Insights · Microsoft · Data · AI

Field notes from enterprise Microsoft delivery

Written by the architects who build and run these platforms for clients. Fabric, Azure, Power BI, Copilot and the governance that holds them together.

Written from delivery

Every piece is written or reviewed by the architects who build and run these platforms for clients.

Sourced, not summarised

Where we state a number we say where it came from, and when we last checked it.

Built to be used

Patterns, runbooks and checklists you can take straight into your own environment.

Latest article
Nine Power Platform responsibilities, from environments to Microsoft support, running past the project handover with no end date
Power Platform

Power Platform managed services: what running the estate involves after go-live

Power Platform managed services cover what a project leaves behind: environments, data policies, capacity, licences, alerts, releases and the support gaps.
Read the article

All articles

Browse by topic.

Pick a topic to filter the list below. Long-form, technical, and written from delivery rather than from a product page.

Three stacked layers showing where access is decided in Microsoft Fabric: a dashed domain box that labels and delegates but controls no access, a solid workspace layer where roles decide who gets in, and an item and data layer covering permissions and row and column security
Microsoft Fabric

Fabric governance: domains do not control access

Fabric governance is built on domains and endorsement, but a domain never decides who can see data. Certification is the control that changes behaviour.
Horizontal bar chart of the most common shape of Power Platform adoption: community running ahead at 300, strategy and vision and business value at 200, and support and governance still at 100, where makers support their own apps and environments are creatable by all
Power Platform

Power Platform adoption is not a usage number

Power Platform adoption is usually reported as a user count. Microsoft’s maturity model says that is the wrong measure, and a maker community proves
Two cards comparing what exists before anything runs: AKS shows six nodes recommended for production before the first request, Container Apps shows zero because it scales to zero with nothing to patch while idle, and AKS earns its place with API access or deny-by-default network policy
Azure

Container Apps vs AKS: the choice is who runs it

Container Apps vs AKS is usually settled by habit. Microsoft names two narrow reasons to need AKS, and neither is that the application is
Two nested boxes: an outer box for what a user can already reach, decided by permissions, and a smaller inner box for what Copilot may process, narrowed by labels, DLP and the EXTRACT right, showing Purview shrinks only the inner box
Copilot & AI

Purview and Copilot: what it can actually stop

Purview and Copilot are sold together, but Purview never changes what a user can reach. It only narrows what Copilot processes, and it has
Two horizontal bars comparing delivery layers: a workload bar that ends at a clear handover point marked done, and a central bar for policy, baselines and standards that runs off the edge of the frame because it never completes
Azure

Managed service vs project delivery model

Managed service vs project gets treated as a commercial choice. It is an operating model decision, and a project cannot deliver the central layer
Two rows of blocks comparing DLP simulation coverage: SharePoint and OneDrive fully evaluated including existing items, against Exchange, Teams and devices where only six of twenty blocks are filled because only new items are in scope
Data governance

Purview DLP policy design starts with one sentence

Purview DLP policy design starts with a sentence, not a template. Microsoft puts intent before configuration, and simulation mode is not a dress rehearsal.

Next step

Most people arrive here mid-problem.

Pick whichever sounds like your situation. It routes to the people who would actually handle it, not to a form.

Schedule a discovery call

We bought Copilot licences and cannot show what they returned.

The readiness assessment we run before any rollout: what Copilot can currently surface, and to whom.

See how it works

We are on Fabric but nobody has tested a recovery.

A continuity review against what the platform genuinely replicates, which is less than most teams assume.

See our Fabric practice