AI for Government & PMO in Saudi Arabia

Report against Vision 2030 from live data, not last month’s deck.

[object Object]

Programme offices spend more time assembling status than acting on it. Orhan reads your own programme records and answers portfolio questions in plain language, drafts the structure underneath a new initiative, and keeps an audit trail of every answer it gave and every action it took.

Signs you need it

• Status packs take a week to assemble and are stale the day the steering committee reads them
• Risks surface at the meeting rather than before it, when there is still time to act
• The same initiative data is re-keyed into three different reporting formats for three different audiences
• Public AI tools are banned outright because nobody can control what would leave the building

Use cases

• Portfolio question answering: Ask about budget, schedule, dependencies or risk across the portfolio and get an answer with the source record attached, limited to what the asker may open.
• Programme mobilisation: Governance structure, councils, cadence and decision rights drafted for a new initiative in minutes, then edited by the people who own it.
• Document and submission intake: Contractor submissions, Etimad documents and correspondence read, classified and checked against your own rules before they reach a reviewer.
• Early drift detection: Schedule and budget drift flagged from live records while it is still a variance, not after it has become a reportable slippage.

What you get

• An assistant grounded in your programme records, with permissions compiled into retrieval so no answer exceeds the asker’s clearance
• Deployment inside your own tenancy or data centre, with data residency in the Kingdom where that is required
• An audit log of every question asked, every answer given and every source cited
• An evaluation set built from the questions your PMO actually fields, re-run on every change
• Source, prompts, runbooks and documentation handed to your team — you own what we build

How it works

• Scope: We agree the data boundary and the classification rules first, with your security and data governance teams in the room.
• Ground: Your programme records and document stores are connected, with existing permissions carried through into retrieval.
• Evaluate: A question set drawn from real PMO queries is scored, so quality is measured against your own work rather than a benchmark.
• Roll out: Released to one programme office first, widened only once the evaluation and the audit trail both hold.

When it is not the right fit

Decisions that must be made and signed by a named accountable official. The assistant prepares the evidence; it does not approve anything.

Can this be deployed inside the Kingdom?

Yes. It runs in your own cloud tenancy or data centre, so data residency, PDPL obligations and NCA controls are met by where it sits, not by a promise in a contract.

How do you stop it answering beyond someone’s clearance?

Permissions are compiled into retrieval rather than filtered afterwards. A document the asker cannot open is never retrieved, so it cannot be summarised or quoted back to them.

Will it invent a number if it does not know?

Every answer carries the record it came from. Where the evidence is thin or conflicting the assistant says so rather than resolving it silently, and unanswered topics are reported so you can see the gaps.

Who owns the result?

You do. Source, prompts, evaluation sets and documentation are handed over. There is no lock-in to us as a supplier.