“Send Summary Email to Owner” action not sending emails

[lumora]

Hey Pickaxe team,

Reporting a persistent bug with the Send Summary Email to Owner action. It has not been working for a long time across my bots - no summary email is ever sent after a conversation ends, regardless of what I do. I have changed the trigger action many times, tried different emails and it never ever sends an email to me.

What I have tried:

• Disconnecting and reconnecting the action multiple times

• Re-entering the owner email and changing emails

• Saving and re-saving the configuration

• Changing multiple trigger action variants

  None of these fix it. The action simply never fires. The trigger prompt is set correctly (conversation appears to have finished) and the email field is populated.

This is blocking a core use case for me - I have created a feedback form using a pickaxe form AI - the goal is after any feedback is submitted - I should receive a summary from the AI - yet I never get any emails so this is quite useless.

Attaching a screenshot of the current config for reference.

Please let me know how and when this is fixed.

[abhi]

Hi @lumora,

Sorry to hear this has been so persistent. I’m taking a closer look from our side. I did a quick test on my end and the summary email seems to be working in my setup, so let’s figure out what’s different here.

Quick question, has this ever worked for you before, or has it never triggered at all?

Could you also share your Pickaxe link (or deployment link) and the email associated with your account? I’ll try to replicate the exact setup and see what might be blocking the action.

Really appreciate you sharing all the steps you’ve already tried. That helps a lot.

[emmabrooks]

Hi @abhi can you confirm if that ‘Send summary email to owner’ action would ever trigger with the prompt ‘Send an email summary of the user’s feedback to the owner after every message’ i.e. the prompt that @lumora is using? My understanding is that we had to use something along the lines of:

’When a conversation appears to have finished, ask the user what their email is. Then, once they’ve given it to you, use this action to send a summary of the conversation to the user.’

If the prompt ‘Send an email summary of the user’s feedback to the owner after every message’ works that would be great, as this is something that is lacking in Pickaxe, as far as I’m aware - some way to see the conversations happening without having to go into Pickaxe and copy and paste them into another source.

[abhi]

Hi @emmabrooks, thnx for the question.

The phrase “when a conversation appears to have finished” is where things usually get tricky. In a chatbot, there’s no real end signal. People just stop replying or close the chat, and the AI has no way of knowing the conversation is actually done. So triggers based on that tend to be unreliable.

Also, wording like “after user feedback” can confuse the model. It is not always clear what counts as feedback. Is it every message, only opinions, or only when the user explicitly says something is feedback? That ambiguity makes the action harder to trigger consistently.

It works much better when the user clearly says something like “that’s all” or “conversation ended” in the chat. That gives the model a clear signal to act on.

One more thing that often gets overlooked. The trigger prompt does not work in isolation. It depends a lot on your main role prompt. If the role prompt itself has strict or limiting instructions, it can conflict with the trigger.

For example, if your role prompt says “do not answer anything beyond XYZ” and your trigger says “when conversation is finished, send email,” the model may never reach that step. It follows the role prompt first, so both need to be aligned.

For the prompt about sending an email after every message, it might trigger, but in practice it can be inconsistent. Actions depend a lot on how clearly the trigger prompt is written and how the model interprets it in context.

Also worth keeping in mind from a practical side. Every action trigger uses credits, and more complex tasks cost more. If someone has a long conversation with 50+ messages, using an expensive model plus actions on every message can add up quickly in cost.

There is also a speed tradeoff. Triggering an action after every message can slow down responses and create a bit of lag. So it becomes a balance between getting instant replies vs receiving every message via email.

One helpful thing here is that full conversation transcripts are already available inside the workspace. So if your main goal is to review conversations, you may not even need an action for that.

Happy to take a look at your setup if you want to refine it together