Jira Field Pack - Pilot Evidence

Purpose

This note defines the minimum Jira configuration needed to make the evidence kit work in a real pilot.

The goal is tagging and traceability, not perfect instrumentation.

Minimum field pack

If Jira admins can create custom fields, use these.

1. AI Pilot Cohort

Type:

• checkbox
• or single select with Yes/No

Purpose:

• marks issues that belong to the pilot evidence set

2. AI Target Workflow

Type:

• single select

Purpose:

• tags the issue to one of the approved target workflows

Example values:

• Developer - Debugging support
• Developer - Unit test drafting
• QA - Defect triage
• Product - Acceptance criteria drafting

3. AI Workflow Mode

Type:

• single select

Values:

• learning
• delivery

Purpose:

• captures whether the work was primarily a learning-mode or delivery-mode use case

Optional field pack

Use these only if the admin burden is low and the team will actually maintain them.

4. AI Assistance Surface

Type:

• single select

Values:

• chat
• IDE assistant
• repo-aware assistant
• other

5. AI Review Follow-Up

Type:

• checkbox

Purpose:

• marks issues where reviewer burden or follow-up became notable enough to inspect

If custom fields are too hard

Use labels as the fallback, but stay consistent.

Suggested labels:

• ai-pilot
• ai-learning
• ai-delivery
• ai-review-followup
• ai-workflow-<slug>

This is weaker than real fields, but it is still better than leaving workflow tagging implicit.

Saved filters

Create at least these filters.

All pilot work

Use:

• AI Pilot Cohort = Yes

or label fallback:

• labels = ai-pilot

Pilot work by workflow

Use:

• AI Pilot Cohort = Yes AND "AI Target Workflow" = "<workflow name>"

Candidate rework or follow-up set

Use one of:

• AI Pilot Cohort = Yes AND "AI Review Follow-Up" = Yes
• AI Pilot Cohort = Yes AND labels = ai-review-followup
• linked follow-up fix issues using the team’s normal defect convention

Do not force one perfect query if the team already has a working rework convention.

Fields to export or inspect weekly

At minimum, pull:

• issue key
• summary
• issue type
• created date
• resolved date if available
• current status
• AI Target Workflow
• AI Workflow Mode
• labels
• link to issue

Use Jira reports or exports for cycle-time direction where possible rather than building a separate timing system from scratch.

Guardrails

• tag pilot issues before meaningful work begins
• if an issue is not tagged, it does not count as part of the pilot evidence set
• do not create many optional AI fields just because the admin console makes it possible
• keep the Jira layer focused on workflow traceability, not on rating people

Translation rule

If the target workplace uses Azure DevOps, Linear, GitHub Projects, or another tracker, keep the same logical field pack:

• pilot flag
• target workflow
• workflow mode
• optional surface or review-follow-up indicator