‹ Back to agilebykillette.com
A Practical Guide by Agile by Killette

The AI + Jira Playbook

How I turn Jira queries into delivery insight — and put AI to work doing the heavy lifting.
From ceremony to ownership.
Infographic: five ways a Scrum Master can implement AI to elevate team delivery and maturity — delivery insights from Jira data, smarter retrospectives, AI-assisted backlog and story readiness, earlier risk prediction, and an AI knowledge coach for the team, alongside a maturity progression from automation to insight to better decisions to team learning to higher maturity
AI doesn't replace the Scrum Master. It removes friction — so the time goes back into coaching the team.

Metrics without visibility are just noise. These are the kinds of Jira queries I use to keep an ART honest — paired with how AI turns raw results into the story leaders actually need. (JQL examples; adapt field names to your instance. A few cross-team queries assume standard link types or Jira Align.)

🔗

Dependencies & Blockers

Across the ART
status = Blocked AND updated <= -5d ORDER BY updated ASC
Blockers stuck 5+ days — the ones quietly rotting while everyone assumes they're handled.
Flagged is not EMPTY AND resolution = Unresolved
Every open impediment that's been flagged, across all teams, in one list.
issue in linkedIssues("PROJ-123", "is blocked by")
Everything blocking a specific critical feature — the dependency chain, made visible.
🤖 How AI assists

Ask AI in plain English — “show every blocker older than a week across Teams A–C” — and it writes the JQL for you. Then paste the export back and ask it to group by owning team and name the dependency causing the most downstream blockage. In seconds you get the story that PI Planning usually needs a room and a wall of sticky notes to uncover.

📈

Flow & Predictability

Is work actually moving?
sprint in openSprints() AND status not in (Done, Closed)
What's still open in the current sprint — your carryover risk, live, not at the retro.
status = "In Progress" AND updated <= -3d ORDER BY updated ASC
Aging work-in-progress — items that started but stalled and are silently blowing your cycle time.
status changed to Done during (-14d, now())
Throughput over the last two weeks — the raw material for a trend.
🤖 How AI assists

JQL pulls the items, but it can't compute cycle time or spot a trend on its own. Feed the results to AI (or a Copilot-connected dashboard) and ask it to calculate average cycle time, flag WIP above the limit, and name the two stories most likely to carry over. That turns a flat list into a forecast — the shift that took teams from 70% to 92% predictability.

🎯

Scope & Commitment

The Say/Do story
sprint in openSprints() AND type = Story AND "Story Points" is EMPTY
Unestimated stories that snuck into the sprint without a size — hidden risk.
sprint in openSprints() AND created >= -5d
Work added after sprint start — scope creep, made visible instead of absorbed silently.
🤖 How AI assists

Point AI at the sprint's planned-vs-current scope and it drafts your Say/Do narrative: what was committed, what changed mid-sprint, and the trade-off conversation that should have happened. That's the exact insight that turns a missed sprint into a productive leadership discussion.

⚠️

Risk & Readiness

What's about to bite
priority in (Highest, High) AND resolution = Unresolved AND updated <= -7d
High-priority items going stale — the fires forming while attention is elsewhere.
status = "In Progress" AND assignee is EMPTY
Work in flight that nobody actually owns.
type = Bug AND resolution = Unresolved AND created <= -14d ORDER BY priority DESC
Aging bugs, worst first — quality debt with a countdown.
🤖 How AI assists

Ask AI to triage the list into a Green / Yellow / Red view with a one-line “why” and a suggested mitigation for each item. That's your health report — pre-drafted, defensible, and ready before the leadership sync instead of scrambled the night before.

🚆

ART / PI Health

The whole train at a glance
project in (TEAMA, TEAMB, TEAMC) AND status = Blocked
Every blocker across the entire train in one view — no team's problem hides.
type = Feature AND "Program Increment" = "PI-2026.3" AND resolution = Unresolved
Feature progress for the current PI, across all teams.
🤖 How AI assists

Across a 15–17 person train, AI turns these cross-project queries into a single PI health summary — features at risk, the dependency web, and the one or two threads that could derail the increment — refreshed continuously instead of hand-assembled every Friday.

🔒 One rule I never break

In regulated environments, sensitive or client data never goes into public AI tools. I use enterprise Copilot with proper data governance and access controls — responsible AI is part of the delivery, not an afterthought. That's how you get the speed of AI without the audit risk.