
Before this reaches a review board
What a Saudi review board actually scores you on
This scorecard's five dimensions map onto five questions a Saudi review board asks before an AI system is approved for production.
| What is checked | The question asked | What passes |
|---|---|---|
| PDPL, lawful basis and data map | Is there a lawful basis for every personal-data field the system touches, and a map of where each one flows? | Field-level data map tied to a named lawful basis, kept current as the system changes. |
| NDMO, data quality and ownership | Does every dataset feeding the system have a named owner and a demonstrable quality check, beyond a defined schema? | Ownership and quality gates recorded per dataset, sampled and re-checked on a fixed cycle. |
| NCA ECC controls | Are the essential cybersecurity controls, access, logging, change management, evidenced continuously rather than asserted at audit time? | Control checks wired into the platform so the evidence a reviewer asks for already exists. |
| In-country residency by data class | Is it clear, class by class, which data must stay in-country and which may run elsewhere, with the boundary enforced rather than assumed? | A residency map per data class, enforced at the platform boundary, not left to a policy document. |
| Arabic handling | Does the system actually work in Saudi Arabic, dialect switching, Hijri dates, Arabic names, or only in a translated English layer? | Arabic handled as the working language from the start, evaluated per dialect rather than as one score. |
Worked example
A Riyadh-based insurer scored strongly on data foundation and use-case ownership but weak on governance and Arabic handling: PDPL lawful basis existed for some fields but not the retrieval index, and its assistant fell back to English whenever a caller switched dialect mid-sentence. The review board's questions were not about the model's accuracy, they were about the missing data map and the fallback path. We closed the data map first, then rebuilt the fallback around a named Arabic dialect boundary, and the same system passed on its next review.
Your context
The questions that decide the architecture
How sensitive is the data the AI would touch?
Where is the AI allowed to run?
What do you actually want to put live first?
How unified are your core data sources today?
Can an assistant retrieve from your documents, contracts and tickets?
How do you know your data is correct?
Who owns the definition of your key business metrics?
How confident are you on PDPL, residency and access controls for AI use?
Is there an agreed path to approve an AI system for production?
How do you measure whether an AI answer is correct?
Can you see what an AI system did in production yesterday?
Do your AI use cases have a named owner and a business metric?
Who can change an AI system after it goes live?
Your readiness score
0/13Answer the context and readiness questions to reveal your score and your recommended Sovereign AI path.
Questions people ask
What is an AI readiness assessment?
A structured check of whether an organisation can put AI into production: how unified and governed its data is, whether documents are retrievable, whether outputs are evaluated, and whether someone owns the outcome.
How long does the scorecard take?
A short set of questions, about a minute, with the score shown straight away on the page.
What happens after I submit my score?
You receive your scored breakdown by email with the recommended next step, and a Bilytica strategist follows up within one business day to discuss options.
Does this scorecard reflect what a Saudi review board checks?
Yes. Alongside the five general dimensions, the page above maps directly to PDPL lawful basis, NDMO data ownership, NCA ECC controls, in-country residency by data class, and Arabic handling, the checks a Saudi review board actually applies.
What if our weakest dimension is governance, not data?
Then the fastest path is usually the PDPL lawful-basis and data map first, since NCA ECC and residency reviews both assume that map already exists.
Does this scorecard reflect what a Saudi review board checks?
Yes. Alongside the five general dimensions, the page above maps directly to PDPL lawful basis, NDMO data ownership, NCA ECC controls, in-country residency by data class, and Arabic handling, the checks a Saudi review board actually applies.
What if our weakest dimension is governance, not data?
Then the fastest path is usually the PDPL lawful-basis and data map first, since NCA ECC and residency reviews both assume that map already exists.