Quorum
Start free trial

Test case design for Jira Cloud

The agent asks.
Your QA decides.

Quorum reads a Jira story, drafts the test cases and points out what the ticket leaves unclear. Where a business rule is missing, it asks you instead of making one up. Nothing reaches Jira until your QA signs off.

Start free trialSee how it works
  • 0edits applied without your approval
  • 0untraced cases. Each maps to a criterion or says why not.
  • 0copy-paste. Accepted cases arrive in Jira as sub-tasks.
Accept & send to Jira
Test cases5 cases · 2 questions
  • TC-1

    Apply a valid percentage code

    PositiveHighAC1
  • TC-2

    Apply an expired code at checkout

    NegativeHighAC2Depends on assumption
  • TC-3

    Apply a code typed in lowercase

    EdgeMediumAC1Depends on assumption
  • TC-4

    Apply a discount larger than the order total

    Proposed
    EdgeMediumAC1
  • TC-5

    Reject a malformed code

    Edit proposed
    NegativeLowAC2

    Expected result

    Shows a generic error

    Shows "Code not recognized" under the field

Open questions
BlockingTC-2

When a code has expired, should checkout show an error or remove the code silently?

Show an errorRemove silently
AssumptionTC-3

I assume codes are case-insensitive, so SAVE10 and save10 are the same code. Confirm?

Chat
TC-3

Why is this an edge case and not a positive one?

AC1 never says whether codes are case-sensitive, so lowercase input is a boundary. I assumed it works and opened a question for you to confirm.

TC-2TC-5
The maximum length check is missing

Most AI testing tools try to take the human out of the loop.

Quorum keeps your QA in it.

A test case built on a made-up business rule is worse than no test case.

So when the ticket is silent, the agent asks.

[01] Questions

It asks
before it guesses.

A missing rule is a question for your team, not a blank for the model to fill. Quorum finds the gaps in a story and asks about them, with concrete options to pick from.

BlockingAffects TC-2, TC-6

When a code has expired, should checkout show an error or remove the code silently?

Show an errorRemove silently

The ticket says “invalid codes show an error” but never mentions expired ones.

AssumptionAffects TC-3

I assume codes are case-insensitive, so SAVE10 and save10 are the same code. Confirm?

Usual for promo codes, but AC1 does not say so.

AnsweredUpdated TC-4

If the discount is larger than the order total, what is the final amount?

The total is capped at zero. Shipping is still charged.

Recorded on the story when you send, so devs and the PO see the decision too.

Blocking

The answer changes what a case should expect, and the ticket gives nothing to go on. Answer it or dismiss it before anything goes to Jira.

Assumption

A safe convention, said out loud: “I assume X, confirm it.” Every case that relies on it is flagged, so nothing rests on a silent guess.

Only questions worth asking

Before it asks, the agent looks for the answer in the ticket, its epic and the linked issues. Related doubts arrive as one question, not five.

Answers flow into the cases

Pick an option or write your own. The agent proposes the matching edits to every case your answer touches.

[02] How it works

Story in.
Sub-tasks out.

Five steps on one screen: the story on the left, the cases in the middle, questions and chat on the right. You make the call at every step.

  1. 01

    Connect Jira

    Point Quorum at your Jira Cloud site with an API token, kept encrypted on the server. Choose how test cases link to the story.

  2. 02

    Pick a story

    Find it by key or text with ⌘K. Before anything runs, you see what the agent will read: the epic, linked tickets and subtasks.

  3. 03

    Generate

    Cases and open questions stream in as they are written. Each case maps to an acceptance criterion, and any case that rests on an open question is flagged.

  4. 04

    Refine in chat

    Select cases, or a single step, and say what is off. The agent can push back with reasons, and every change it suggests waits for your approval.

  5. 05

    Send to Jira

    Each accepted case becomes a sub-task of the story, linked and labeled. Your answers go on the story as a comment, so the decisions are written down.

[03] Review

Reviewed like
a pull request.

Traced to the ticket, diffed field by field, approved by a person. The bar your team already sets for code.

See the gaps before you send

Every case points to the acceptance criterion it covers, and criteria with no cases stand out. No criteria on the ticket? The agent drafts them from the description, marks them as inferred and asks you to confirm.

Proposals, not overwrites

The agent never touches a case directly. Edits arrive as a field-by-field diff, new cases as drafts, deletions with a reason. Accept or reject each one, or all at once.

Your selection is the context

Select a case, a few, or a single step, and the agent answers about exactly that. Select nothing and you are talking about the whole story.

Sub-tasks, not spreadsheets

Accepted cases become sub-tasks of the story, linked the way you chose and labeled so you can find them with JQL, including the ones ready for automation.

[04] Compared

Agent speed. QA judgment.

Any model can write test cases. Getting the business rules right is the hard part, and it takes a person.

  • The ticket leaves a rule open

    Quorum
    Asks you, with concrete options
    One-shot AI generators
    Usually fills the gap with an assumption
    Manual test design
    Depends on who writes the cases

  • Link to acceptance criteria

    Quorum
    Every case, plus coverage per criterion
    One-shot AI generators
    Varies by tool
    Manual test design
    If someone keeps the matrix

  • Changing a case

    Quorum
    Proposed as a diff you approve
    One-shot AI generators
    Edit the output or regenerate
    Manual test design
    Edit by hand

  • Who signs off

    Quorum
    The QA, case by case
    One-shot AI generators
    Up to your process
    Manual test design
    The QA

  • Where the cases end up

    Quorum
    Sub-tasks on the Jira story
    One-shot AI generators
    Exported or copied into your tool
    Manual test design
    Wherever the team keeps them

[05] FAQ

The usual questions.

Do we need to change how we write tickets?
No. Quorum works from the stories you already have. If one has no acceptance criteria, the agent drafts them from the description and asks you to confirm.
Does it replace our QA?
No. The agent drafts and asks; your QA accepts, edits or rejects every change. When they disagree, the agent explains its reasoning, and the QA has the final say.
Does it work with our Jira?
Yes, if you are on Jira Cloud. It connects with an API token, with no Marketplace app to install. Jira Server and Data Center are not supported.
Is our data safe?
Your Jira token is encrypted and never leaves the server, and every account sees only its own data. Ticket content goes to the AI model only to generate and refine your cases.
Does it support Xray or Zephyr?
Not today. Cases are plain Jira sub-tasks with preconditions, steps and expected results, so you need no test management add-on to use them.
What about tickets written in Spanish?
Cases and questions come back in the language of the ticket. A story in Spanish gets its test cases in Spanish.

Test cases your QA actually agrees with.

Start free trialSee how it works