Connecting Jira to AI: query the board without learning JQL
Jira can be connected to AI. IntraGPT looks up issues in the Jira environment you point it at and returns the details that belong to the question. On request the assistant creates an issue, updates one or posts a comment. People who do not work in Jira themselves still get an answer to the question of where something stands and who is holding it.
The four things the assistant does
Searching in the language of Jira
The assistant turns your question into a search on project, status, assignee or last change, and keeps the number of results limited. It also retrieves only the details the question needs, rather than every field of every issue.
Creating an issue
A new issue lands in the project with an issue type, a summary and a description. Because issue types differ per project, the assistant first checks which ones exist there before writing anything.
Updating an issue
With the issue key at hand the assistant can set a status, assign someone or fill in another field. That happens on request only, because an updated issue shifts the work of a team that is busy with it.
Posting a comment
The assistant can put a comment under an issue. It is visible to everyone with access to the project, so it shows the text before writing it and keeps internal considerations out.
Working with the issue key
Issues in Jira carry a key such as ABC-123, and that key is needed to update or comment. The assistant never invents one: it looks the issue up first and works with the key that comes out of it.
What teams do with it
Government
- Situation
- An information manager needs to know which reports about a public service process are still open, but does not work in Jira.
- What the agent does
- The assistant looks up the open issues in that project and returns the keys with their summaries.
- Result
- An answer without an extra licence and without a lecture about filters.
Customer service
- Situation
- A customer question really belongs with the development team, but nobody picks it up.
- What the agent does
- After approval the assistant creates an issue in the right project, with an issue type that exists there and a summary that can be understood.
- Result
- No retyping, and no report left lying in a mailbox.
Project management
- Situation
- At the end of the day the project lead wants to know what moved today.
- What the agent does
- The assistant looks up the issues that were updated most recently and keeps the list short.
- Result
- An overview of what is changing, instead of a dashboard full of widgets.
Quality and compliance
- Situation
- During an audit it must be demonstrable what happened to a recorded deviation.
- What the agent does
- The assistant looks the issue up by its key and returns the status, the assignee and the latest comment.
- Result
- The course of events can be reconstructed without the auditor needing access to the system.
Four things to arrange
-
1
Signing in with your own account
The connector uses an Atlassian account whose access you grant yourself and can withdraw again. The rights of that account decide what the connector gets to see.
-
2
Pointing at the Jira environment
You connect to your own Jira environment. That fixes which Jira the assistant talks to and which projects come within reach.
-
3
Releasing writing on purpose
Searching has no consequences. Creating, updating and commenting are released separately, and you decide which roles inside IntraGPT may use them.
-
4
Looking back afterwards
Every action is recorded with the moment, the issue and the outcome. So it stays visible later why a status changed or where a comment came from.
Rights, visibility and the team board
An issue in Jira is internal deliberation: customer names, estimates and sometimes a discussion that is not meant for everyone. That is why it is fixed who may use the connector and what can be written.
Where the data lives: Nederland of de EU
The account sets the view
The connector sees exactly what the connected Atlassian account sees, no more. Projects that account has no rights on stay invisible to the assistant, even if someone asks for them explicitly.
A comment is teamwork, not a scratchpad
What the assistant puts under an issue is read by everyone with access to the project. That is why it comments only on explicit request and shows the text first.
Changing a status moves work
A status change slides a card on the board of a team that works with it themselves. The assistant does not do that on its own initiative, and never just to round off a conversation neatly.
Runs in your own environment
The model sits on our own server in the Netherlands. Summaries, descriptions and comments from Jira are not sent to an American model provider and are not used for training.
Frequently asked questions about the Jira connector
Why does my search return no results?
Usually because the search does not quite match how this project is set up. A project name spelled differently, a status that is not called Done here but something else, or a field that does not exist in this project: Jira then politely returns zero results without complaining. That is why the assistant first checks which statuses and issue types exist in the project and tunes the search to them.
Can I connect Jira to ChatGPT?
There is no built-in connection between ChatGPT and your Jira, and putting the credentials of an Atlassian account into a public chatbot is unwise. IntraGPT offers the same way of asking, but on your own Jira environment, with the rights of one account and with every action recorded.
Can the assistant close an issue?
Updating is possible, so changing a status is too, but only if that has been released and you ask for it explicitly. What closing means differs per project, so the assistant reads the statuses that project actually knows instead of guessing. A card never slides because someone mentioned it in passing.
Does IntraGPT work with Jira?
Yes. Jira is in the IntraGPT connector catalogue, with searching, creating an issue, updating an issue and commenting under one. The connector works inside your own Jira environment.
How is this different from GitHub issues for the assistant?
GitHub issues belong to a repository and are mostly about the work of developers. Jira is the board of the wider organisation, with projects, issue types and statuses configured per project. If you use both, you set up both connectors: Jira tells you where something sits in the planning, GitHub what is happening in the code.
Do I need to know JQL to use this?
No, you ask in plain language and the assistant builds the search. It does help to say which project it concerns, because without a boundary the outcome is too large to work with.
Does the assistant invent issue numbers?
That is precisely ruled out. A key such as ABC-123 is needed to update or comment, and the assistant takes it from a search. If it cannot find the issue it says so, instead of inventing a number that happens to exist.
How long does setup take?
Connecting itself takes a few minutes. The rest of the time goes into choosing which Atlassian account is used, because the rights of that account decide what the assistant sees and may do.
What does a Jira connector cost?
It belongs to the project we run together. What drives the work is the number of sources being connected, the rights that come with them and whether the assistant may update as well. We go through that in a thirty minute call.