Messages marked as not delivered

Plan requirement

App pluscloud AI Agent Preview
Subscription Suite Team or higher, Support Team or higher
Access Client admin, Client editor, Client user

When a channel would reject an AI agent message, the preview dims it and adds a red note with the reason. This is what the note means and how you fix it.

What you see

The message is shown greyed out, with a red note under it that starts with Not delivered. or Probably not delivered., followed by the reason. Click the note to open the message in the inspector, where the same finding is listed under Checks. The badge on the Inspector button counts the messages with a warning or a delivery problem.

A WhatsApp preview in which the last message is greyed out. The red note under it reads Probably not delivered. WhatsApp rejects reply buttons longer than 20 characters: Book a service appointment has 26, followed by the note that Zendesk documents this limit for AI agents on WhatsApp and that a longer label can cause the Technical error reply. The inspector next to it shows the same finding under Checks, and a channel note for WhatsApp.

Not delivered or probably not delivered

  • Not delivered: the channel or Sunshine Conversations states that it rejects such a message. The customer does not receive it.
  • Probably not delivered: the limit is documented, but what happens above it is not. Expect the message to fail.

Every limit the preview checks, per channel and with its source, is in Channel limits the preview checks.

Fix the message

  1. Read the reason in the note. It names the limit and, for a button or a list option, the label that is too long.
  2. Change the reply, the button or the carousel in the dialogue builder of your AI agent: shorten the text, split a long reply over two messages, shorten the label, or use fewer buttons.
  3. For Viber, check the display name of the AI agent in the AI agent settings. Viber does not send a message whose sender name is longer than 28 characters, so a long name stops every message.
  4. Click Restart conversation and run the test again on the same channel.

A limit can differ per channel. A reply that is fine on the Web Widget can be too long for Instagram or SMS, so test on every channel your AI agent uses. Which channels that are is in Which channels an AI agent can work on.

Checks on the AI agent's own output

Under Checks the inspector also shows warnings that apply on every channel:

  • Quick replies combined with link or postback buttons in one message. AI agents then answer with the Technical error reply, as Zendesk describes in Technical errors in AI agents.
  • More than 10 buttons: only the first 10 are delivered. Zendesk sets the maximum of 10 buttons in Using the customer message block in conversation flows for AI agents.
  • A button without a label, or a link that is not a valid http or https address.
  • A parameter that was not filled in. Customers would see the placeholder text.
  • Text after a template shorthand, which is dropped.

The limit of 20 characters for a button label is not one of these checks. Zendesk documents it only for AI agents on WhatsApp, in Using the customer message block in conversation flows for AI agents, so the preview checks it as a WhatsApp channel limit, as in the example above. See Channel limits the preview checks.

The Technical error reply is one of the system replies, described in Viewing and editing system replies.

Turn the check off

In Preview settings, turn off Simulate channel limits. The notes disappear and the messages are shown normally. The option is not shown for email, where nothing is checked. Turn it off for a demonstration in presentation mode if you do not want your audience to see the notes.

See also

Was this article helpful?

0

Still stuck?

Our support team will take a look with you.

Comments

0 comments

Please sign in to leave a comment.