Free Tool · No signup to see the result
Sovereign AI Readiness Checker
Five questions about your regulator, where the data must stay, your GPUs, the first workload and the timeline. One of five reference architectures comes back — with the model family to evaluate and what to do in the first thirty days.
5 questions · 2 minutes
Which sovereign architecture fits your data, your regulator and your hardware?
Regime → residency → GPUs → first workload → timeline. You get one of five reference architectures, the model family that fits, and what to do in the first thirty days. No signup to see the result.
Frequently asked questions
What are the five reference architectures?
Air-gapped sovereign (nothing leaves your data centre), in-country private cloud (sovereign by policy — CMEK, service controls, egress block), hybrid with a sovereign core (split by data classification), sovereign edge (quantised models on offline devices), and cloud-acceptable but sovereign-ready (same open-weight stack, reversible decision). Each maps to deployments MindMap has shipped.
Why does the checker ask about GPUs?
Because hardware, not software, usually fixes the go-live date. Our standard stack goes from a clean cluster to first prompt in about 11 days; H100-class procurement runs 8 to 12 weeks. Knowing which situation you are in changes the plan more than any other answer.
Is the result a recommendation to buy something?
It is a reference architecture and a model family to evaluate. The evaluation itself — 20 to 50 of your documents, your acceptance criteria, infrastructure you control — is free of production data and you keep the harness whether or not you proceed.
How does this relate to the AI Readiness Assessment?
The assessment scores organisational readiness across six dimensions. This checker answers a narrower engineering question: given your regulator, residency and hardware, which architecture should you build and which model family fits. Use both.
Then run the family on your own documents.
Two weeks, fixed scope, infrastructure you control, no production data. You keep the evaluation harness either way.
Request a model evaluation →