What a Data Platform Health Check Actually Checks - Before You Hand It to an Agent

Before you let AI build on your data, check what it’s building on. Get a Data Platform Health Check before any agent gets near your data. It’s worth being specific about what that actually involves, because “check your foundation” is easy to say and vague enough to put off.

A Health Check looks at four things in particular. Governed access comes first - whether your security model is actually one consistent set of rules at table, row and column level, or a patchwork that happens to work for the reports currently in use. Platform health is next - capacity sizing, duplicated pipelines, and whether refresh patterns match what the data actually needs, rather than what got built under time pressure. Semantic model clarity matters just as much - whether your key metrics have a single, documented definition an agent (or a new starter) could build from without guessing. And governance posture closes it out - whether sensitivity labelling and lineage are actually tracked, or exist in theory but nobody’s checked recently.

It isn’t a redesign, and it isn’t scoped like one. The engagement is short and assessment-only - weeks, not months - and the output is a clear picture of what’s already solid against what needs attention, not a mandate to rebuild. Most organisations find most of their foundation is fine; a Health Check exists to find the specific gaps worth fixing before they get inherited by something faster than a person.

Where it does turn something up - inconsistent access rules, undocumented metrics, missing classification - a focused Data Governance engagement is the next step, not a return to the Health Check. That’s where OneLake Security gets defined as one model rather than several, Purview classification and lineage get put in place properly, and the access rules an agent would inherit get reviewed and tightened rather than assumed. It’s deliberately a separate, smaller step from a full platform consultancy engagement, because most of what a Health Check finds doesn’t need one.

A sound foundation and a trained reviewer solve different problems, and you need both - a governed platform with nobody able to judge what’s built on it is still a risk, just a different one. That’s why PTR pairs this work with the training argument we’ve made separately: the platform needs checking, and the people reviewing what gets built on it need to know what to look for.

If you’re not sure whether your own Fabric environment would hold up to an agent reading from or writing to it, a Data Platform Health Check is the place to start. Contact the PTR team through our enquiries page, by email at info@ptr.co.uk, or on 0118 979 4000.