Overview and setup
What the assistant does, who configures it, and how to get it live on one service desk.
What it is. A chat window in your customer portal. The customer describes their problem in their own words. The assistant answers it, or creates the right request for them.
Why you would want it. Customers do not know which of your request types matches their situation. So they pick the closest one and leave half the fields empty. Your agents then spend two replies asking for information the form should have collected. The assistant asks those questions up front, and does not create a ticket at all when the answer is already in your knowledge base.
Who sets it up. A Jira administrator. Individual service desk teams can add temporary notices themselves, which is covered in Operational notices.
What the customer sees
The assistant is offered in the portal header — on the Help Center above your normal request types, and on the individual portals underneath it — plus an entry in the customer's own menu. Nothing else about the portal changes.
Clicking Start a conversation opens the chat window, which explains what it can do before the customer types anything.
What it can do in a conversation
- Answer from your knowledge base
- It searches the Confluence spaces you choose and answers before offering to create anything.
- Choose the request type and fill it in
- It picks the right request type and shows its fields as a small form inside the chat, so the customer never browses your portal menu.
- Accept file attachments
- Up to 5 MB per file and 10 MB per conversation. Only when the request type accepts attachments; if it does not, the assistant says so instead of silently dropping the file.
- Ask for confirmation
- It shows what it is about to create and waits for a yes. You can switch this off, but we recommend leaving it on.
- Collect feedback
- A thumbs up or down at the end of the conversation, which feeds the statistics.
ITSD-8841 — Report a hardware or connectivity problem · Click to follow upWhat you need before you start
- Jira Service Management, with at least one service desk that customers can reach through the portal.
- Jira administrator rights.
- Confluence with at least one space of customer-facing articles, if you want knowledge base answers. This is optional but it is where most of the value is.
- Nothing else. There is no API key to buy: the AI runs on Atlassian's own Forge LLM service.
Set it up on one service desk
Do not enable every desk at once. Pick the busiest one, get it right, then copy the pattern.
- 1Install the app, then open Jira → Apps → AI Portal Chat.
- 2Go to the Portal Assistant tab. In the Company Instructions card, set Ticket & Knowledge Base language and write your Custom instructions for AI Portal Chat — tone, and anything the assistant must never promise. Two or three sentences is enough. Click Save Instructions.
- 3In the Service Desk Configuration card, pick your service desk and switch Enable AI Portal Chat on.
- 4Leave Auto-confirm tickets off. Customers then see what will be created before it happens.
- 5Switch Enable Knowledge Base Search on. A Confluence Space Keys field appears: type the space keys to search, separated by commas, for example
KB, IT. - 6Add service desk instructions for this desk: what it covers, and any rule an agent always repeats, such as an approval requirement.
- 7For your two or three busiest request types, add request type instructions. Write the questions your agents always end up asking.
- 8Go to the Branding tab, choose the trigger layout, set your colours and give the assistant a name.
- 9Open the portal as a customer and have a real conversation with it.
Where each screen lives
| Screen | Where | Who can use it |
|---|---|---|
| The chat itself | Your JSM portal: the header, plus an entry in the customer's menu | Customers |
| Statistics, Portal Assistant, Branding, API, Audit Log | Jira → Apps → AI Portal Chat | Jira administrator |
| Agent Instructions (temporary notices) | The service desk project → Agent Instructions | Project administrator |
Next: Instructions and knowledge base, which is where you actually control how the assistant behaves.