Power Platform Community Forum Thread Details

I have a Power Automate flow calling a Copilot Studio agent. The "Execute Agent and wait" action fails on every run with a fixed error, while the plain "Execute Agent" action against the same agent, same connection, same inputs succeeds. This looks like the async result path is broken rather than a configuration problem, and I can find no connector action that retrieves a conversation result to work around it.

**Exact error text**

"Enhanced task completion preview has ended. Go to copilotstudio.microsoft.com and republish the agent."

Failing action: Execute_Agent_and_wait, channelId: pva-autonomous

**Proof the configuration is correct**

Plain "Execute Agent" against the same agent returns HTTP 202 with a valid conversation ID:

```
{
"statusCode": 202,
"body": { "conversationID": "3159a48a-e272-455d-b71d-37e85010f639" }
}
```

Response headers confirm correct routing: x-ms-environment-id resolves correctly, x-ms-pva-engine-routing: prd-prod-mt-eu-107-neu.

So the connection, environment ID, agent selection and message inputs are all valid. Only the "and wait" variant fails.

**Environment details**

- Environment type: Default, Region: Europe
- Agent status: Available, Orchestration: Generative
- Agent published successfully (confirmed in Power Platform admin centre)
- Agent has a valid Microsoft Entra agent ID registered
- Authentication: None
- "Allow other agents to connect": enabled
- "Memory (Preview)": disabled

**What I have already tried, with no change**

- Republished the agent multiple times, including after forcing an actual instruction change
- Created a brand new connection between Power Automate and Copilot Studio
- Deleted and re-added the Execute Agent and wait action entirely
- Populated the optional Environment ID parameter explicitly
- Enabled the Demo Website and Teams/Microsoft 365 publish channels
- Confirmed the agent works perfectly in Copilot Studio's own test panel, producing correct output using its knowledge source
- Checked Settings > AI & behavior > Orchestration: no "Enhanced task completion" toggle exists anywhere in this environment's UI
- Verified via run history that the failing action shows no Inputs section at all, suggesting the connector fails before constructing the request

**Key observations**

1. The error instructs me to republish, but republishing has no effect and the feature it references has no toggle in my environment. My environment is not enrolled in the early release cycle, so I cannot access the preview portal setting either.

2. There appears to be no way to retrieve the result of a conversation started by plain "Execute Agent". The connector exposes Evaluate Agent, Execute Agent, Execute Agent and wait, and four "Get Agent Test..." actions, none of which take a conversation ID and return the agent's response.

**Questions**

- Is "Execute Agent and wait" expected to still function following the Enhanced Task Completion preview retirement, or has this capability been withdrawn?
- If withdrawn, what is the supported replacement for retrieving an agent's response inside a Power Automate flow?
- If it should still work, what environment-level setting or flag needs changing, given the toggle is not exposed in a standard (non early release) environment?

Happy to provide correlation IDs or full run history on request.

Categories:

Building flows