A governed, trusted, AI-ready Power BI estate, with a roadmap to get there

Power BI advisory and governance is the strategy, governance framework and roadmap for your Dynamics 365 Power BI and Fabric estate. It sets six things: an estate and maturity assessment, a target operating model, a governance framework, a roadmap, a licensing and capacity strategy, and an honest Copilot and Fabric readiness verdict, so that whatever you build and run is governed, trusted and AI-ready by design.

Most estates do not fail on a single report. They drift.

The pattern is familiar to any Head of Data who has watched it happen. Power BI takes off, the business finally likes self-service, and then the estate starts to get away from you. Workspaces multiply. The same measure means something different in every report, so no two numbers agree and nobody quite trusts the pack. The roadmap is a slide nobody has updated. And the board is asking about AI and Copilot before the foundations underneath are anywhere near ready.

The fix is not lockdown. It is direction and a lightweight governance framework: decide where you should sit between self-service and central control, put the certified, reconciled numbers on a foundation the business builds on, and get an honest read on where you actually are before you spend on AI. That is what this page is about, and it is a different job from diagnosing a broken report or building the model. It is deciding what to do with the estate, and how to govern it.

You land on managed self-service: the business builds its own reports on a governed foundation of certified semantic models, workspace standards and security patterns owned by a central team, and for most SMEs you start more centralised to build that certified foundation, then increase self-service as maturity grows. The goal is the lightest governance that meets the objective, not lockdown.

This is the single biggest call the governance owner has to make, and getting it right kills the objection that governance will kill the self-service the business finally likes.

There is a spectrum, and both ends fail. Fully business-led self-service is fast and popular, and it sprawls: thousands of workspaces, no accountability, the same figure computed a dozen ways. Fully centralised enterprise BI reconciles, and it becomes a queue the business routes around. Managed self-service is the balance point Microsoft and the wider market have converged on, and it is where a credible governance framework sits: a central team owns and certifies the shared semantic models, sets the workspace standards and the security patterns, and the business builds freely on top of that governed foundation.

For an SME the honest recommendation is almost always the same shape. Start more central, build the certified foundation the numbers depend on, then open up self-service as the estate matures and the guardrails prove themselves. Governance is scaled to where you are, not imposed as a regime you are not ready for.

A Power BI governance framework is a tiered workspace and tenant governance model, certified and endorsed semantic models, row- and object-level security, Microsoft Purview sensitivity-label and DLP data-protection design, and a right-sized centre-of-excellence (CoE) operating model, rolled out lightweight and iteratively, not as a lockdown regime.

Ask most vendors what a governance framework actually consists of and you get a policy document. Here is the concrete design work, the layers that stop sprawl and make the estate trustworthy, each one something we design with you, not a slide.

• Tiered workspace and tenant governance. Not every workspace needs the same oversight. We design the tiered model Microsoft recommends, governed and ungoverned levels kept as simple as possible, set by four decision criteria: who owns the content, its scope, its subject area, and whether it is business-critical. Naming conventions, domains and a decision log come with it, so the tenant is governable without being a bottleneck.
• Certified and endorsed semantic models. This is the trust layer. We design the endorsement policy, promoted for good-to-reuse, certified for the authoritative, documented, change-managed models the business can rely on, and the minimise-and-decouple pattern that ends the "same number means something different in every report" sprawl. One reconciled number, owned centrally, built on once.
• Row- and object-level security design. We design the row-level and object-level security model against your actual roles, for example legal-entity-level row security for a multi-entity Dynamics 365 group, so people see exactly the data they should and no more, enforced through workspace separation so the certified model stays protected.
• Data protection: Purview, sensitivity labels and DLP. We design the data-protection model on Microsoft Purview: how sensitive data is classified, labelled and protected with sensitivity labels, and where DLP policies apply. This is policy and design work we own; where a full Purview implementation is in scope, we lead it and deliver it with the right partner team.
• A right-sized centre-of-excellence operating model. We design the centre-of-excellence operating model on paper, the structure, roles and an accountability matrix, sized to an SME, not an enterprise regime. This is the operating model and the governance charter, not a staffed CoE build; where a staffed build is what you need, we design it and deliver it with the right team.
• The governance roadmap and rollout. We design the rollout the way Microsoft recommends, lightweight and iterative, alongside delivery rather than as a two-year programme before anything ships: a governance and CoE charter, prioritised policies, quick wins, and the KPIs that tell you it is working.

The right licence footprint depends on your estate: Pro per user for standard sharing, Premium Per User where you need premium features for fewer than about 250 users, and a Fabric F-SKU capacity when you need lakehouses, warehouses or free-viewer access at F64 and above, sized to your consumption and costed before you commit, and planned around the retirement of the Power BI Premium P-SKUs to Fabric F-SKUs.

Licensing is the part of the estate the CFO funds, and it is genuinely non-trivial, which is exactly why it belongs in the strategy rather than being left to a renewal date to decide by accident.

The footprint is a real decision, not a default. Pro per user covers standard sharing. Premium Per User is cost-effective when you need premium features for fewer than roughly 250 users, but it does not give you a Fabric capacity, so no lakehouses or warehouses. A Fabric F-SKU capacity is what you move to when you need that, or when you want free-licensed users to view Power BI content, which needs F64 or larger. We size the capacity to your actual consumption and cost it before you commit, so the number is defensible at the board rather than a surprise.

