What Aura AI is
The AI agent that ships with Auralis, where it lives in the product, and what each page in this guide covers.
Aura AI is the AI agent that comes with Auralis. Aura is the agent; Auralis is the company that makes it. Both names are real and neither replaces the other — a footer reading "Powered by Auralis" is the company, and a message signed "Aura" is the agent.
It is a product of its own
AI Agents ships unbundled from the helpdesk. That is a product decision rather than a packaging one, and it cuts both ways:
- A workspace can run the Auralis helpdesk with no AI agent at all. Nothing in the ticketing product requires Aura to be enabled.
- Aura is built to sit on top of a helpdesk you already have — Zendesk, Freshdesk, Intercom — as well as on this one.
So read this guide as the documentation for a second product that happens to be in the same sidebar, not as a chapter of the helpdesk guide.
Where it lives in the product
The agent has its own section, Aura AI, with four pages:
| Page | What it is for |
|---|---|
| Overview | Turning the agent on, the connection, and where it answers |
| Performance | How the agent is doing — conversations, outcomes, ratings |
| Agent | Tone, persona, role, goals and the words it must get right |
| Workflows | The graph a conversation takes through the agent |
What the agent is allowed to do is configured somewhere else, under Settings, AI agents: Autonomy, Auto-resolve, Categorization and Playbook. The split is deliberate. The Aura AI section is the agent; Settings is the permission you grant it.
Decisions the agent asks a human to make land in Approvals, on the product rail.
What is in this guide
Getting it running
- Turning Aura AI on — the Enable switch, existing agent identifiers, testing the connection, and disabling.
- The Agent page — the seven settings that decide how it writes, and the ones that are not here.
Deciding what it may do
- Autonomy — where Aura may answer alone, where a human approves first, and where it stays silent.
- Auto-resolve — letting it reply and solve a ticket on its own, and the guardrails that all fail closed.
- Ticket categorisation — classifying inbound tickets into your own ticket types.
- The AI playbook — the wording, the questions and the refusals for service-call work.
Working with it, and watching it
- Agent Assist — the agent beside a ticket, drafting for a human to send.
- Workflows — the conversation graph.
- Performance — the ten measures and the four lists.
- Approvals — the queue where a person decides.
Where its knowledge comes from
The help centre and the agent's knowledge base are two different things. Publishing an article does not put it in the agent's index; handing an article over is a deliberate act taken from Knowledge, Content. See Handing articles to the AI.