Governance-grade AI, delivered as working software — not slideware.
Secure AI platforms, retrieval systems and operational data products for enterprises that can't afford to get it wrong. Built by a twenty-year security and data-governance specialist who ships.
Enterprises are racing to adopt AI without the security, governance and data foundations to do it safely. Most consultants make that worse — they move fast and break governance. The rare thing is doing both: speed and control.
Four things, done with a governance specialist's discipline.
Not AI as a buzzword — AI and data systems built the way a security person builds: provenance, reconciliation, least privilege, and exceptions surfaced rather than hidden.
Private, sovereign AI your data never leaves.
Secure knowledge platforms with retrieval-augmented answers, orchestration into real workflows, role-based access — running inside your own perimeter, not someone else's cloud.
⟩ source: private hospital group · internal-perimeter AI, RAG + orchestrationOne trustworthy view across scattered systems.
Live pipelines and dashboards that stand on reconciled data — store code as the join key, freshness and provenance visible, no source system disturbed.
⟩ source: national coffee retailer · 75-store portal, 52-tool governed assistantThe controls that make AI safe to turn on.
Data-governance foundations, classification and access design, cloud-maturity roadmaps, and audited content delivery with per-partner scoping and full traceability.
⟩ source: drilling-services firm · IP protection, ISO 38505 / DAMAWorking apps your teams actually use.
Deployed operational tools with real roles, row-level security and live workflow — built on the platforms you already license, not a new subscription stack.
⟩ source: national coffee retailer · deployed ticketing app, live across the estateShipped work, measurable results.
Client names are held in confidence. The numbers are real.
Caught a seven-figure error the client's own reports missed.
A live budget pipeline reconciling invoice-level spend to the cent surfaced a capital figure their spreadsheets had silently dropped as text — corrected at source.
Turned two decades of dormant content into engineering speed.
An LLM pilot over unstructured maintenance-tactics content lets engineers source, aggregate and draft deliverables that used to take days — in a fraction of the time.
Enterprise AI that never touches the public internet.
A multilingual, role-aware knowledge platform with retrieval and an orchestration layer — designed for full data sovereignty inside the internal IT perimeter.
One store-centred portal across a whole estate.
A governed operations portal unifying sales, maintenance, budget and property data — with a natural-language assistant spanning six data domains, live in one intensive build.
The principles are the product.
Anyone can stand up a model. What makes a system trustworthy is the discipline underneath it — and that discipline is non-negotiable on every engagement.
Every important number carries its source, definition and last refresh. If you can't trace it, you can't trust it.
Your systems of record stay authoritative. Platforms I build cache, present and enrich — they never quietly become the new source of truth.
Missing is stated as missing. Uncertainty is flagged, never smoothed over to make a dashboard look finished.
Anomalies are shown, not hidden. The errors a system reveals are worth more than the ones it politely conceals.
People see only what their role should. Access design is part of the build from the first line, not a bolt-on afterwards.
Delivery is a running system, not a deck. Prove it end to end on the tools you already own, then improve it incrementally.
Drew van Vuuren
Twenty years in IT, eighteen of them in information security and data governance across financial services, government, telecoms and healthcare — in Europe and South Africa.
Now building the AI and data platforms those same disciplines demand: fast enough to matter, controlled enough to trust. Based in Cape Town, working with enterprises wherever they are.
Have an AI problem worth getting right?
If it needs to be fast and defensible, that's the whole point of iDatasec. Tell me what you're trying to do.
Start a conversation ↗