A support ticket may need a category, an owner and a check of the relevant policy before an agent can answer it. Norml builds AI assistance inside that workflow. An eligible Zendesk ticket receives a reviewable classification and reply suggestion based on approved information. The agent keeps control of the customer-facing response, while unclear questions and failed processing remain visible to the team.
We inspect your existing Zendesk features and triggers first, then scope the additional workflow around the decisions your agents need help making.
We map eligible messages to your ticket categories and escalation rules. Any automatic field updates are limited to agreed fields and conditions. Human-set priorities and assignments remain protected, and conflicting or insufficient information produces a review state instead of a silent guess.
The workflow uses the permitted knowledge content and ticket context to prepare an internal note with a reply suggestion and source references. It explicitly marks the note private. An internal note is not a saved draft in the agent’s reply composer; the agent reviews and transfers the approved wording into the public response.
We track the triggering comment and processing result, exclude the automation’s own notes and recheck ticket state before updating it. Reopened, solved or merged tickets follow agreed rules. Failed writes and unclear requests reach an owner rather than creating repeated notes or marking a ticket resolved.
Share anonymized questions, expected categories and examples that need a person immediately. Define who owns complaints, sensitive requests and cases that cannot be answered from the approved information.
Identify the policies, product guidance and support articles the workflow may use. We need an owner for updates and a rule for missing or conflicting answers. Ticket access must not expose restricted source content to an unauthorized user.
We inspect the permitted API actions, custom fields, groups and existing triggers. Internal notes or field changes can still trigger account automations, so notification behavior must be checked before release.
We compare classifications and proposed replies with decisions reviewed by your support team. Missing evidence, policy conflicts and unsupported questions define where the workflow should hand off.
We build eligible-comment intake, source retrieval and the allowed ticket updates. The note is explicitly private, field ownership is enforced and the processing record distinguishes a completed update from a failed attempt.
We test duplicate events, concurrent agent edits, outdated sources and existing Zendesk triggers. Agents check that suggestions are useful and learn how to report an incorrect answer, pause assistance and inspect failed runs.