Power Platform Center of Excellence: An Operating Model, Not a Toolkit
Here's a pattern we see often enough to name. An organization decides to get serious about Power Platform governance. Someone finds Microsoft's CoE Starter Kit, deploys it over a sprint, and the dashboards light up: hundreds of apps, thousands of flows, orphaned resources everywhere. For two weeks, it feels like control.
Six months later the Starter Kit is a dashboard nobody opens. The sync flows need patching, the compliance process was never turned on because nobody agreed on the rules it should enforce, and the sprawl it so vividly displayed is still sprawl — now with better charts.
Nothing failed technically. What was missing was the actual Center of Excellence: the people, decisions, and operating rhythm the tooling was supposed to serve. A CoE is not something you install. It's something you run. That's even clearer now that Microsoft no longer actively maintains the CoE Starter Kit and has moved its core capabilities into the Power Platform admin center. The dashboards are part of the product now. The Center of Excellence still isn't.
Illustrative screen. Names and figures are examples, not a client.
The utility, not the wiring
That's the Center of Excellence: the utility that operates the grid.
On our governance page we make the case that you don't ban electricity — you build a grid: wiring standards, circuit breakers, meters. Governance policies, DLP rules, and environment strategy are that wiring. But wiring doesn't run itself. Somebody maintains the standards, reads the meters, answers the calls, connects new construction, and decides the cases the rules didn't anticipate.
That's the Center of Excellence: the utility that operates the grid. Concretely, a functioning CoE does a handful of things on a schedule:
Sets and evolves the standards
— which connectors are open, what "production-ready" requires, when a personal tool must graduate to managed status. Standards that never evolve become either ignored or obstructive.
Runs intake and promotion
— a known path for "I built something useful, now what?" so good solutions get hardened instead of lingering in the shadows.
Watches the estate
— inventory, usage, orphaned resources, risk signals. Not to surveil makers, but so the organization learns about critical apps before their builder resigns, not after.
Grows the makers
— training, office hours, patterns and templates, an internal community. This is the part lockdown-style governance never budgets for, and it's the part that pays.
Decides the hard cases
— the app that touches regulated data, the flow that crosses departments, the agent that wants write access. Rules handle the routine; the CoE handles the rest.
None of that is software. All of it can be supported by software, which brings us to the kit and what has replaced most of it.
The CoE Starter Kit, now that Microsoft has stepped back
Microsoft's CoE Starter Kit is a free, substantial set of solutions built on the platform itself: inventory sync, dashboards, compliance workflows, maker onboarding components. Microsoft no longer actively maintains it. Issues aren't reviewed or addressed, it won't get new capabilities, and it remains available for existing and new deployments. So the question has shifted from "should we install it?" to "what, if anything, should still be running?" We know the kit well, and our advice on it is situational, because we have no incentive either way: our recommendations aren't driven by software sales or license quotas.
What the admin center covers now.
Inventory of apps, flows, and agents across the tenant, including agents built in Copilot Studio and Microsoft 365 Copilot. Usage and top owners. Operational health through Monitor. Governance recommendations through Actions. Managed Environments add sharing limits, environment groups, weekly usage insights, and pipelines. It's supported, it ships with the product, and nobody on your team has to patch it.
What the kit is now.
An unmaintained solution with admin rights. It runs under an identity that must hold Power Platform admin or global admin permanently, and it's dozens of flows and apps that nobody upstream will fix when something changes. It is, with no irony, an estate that needs governing. When its sync flows fail quietly, the inventory goes stale, and a stale inventory is more dangerous than no inventory, because people trust it.
What neither one does.
Decide. The kit displays, and so does the admin center. Every meaningful view poses a question (who owns this, is this compliant, should this exist) that a human process has to answer. Microsoft now ships the meters. Running the utility is still your job.
Our advisory position, plainly:
- If you're standing up a CoE now, start in the admin center. It covers the visibility most organizations went to the kit for. Managed Environments are an entitlement of standalone Power Apps, Power Automate, Copilot Studio, Power Pages, and Dynamics 365 licenses, so check how your makers and users are licensed before you build your plan on them.
- If the kit is already running, triage it rather than ignoring it or tearing it out: confirm it still syncs, move inventory and usage reporting to the admin center, and keep only the components that fill a gap you can name, each with a named owner.
- If you have a gap the admin center doesn't fill and the platform capacity to own the fix, a scoped kit component can still earn its place. Treat it as your own solution from here on, because it is.
- If a vendor or partner proposes the Starter Kit as your governance strategy, that's the tell to keep interviewing. It was a component of one at most, and now it's a component you'd maintain alone.
