PowerLeap Logo

Copilot Studio Friction

Get notified when this record changes

One email when the status or the fix changes — double opt-in, no tracking, unsubscribe in every email.

Teams shows "SystemError" although the agent answered correctly

Mitigatedsince 8 July 2026

Last verified

Details & related

Assessment

Confidence
CorroboratedMultiple independent sources describe the same behaviour.
Severity
Blocking
Typical time lost
Hours

Identification

Publish & channelsMicrosoft Teams

Verification & changes

  1. Change

    Pipeline: new evidence from r-copilotstudio — review required.

  2. Verified

    Doc check by human: Source threads and dossier section B2 re-read during seed migration; stale persisted channel version behavior still reported as of mid-2026.

  3. Change

    Provisionally approved by the Product Owner; external LLM quality review pending.

  4. Change

    Initial record created from the seed dossier (migration wave 1).

Related

Are you in the right place?

  • Your agent worked in Teams before, or works in the test pane right now.
  • In Teams, every question now returns Sorry, something unexpected happened... Error code: SystemError.
  • The "Activity" view in Copilot Studio shows that a correct answer was generated.
  • The error hits every user in Teams, not only you.
  • You changed the agent or its channel configuration before this started.

If instead Teams answers, but with outdated content → see Teams answers with an old version of your agent. If instead the agent returns nothing at all → see Agent gives no answer and no error message. If a flow inside the agent times out → see Flow stops with "FlowActionTimedOut".

What's happening

Teams does not talk to your draft agent — it talks to a persisted copy created when you publish. That stored channel version can go stale or invalid while the agent itself stays healthy. Your agent still generates the correct answer — you can see it in the "Activity" view. But Teams cannot deliver it and shows SystemError instead. Think of a restaurant where the kitchen cooks every order, but the serving hatch is jammed. Teams also enforces requirements the test pane never checks, such as a greeting topic and app permission policies. A gap there produces the same broken delivery.

For technicians

Persistent channels such as Teams pin a published agent version; a republish forces creation of a new one. The source thread shows the "Activity" view with the generated answer while Teams shows the error. Reported Teams-side prerequisites: channel configuration, the "Conversation Start" system topic, app permission policies, and connector permissions. None of these gaps get a specific message — everything collapses into SystemError.

How to fix it

Solution 1

Community workaround

Republish to force a new channel version

  1. In Copilot Studio, select "Publish".
  2. Confirm the publish dialog.

✅ You should now see: a confirmation that publishing completed, with a fresh timestamp.

  1. In Teams, close the conversation with your agent.
  2. Start a completely new conversation.
  3. Ask a test question.

✅ You should now see: a normal answer instead of the SystemError message.

Solution 2

Community workaround

Check the Teams channel prerequisites

Use this when the error survives a republish.

  1. In Copilot Studio, open "Topics".
  2. Confirm the "Conversation Start" system topic exists and is enabled.
  3. Open "Channels" and confirm "Microsoft Teams" is connected without warnings.
  4. Ask your admin to check the Teams admin center. The agent's app must be allowed under "Teams apps" → "Permission policies".
  5. Check that every connector the agent uses has a working connection.
  6. Publish again after any change.

✅ You should now see: the agent answering in a brand-new Teams conversation.

Check that it worked

Ask your agent a question in Teams, in a brand-new conversation — not in the test pane. Expected: a normal answer, with no SystemError. The "Activity" view and the delivered answer now match.

If it didn't work

  • Propagation is still running. Publishing updates Teams asynchronously, and delays are common. See Teams answers with an old version of your agent — suspect number one after any publish.
  • You retested in an old conversation. Old sessions can keep the broken state. Start a new conversation for every retest.
  • An admin policy blocks the app. No republish helps until the Teams app permission policy allows your agent.
  • The "Activity" view shows no answer either. Then the failure is inside the agent, not in Teams delivery. Note the exact error code and start from there.

Prevent it next time

  • Republish after every channel configuration change, following the publish ritual.
  • Judge every change in the published Teams channel, not only in the test pane.

Evidence

  • Community threadlearn.microsoft.com

    Makers report Teams serving a stale persisted agent version until a republish forces a new one.

  • Community threadlearn.microsoft.com

    An independent maker documents a published agent failing in the channel while the identical agent works in the test pane.

  • Official documentationlearn.microsoft.com

    The system topics reference documents the Conversation Start topic that channels such as Teams run when a conversation begins.

  • Community threadreddit.com

    User reports agent works in test pane but fails to respond in Teams chat despite successful app addition, matching the pattern of Teams-specific communication failures.