Skip to content

Data, sovereign platforms and technology sourcing

Run AI on your data, in the Kingdom, without vendor lock-in.

Watercolour of a modular data centre on the northern edge of Riyadh: a gantry crane lifts one module off a shared concrete base while a new module waits on a truck, and the Kingdom Centre stands small on the horizon.

The intelligence comes to your data.

We design the data foundations and the hosting that AI needs to run every day: on-premises, on a sovereign cloud in the Kingdom, or both. Then we help you decide what to build, buy or partner on, and choose models and vendors on your criteria, with no alliance to defend.

The strategic question

What do we build, buy or partner on, where does it run, and how do we avoid lock-in?

The problem

Every vendor has an answer. Few are tested on your material.

The decision needs evidence from your own documents and constraints, not from a demo.

More context

Model providers, cloud platforms and integrators each pitch a complete stack, but data classification, hosting rules and transfer limits narrow the real options, and a wrong commitment is expensive to undo.

  1. Data that is not ready

    Quality, access and classification gaps surface halfway through the build.

  2. Hosting decided late

    Security and data teams approve an architecture already chosen.

  3. Lock-in by default

    Contracts give away data portability and exit terms.

What we do

Five lines of work. One accountable team.

  1. Data readiness for AI

    We assess the quality, access and classification of the data each initiative needs, aligned with NDMO, and close the gaps before design.

  2. Hosting architecture

    On-premises, a sovereign cloud in the Kingdom, or hybrid, chosen against your data classification, your security controls and the cost of running it.

  3. Build, buy or partner

    A decision for each capability, weighed on control, cost, speed and the skills your team will need to run it.

  4. Model and vendor due diligence

    Shortlisted models and vendors tested on your own Arabic and English material, with the criteria and the results written down.

  5. Tender support

    RFP design, evaluation scorecards and contract terms on data and exit, so the award rests on evidence you can defend.

Where we help you write an RFP or tender, we do not bid on it.

How an engagement runs

Four stages, applied to this practice. A gate after each.

Each gate is a decision: continue, change course or stop. That keeps spend tied to results, and keeps you in control.

The Live Path

  1. Discover

    We map the data, the constraints and the options, and test the shortlist on your material.

    Gate A ranked roadmap with business cases.

  2. Design

    Target architecture, hosting and sourcing decisions are signed off with your security and data teams.

    Gate Design sign-off and a value baseline.

  3. Lead the build

    We hold partners to the agreed architecture and contract terms while the system is built.

    Gate The Go-Live Gate Checklist, signed.

  4. Run & transfer

    The exit and portability plan stays current, so changing provider later remains a real option.

    Gate Value signed off by finance, and the next operation chosen.

We count systems running in live operations, not pilots.

See The Live Path in full

What you get

What your team keeps.

What you get

  • Target data and hosting architecture
  • Build, buy or partner recommendation
  • Model and vendor shortlist with test results
  • RFP pack and evaluation scorecard
  • Exit and portability plan

Typical outcomes

  • A hosting decision your security team signs off early.
  • Models and vendors chosen on evidence from your own material.
  • Freedom to switch providers later without starting over.

How we measure

What we baseline, then track.

Each measure gets a baseline in the Design stage and is reported against it once the system is live. This is how we measure, not a result we promise in advance.

  1. Data readiness for priority use cases

    The share of priority use cases whose data meets the agreed quality and access bar.

  2. Model quality on your own material

    Accuracy on a test set of your Arabic and English material, for each shortlisted model.

  3. Cost per transaction once live

    What each transaction costs in production, against the estimate in the decision.

  4. Exit readiness

    The share of critical platforms with a documented exit and portability path.

How we measure, step by step

Ways in

Start with a fixed scope. Then decide what comes next.

See every way in
  • 3 to 5 weeks · fixed scope

    AI sourcing and platform decision

    Build, buy or partner: decided with models tested on your material.

    • A build, buy or partner recommendation
    • A hosting option: on-premises, sovereign cloud or hybrid
    • Three to five models or vendors tested on your Arabic and English material
    What's included

    Decide what to build, buy or partner on, and where it should run, with models and vendors tested on your own material rather than judged on their pitch.

    Best for: Organisations weighing many vendor offers or preparing a tender.

    • An RFP pack or a vendor due-diligence report
    • An exit plan

    Where we help you write an RFP or tender, we do not bid on it.

  • 4 weeks · fixed scope

    The 4-week diagnostic

    AI opportunities ranked by value, with business cases and a roadmap.

    • Interviews with your teams in Arabic and English
    • A map of processes, data and handoffs
    • Opportunities ranked by value, data, fit and risk
    What's included

    Four weeks, fixed scope. We map how the work really happens, then hand you AI opportunities ranked by value, with business cases and a roadmap. You decide what comes next.

    • Business cases for the top initiatives
    • A roadmap with gates for the next stages

Questions

Questions buyers ask.

Where will our AI system be hosted?

On-premises, on a sovereign cloud in the Kingdom, or hybrid.

More detail

We choose against your data classification, your security controls and the cost of running it, and your security team signs off early. Data stays in the Kingdom by default. Any exception needs your written approval.

Do you sell or favour a particular AI model?

No. We are model-neutral and have no model to sell.

More detail

Global, open-source or local sovereign providers are chosen for your data rules and your goals, and shortlisted models and vendors are tested on your own Arabic and English material.

Can you help write our AI tender and then bid on it?

No. Where we help you write an RFP or tender, we do not bid on it.

More detail

We help with RFP design, evaluation scorecards and contract terms on data and exit. The AI sourcing and platform decision takes 3 to 5 weeks with a fixed scope.

Weighing vendor offers or preparing a tender? Start with the sourcing decision.

30 minutes with our team. We tell you plainly where we would start.