- In every case: design the operating model first, then choose tooling to fit it. Tooling-first CoEs produce dashboards. Model-first CoEs produce decisions.
What we actually build with you
An IMP0WER CoE engagement is organizational design with technical scaffolding, in that order. Following our GRID methodology — Gauge, Route, Install, Distribute — it typically covers:
The mandate.
What the CoE decides, what it advises on, and what it deliberately leaves alone. A CoE that tries to review everything becomes the bottleneck it was created to remove.
The people.
Who sits on it — platform admin, security, business unit representatives, a maker voice — sized to your reality. In a mid-sized organization this is a rhythm and a set of hats, not a department. We design for the staffing you have, not the org chart of a Fortune-100 program office.
The rhythm.
What happens weekly (intake, exceptions), monthly (estate review, standards), quarterly (strategy, retirement pass). Cadence is what separates an operating model from a committee that met twice.
The processes.
Intake, promotion criteria tied to our published Ø Standard, exception handling, the escalation path for hard cases, and maker enablement — office hours, patterns, training paths.
The tooling, last.
Admin center and Managed Environments capabilities first, Starter Kit components only where they fill a named gap and someone owns their upkeep, third-party tools only where scale or compliance demands them. Tooling recommendations come with their ongoing cost of ownership attached in writing.
A founder leads the engagement personally, extended by the IMP0WER delivery team. And because we work both the app layer and the content layer, your CoE's standards can cover the SharePoint foundation your apps and agents actually stand on, not just the apps themselves. As agents arrive on the platform, the same operating model extends to them; that's by design, not luck.
Where to start
Not with an org chart — with facts. The Power Platform Health & Governance Gauge ($3,500 focused / $7,500 extended) inventories the estate, maps risk, and scores maturity against the GRID stages. The result tells you what your CoE will actually be governing, which is the only sane basis for designing it. If you engage us for the build, 100% of the focused-tier fee is credited toward any implementation engagement of $25,000 or more signed within 90 days of your readout. We respond within one business day.
Frequently asked questions
Do we need a full-time team to run a CoE?
Almost certainly not at first. Most mid-sized organizations run an effective CoE as a defined rhythm across existing roles — a platform owner, a security stakeholder, business representatives — with clear time expectations. What kills CoEs isn't headcount; it's undefined mandate and no cadence. We design for the staffing you can sustain, and we'll say so if your ambitions exceed it.
Is the CoE Starter Kit deprecated?
Microsoft doesn't use that word. Its documentation says the kit is no longer actively maintained: issues aren't reviewed or addressed, it won't be enhanced with new capabilities, and it remains available for existing and new deployments. Its core capabilities are now part of the Power Platform admin center, which Microsoft encourages organizations to use as the central experience for governance, monitoring, and insights. The label matters less than the consequence: whatever kit components you run, you maintain.
Should we still deploy the CoE Starter Kit?
For a new CoE, start with the admin center instead. Inventory, Usage, Monitor, and Actions cover the visibility most organizations deployed the kit for, and they're maintained as part of the product. Deploy a kit component only for a gap you can name, scoped to that gap, with someone on your team who can maintain it indefinitely. Which gaps you actually have is a question the Gauge answers with evidence.
We deployed the Starter Kit and nobody uses it. Where do we start?
With the operating model, not the kit. The problem is rarely the tooling: the kit assumed a mandate, owners, a cadence, and decision processes for its dashboards to feed, and nobody built them. Build those first. Then decide which views the model needs and get them from the admin center where you can. Any kit components you keep get an owner and a review date. That's a common starting point; you're not behind, you're mid-sequence.
What's the relationship between the CoE and IT?
The CoE isn't a rival to IT — it's the joint table where IT's accountability for safety and the business's appetite to build meet on schedule. IT typically anchors platform administration, security, and ALM standards; the business anchors priorities and ownership. The design work is drawing that line explicitly for your organization.
Does the CoE govern AI agents too, or is that separate?
The same model, extended. Agents built on the platform are estate like everything else — they need owners, intake, standards, and monitoring, plus autonomy-specific controls as they move from answering questions to taking actions. The admin center's inventory already lists agents from Copilot Studio and Microsoft 365 Copilot alongside apps and flows, so the estate view is already one view. Building a second, separate governance structure for agents is how organizations end up with two half-run grids. One utility, whole estate.
How is this different from just hiring you to write a governance policy?
A policy document is an artifact; a CoE is the mechanism that keeps artifacts true. We write plenty of policy in these engagements — DLP design, standards, promotion criteria — but every one of them is attached to an owner and a cadence before we leave. Documents don't govern. People operating a rhythm do. That conviction is the whole page you just read.