For AI agents: the complete documentation index is available at https://docs.clickmax.io/en/llms.txt, the full documentation bundle is available at https://docs.clickmax.io/en/llms-full.txt, and this page is available as Markdown at https://docs.clickmax.io/en/features/max/manage-automation-runs-with-ai.md.
  • English

      • How do I ask Max to test, run and follow an automation?

      Open a chat with Max and say what you want to know or do with the automation, by name: "test the Welcome automation with my contact", "run the Black Friday automation for the Customers list", "why is the Abandoned cart automation failing?", "which message of the cadence is Maria on?". Max does what you would do with the editor's Test and Run manually buttons and the Executions tab, and answers with contact and block names, not codes.

      Before you start:

      • Role: any member with access to the workspace. Max acts with your permissions. Starting, cancelling or moving up a run requires edit access to automations.
      • Your own AI (Claude, Cursor, Codex…) can make the same requests: see Connect your AI to Clickmax (MCP).
      • Max is in Beta: check the result on the Automations screens before you decide.

      Test the automation with a contact

      Ask for the test naming the automation and the contact: "run the Welcome automation for the contact Maria". Max confirms who will receive the messages and only then runs it.

      The test is the same as the editor's Test button, and it runs for real:

      • it sends WhatsApp, email and SMS to the contact, applies tags and calls integrations;
      • it honors waits: a "wait 2 days" step holds the test for two days;
      • it counts in the automation's metrics and shows in the Executions tab with the Test badge;
      • it works with the automation in draft and runs the saved version.

      Prefer one of your own contacts. The test starts directly at the chosen contact and does not prove the real event fires the automation; that requires publishing.

      After it runs, ask "show me step by step what happened in the test": Max tells you which path the execution took, what each step did and why a condition came out false. To reuse the data from a real trigger, ask "use the last time the automation fired as the base".

      Run the automation for a list, tag, segment or contacts

      Ask naming the automation and who should enter: "run the Black Friday automation for the Customers list", "run Reactivation for the Inactive tag tomorrow at 9am", "send Welcome to the October leads segment". It is the same as the editor's Run manually: each request becomes a run. See How do I run an automation for a contact, list, tag or segment?.

      1. Max checks the automation. It must be published; if it is a draft, Max asks whether you want to test it with one contact or publish it first, and does not publish it on its own.
      2. Before running, Max tells you how many contacts will enter and how many are on cooldown (they went through the automation recently and are left out).
      3. Max asks whether contacts on cooldown should enter anyway, and whether to start now or schedule it for a date and time.
      4. You approve and the run starts (or becomes Scheduled).

      The run executes for real for everyone in the source: it sends the messages, applies tags and calls integrations. The audience is locked when the run starts: anyone who joins the list, tag or segment afterwards does not receive it. For a scheduled run, that happens at the scheduled time.

      To follow it, ask "how is the Black Friday automation run going?": Max tells you the status (Scheduled, Running, finished, Cancelled or Failed, with the reason), how many contacts entered, how many the cooldown skipped and how many failed, and who started it. Ask "when will it finish?" for the time of the last cycle, "who failed in that run?" for the list of contacts, and "move up the next cycle" to put a pending cycle in the queue now.

      To cancel, ask "cancel the Black Friday automation run". What has not gone out does not go out; contacts who already entered keep going through the flow, and the automation stays published. Max asks for your approval before starting, moving up or cancelling a run.

      Limits that also apply through Max: one run at a time per automation (a scheduled run counts too), up to 500 contacts picked one by one, and automations with steps that depend on the trigger event (such as replying to an Instagram comment) cannot be run.

      Understand why executions failed

      Ask "how many executions of the Abandoned cart automation failed, and why?". Max summarizes:

      • how many executions are in each state (running, finished, failed, stopped);
      • the top errors, most frequent first, with the number of contacts in each;
      • for one execution, which step it stopped at and the reason, or whether it is waiting for a deadline or a reply from the contact.

      You can filter in the request: "only the ones that failed at the WhatsApp step", "Maria's execution", "only the tests". Unless you ask, Max looks at real executions and leaves tests out.

      For a test execution Max shows the full detail of each step. For a real execution it shows the order of the steps and the error of the step that failed.

      Retry or stop

      • Retry one execution: "retry Maria's execution". It runs again from the step where it failed.
      • Retry all executions of one error: "retry everything that failed with this error". Max tells you how many contacts that touches and retries in batches of up to 1000; if some are left, it tells you and asks before the next batch.
      • Stop one execution: "stop Maria's execution". No later step runs and it cannot be resumed; a step already in progress may finish what it started.

      The step that failed runs again: if it sends a message, the message goes out again. That is why Max asks for your approval before each of these actions and fixes the cause first when it is known. Retrying without fixing the cause repeats the failure. Stopping an execution does not pause the automation: new contacts keep entering; to stop that, ask Max to pause it.

      An archived automation does not accept retries. A paused one does.

      Check before publishing

      Ask "check the Welcome automation before publishing". Max lists the problems by severity, each pointing at the block:

      • Errors: no entry trigger, a block with no destination, a channel with no number or sender, a WhatsApp template that is not approved, a deleted tag, stage, offer or list, and a sending number that is archived, deactivated or has automations turned off.
      • Warnings: a block that is not connected, a flow that ends at a block, an empty condition branch, a trigger missing its required filter, and a message that uses a variable the trigger does not provide (it would go out empty).

      A warning does not stop you from publishing. Errors about a number, a template or a deleted reference do not block publishing either, but the send fails later: fix them first.

      Ask "which automations start when I apply the VIP tag?" or "who applies the tag that starts the Welcome automation?". Max lists the automations on the other end of the link, with each one's status (draft, active or paused) and how many executions it has had.

      It covers a tag applied or removed, entering a stage, an opportunity won or lost, and an email delivered, opened, clicked or failed, a link click, a call result and a webhook. Only links with a fixed choice (a specific tag or stage) show up: if a step picks the tag another way, Max cannot see the link. Drafts and paused automations are listed but do not run.

      Find out which automation message a contact is on

      Ask "which automation message is the contact Maria on?" or "what is the next cadence message for the Company X opportunity?". For each automation with messages the contact went through, Max answers:

      • the message the contact is on right now (for example, waiting for a reply) or the last one received, with its delivery (sent, delivered, read, clicked or failed);
      • the next message and when it goes out;
      • the ones still ahead and the ones left out because the contact took another path.

      When the next message depends on the contact (replied or not, button chosen), Max tells you both paths. The answer covers up to the 50 most recent executions and leaves tests out, like the opportunity's Cadences tab: Which automation message is the opportunity contact on?. To take the contact out of the sequence, ask Max to stop that contact's execution.

      If it did not work

      • Max says it could not find the automation or the contact: use the exact name, or the contact's email or phone. Contacts from another workspace are not found.
      • "Nothing to stop": the execution already finished or was already stopped. "Cannot be retried": the execution did not fail (only failed executions can be retried) or the automation is archived.
      • The test does not start: the automation needs at least the trigger block.
      • The run does not start: the automation is not published, already has a scheduled or running run (cancel it or wait for it to finish), the source has no contacts, the WhatsApp number is down, or a message block has no sender.
      • A scheduled run shows as Failed or Cancelled: something prevented the send at the scheduled time (for example, the automation was paused). Ask Max for the reason, fix it and ask for a new run.
      • An action was blocked by an error: see the Error codes.