A readiness review gives you an honest verdict on whether your Power BI estate is ready for Copilot: it is ready when your semantic model is well-modelled on a star schema, human-readably named, documented, certified and simplified, because without that preparation Copilot returns less helpful and inaccurate results. You get where you stand and the governed path to get there, a verdict, not a promise to implement AI.

Here is the part most vendors will not say plainly: the governance work and the AI-readiness work are the same work. Microsoft is unambiguous that without a well-prepared semantic model, Copilot generates less helpful and inaccurate results, and the preparation it lists, star-schema design, human-readable names, documented descriptions, removed complexity, endorsement, is exactly the certified, governed model a good framework already builds. The certified semantic model is the AI-ready foundation. Get the governance right and you are getting ready for AI at the same time.

So what we do here is a readiness review. Beat one, what we assess: is your semantic model well-modelled, named, described, certified and simple enough for Copilot to answer correctly; whether Fabric and OneLake are the right platform fit; and where you stand on advanced-analytics readiness. Beat two, the honest verdict: here is where you are, and here is the governed path to get there. The value-pressured CIO and the AI-ROI-sceptical CFO reward the honest read over the hype, and we would rather tell you that you are not ready yet than sell you an AI dream.

A phased engagement: assess, target operating model, governance framework, roadmap, licensing and capacity strategy, and an honest Copilot and Fabric readiness verdict, that lands an executable roadmap and a lightweight, iteratively-implemented governance framework, not a shelf-ware strategy deck.

The objection to pre-empt is "we will get a strategy deck that changes nothing". So the engagement is built the way Microsoft recommends governance is rolled out, iteratively and alongside delivery, and each phase lands something real.

1. Assess

   We build the estate and maturity view: where the sprawl is, where trust in the numbers breaks, and how governed you actually are today, so the plan is grounded in your estate, not a template.

2. Target operating model

   We set where you should sit between self-service and central control, the managed-self-service point that fits your maturity, so the rest of the framework has a clear model to encode.

3. Governance framework

   We design the framework: the tiered workspaces, the certified-model and endorsement policy, the row- and object-level security, the Purview, label and DLP design, and the centre-of-excellence operating model, all as a design you can roll out lightweight.

4. Roadmap

   We turn it into a funded, sequenced plan: the build, managed and readiness steps in the order that de-risks them, so you have a roadmap you execute, not a slide you file.

5. Licensing and capacity strategy

   We set the right licence and capacity footprint for your estate, sized to your consumption and costed before you commit, and planned around the P-SKU to F-SKU transition.

6. Copilot and Fabric readiness verdict

   We give the honest readiness position and the governed path, a verdict, not a promise to implement AI.

The engagement lands an executable roadmap and a governance framework implemented lightweight and iteratively, phased, not a two-year transformation. The scope is transparent and the cost is shown before you commit.

The advice is genuinely independent: it sets the direction and governance across your whole estate and can recommend, or decline to recommend, any tool or build, including our own Reveal. The Copilot and Fabric readiness verdicts are honest too, so if you are not ready we tell you and set the governed path rather than sell you an AI dream.

The sharpest question on any advisory page is whether the advice is genuinely independent or a build pitch wearing advisory clothes. Ours is independent, and it is independent in two specific ways.

First, the advice is not tied to a build. It sets the direction and the governance across your whole estate and can recommend, or decline to recommend, any tool or platform, including PrecisionPoint's own Reveal where it genuinely fits. Reveal is one option the advice may land on, never the foregone conclusion. Second, the readiness verdicts are honesty-gated. If your estate is not Copilot-ready, we say so and set the governed path to get there, rather than sell you an AI outcome we cannot honestly stand behind.

Under both sits the thing a generic governance advisor does not have: two decades of doing nothing but Dynamics data, so the governance is grounded in your finance numbers reconciling, not generic BI hygiene. Independent advice, grounded in the numbers, with an honest read on AI. That is the wedge.

Usually yes: advisory sets the direction and the governance framework, a build then delivers it, and managed BI runs and governs it day-to-day. Advisory can also follow a reporting health-check, which diagnoses the symptom while advisory sets the strategy so it does not recur. You choose the next rung on the evidence.

Advisory is the strategic layer. It sits beside build and managed and is usually the step before them: set the strategy and the governance, then let a build deliver it and a managed run-state run it. It can also follow a health-check, where a diagnostic finds what is broken and advisory sets the strategy so it does not come back. You are choosing the next rung on the evidence, not signing up for a transformation.

1. Advise.

   This page: the strategy, governance framework and roadmap, set once so the estate is governed by design.

2. Then build.

   When you are ready to deliver it, see Power BI build for Dynamics 365 for the governed, reconciled build.

3. Then run and govern it.

   For the run-state, see how we run managed BI governance day-to-day. The boundary is deliberate: here we design the governance framework; managed BI runs it for you as governance-as-a-service.

4. Or start with the symptom.

   If a report is broken right now, start with the reporting health-check, then set the strategy so it does not recur.

For the full programme this strategy sits inside, see Power BI for Dynamics 365. And if you are weighing the platform question, talk to us about whether Microsoft Fabric is right for Dynamics 365.