AI Agent: Customer identification for forwarded emails to the ticket system

Title: AI-Agent: Customer Identification for Forwarded Tickets

  1. What is your original issue/pain point you want to solve?
    Our team frequently handles customer inquiries that are manually forwarded to Zammad by employees from a central email address. Currently, Zammad identifies customer as the sender instead of the original inquirer. Manually updating the customer for every single ticket is a significant time sink for our agents and is highly prone to human error.

  2. Which are one or two concrete situations where this problem hurts the most?

High Daily Volume: On days with high traffic, agents spend much time just copy-pasting email addresses from the forwared mail bodies into the customer field to ensure the ticket is correctly linked to the original customer.

Context Loss: If an agent forgets to manually change the customer, replies are accidentally sent to the internal colleague who forwarded the mail, rather than the actual customer, causing confusion and communication gaps

  1. Why is it not solvable with the Zammad standard?

While we are aware of manual workarounds, they are not scalable:

Manual “Change Customer”: Requires multiple clicks and manual data entry for every forwarded mail.

Integrating the source mail accounts directly into Zammad is not an option for us. We need the manual “forwarding” step as a first line of defense to prevent spam and irrelevant internal chatter from cluttering the helpdesk.

Standard filters cannot “read” the context of a message. They cannot distinguish between an actual original sender and other email addresses mentioned in the conversation or signatures.

  1. What is your expectation/what do you want to achieve?

An AI agent solution that recognizes the context of messages and can assign the correct customer. It should be possible to specify which email addresses should not be assigned as customers.

If the AI ​​agent is unsure, it should check with the agent to confirm the correct customer or perhaps simply suggest changing the customer to X, which the agent then confirms.

It should also be clear that the AI ​​agent has changed the customer so that the Agent is aware of the change.