Run the check here

The Microsoft Landscape Self-Check

Thirteen questions to tell whether your Microsoft estate is ready to run AI in production, across three domains. Answer for the estate you have, not the one on the roadmap; a hesitant yes counts as a no. The three domains are sequential: a gap in the Azure foundation usually shows up as a data problem, and a data problem usually shows up as an AI failure, so fix them in that order.

Azure foundation

The layer everything else depends on: if governance drifts here, the gaps show up downstream, not here.

Is your landing zone built and measured against a known standard (CIS, Well-Architected), or set up once and left to drift?
Are identity, network, and security controls enforced in code, or clicked together by hand?
Do you know what each workload costs to run, or is Azure one monthly invoice?
Do you know which regions your data can end up in, or would you have to check before answering a customer?

Data platform

An AI use case reads whatever data it is given. A number nobody trusts does not become trustworthy because a model touched it.

Is there one source of truth for the data an AI use case would read, or copies no one fully trusts?
Can you trace where a number comes from, end to end?
Are access and sensitivity controls on the data itself, not just the reports?
Are the pipelines that feed the data platform documented and monitored?

AI readiness

The controls specific to running a model in production: classification before the build, evaluation after it, monitoring once it is live.

Could an AI service reach the data it needs through governed access, or would someone be exporting CSVs to feed it?
Do you know which use cases carry EU AI Act obligations before you build them?
Is there a way to evaluate an AI output for accuracy, or would you be trusting a demo?
Once live, is there monitoring to catch a model drifting, or do you find out from a complaint?
Does every live AI system have a named owner for its behaviour and running costs, or did ownership end with the project?

0 of 13 answered. The result appears with the last answer.

Common questions

How is this different from the Azure Well-Architected Self-Check?

The Well-Architected Self-Check scores one Azure workload against the five pillars. This one is broader and shallower on purpose: thirteen questions across the Azure foundation, the data platform, and AI readiness, built to tell you which of the three is the actual blocker before you commission a deeper review of any one of them.

We already have Azure and Power BI. Why would we need this?

Owning the platforms answers a different question than whether a use case can actually reach the data it needs through them. The self-check separates 'the tools exist' from 'the tools are governed and connected', which is where most estates actually stall.

What does the result actually tell us?

Which of the three domains, Azure foundation, data platform, AI readiness, scores weakest, and that maps directly onto whether the Microsoft Landscape Assessment starts with the landing zone, the data estate, or the first use case.

Is this only relevant if we're planning to use AI?

The Azure foundation and data platform questions hold on their own: a landing zone that's drifted or a data estate with no single source of truth costs money whether or not AI ever enters the picture. The AI readiness section is what changes if AI isn't yet on the roadmap.