Power Automate connector for Customer Insights

What it does

A Power Automate connector that fires cloud flows on Customer Insights events — a data source refresh finishing or failing, a scheduled or unification refresh completing, a segment or business measure crossing a threshold — and offers a small set of actions for reading items back out. It's the operational glue for alerting and light orchestration around the refresh cycle.

Key facts

The connector is premium in Power Automate and Power Apps, and standard in Logic Apps. Everyone touching the flow needs an appropriate premium licence.
Documented triggers: data source refresh succeeded, data source refresh failed, segment threshold crossed, business measure threshold crossed, full scheduled refresh completed, and unification process refresh completed.
Threshold triggers only fire when the value crosses above the threshold. Nothing fires on the way back down.
The business measure trigger works on measures without a dimension. A dimensioned measure won't drive it.
The full refresh trigger only responds to scheduled refreshes. A manually started refresh does not raise it.
Throttling is 100 calls per connection per 60 seconds. Page with $skip if you're reading volume.
Set-up is from Settings > Connections in Customer Insights, on the Power Automate tile, rather than from inside Power Automate.
The connector's actions include Get entities, Get item, Get items from an entity and partition retrieval, with several — including Create item and Update item — still marked preview on the connector reference.
Not available in US Government (GCC High), DoD or the China cloud.
Microsoft's own recommendation is to build against Dataverse in Power Automate as the default, and use this connector for the event triggers it uniquely provides. The Customer Insights docs still describe this connection as preview.

When to use / skip

Use it for exactly one thing: reacting to Customer Insights lifecycle events. Post to Teams when a source refresh fails, email the data owner when a segment crosses a size the business cares about, kick off a downstream job when unification completes. That's real value and there's no other clean way to get it. Don't use it to read profile or segment data — the Dataverse connector against the msdynci_ tables is faster, better at filtering, not throttled at 100 calls a minute, and doesn't need a preview-flavoured connector in the middle. And don't reach for the Create item and Update item actions at all; writing back into Customer Insights from a flow is not a pattern you want to defend later.

Configuration decisions

Which events genuinely need a flow, versus which are better handled by the built-in data refresh notifications.
Which account owns the connection, because a flow tied to a departing consultant's account is a go-live outage waiting to happen. A service account with the right Customer Insights permission is the answer.
Whether segment and measure thresholds are the right signal at all, given they only fire one way and need a fixed number agreeing with the business.
Where the flow reads its data from once triggered — almost always Dataverse, not this connector.
Premium licence coverage for flow owners and, depending on how the flow is invoked, for users.
Whether the orchestration belongs in Power Automate at all, or in Azure Data Factory or Fabric pipelines if the downstream work is heavy.

Gotchas

Threshold triggers going one way only is the thing that catches people. "Alert us when the segment drops below 500" is not something this does, and the client will assume it does.
Manual refreshes not raising the full refresh trigger means your testing during build, where everyone refreshes by hand, will show nothing happening. Test against a scheduled run.
A dimensioned business measure silently fails to drive the trigger. There's no error, just no flow run.
The 100 calls per 60 seconds limit is per connection, so several flows sharing one connection compete for the same budget.
Premium licensing gets discovered at go-live when the flow moves to a service account nobody bought a licence for.
Connection ownership and the underlying Customer Insights permission are separate things. A connection can look healthy while the owning identity has lost Viewer access, and the flow just returns nothing.

Consultant notes

Sell this as monitoring, not integration. It's the answer to "how will we know if the overnight refresh broke", and that's a question every client eventually asks.
Set up the refresh-failure flow into Teams on day one of build. Your own delivery team benefits from it more than the client does at that stage.
If someone proposes a flow that loops over profiles through this connector, redirect to the Dataverse connector before it gets built. Retrofitting after a throttling incident is unpleasant.
Before go-live, move every flow onto a service account, document the connection, and check the premium licence is assigned to it.
Keep a note in the handover pack of which events fire and which don't — the one-way thresholds and the scheduled-only full refresh will be a support ticket otherwise.

Worth another look when the connector's preview actions settle, or if the threshold triggers ever gain a downward direction.