Team collaboration and planning

Atlassian Intelligence in Jira: How AI Is Transforming Sprint Planning in 2025

Atlassian's AI features are moving beyond simple automation to actively assist with estimation, work breakdown, and sprint planning. Here's what's changed and how teams are adapting.

Agile Team
FreeScrumPoker Team
December 11, 2025 ยท 10 min read

Atlassian Intelligence, the AI layer built into Jira and Confluence, has evolved significantly since its 2023 introduction. What started as natural language search and summarization has expanded into features that actively assist with core agile practices including work breakdown, estimation suggestions, and sprint capacity planning.

For scrum teams, these capabilities represent both opportunity and challenge. The opportunity: reduced time on mechanical planning tasks and potentially better-calibrated estimates. The challenge: maintaining team engagement and avoiding over-reliance on AI suggestions that may not account for context the system can't see.

What Atlassian Intelligence Actually Does Now

The current Atlassian Intelligence feature set for sprint planning includes:

Automatic Work Breakdown

When you create an epic or large user story, Atlassian Intelligence can suggest a breakdown into smaller stories or tasks. The AI analyzes:

  • The description and acceptance criteria you've provided
  • Similar work items from your project history
  • Common patterns from anonymized data across Atlassian's customer base
  • Technical dependencies mentioned in the description

The result is a proposed set of child issues, each with suggested descriptions. Teams can accept, modify, or reject these suggestions.

Estimation Suggestions

For teams using story points, Atlassian Intelligence can suggest estimates based on:

  • Historical data from similar stories in your project
  • Complexity signals in the story description
  • Comparison with team velocity patterns

These suggestions appear as recommendations during backlog grooming, not as automatic assignments. The team still votes and reaches consensus.

Natural Language Queries

Instead of building complex JQL queries, teams can ask questions in plain English:

  • "Show me all bugs related to authentication created this month"
  • "What stories are blocking the checkout epic?"
  • "List unestimated items assigned to the next sprint"

The AI translates these into JQL and returns results, making Jira more accessible to team members who don't know query syntax.

Sprint Health Insights

During sprints, Atlassian Intelligence provides observations about:

  • Items at risk of not completing based on historical patterns
  • Scope changes compared to sprint start
  • Unusual patterns in story movement or blocking

How Teams Are Using AI-Assisted Planning

Early adopter teams report varied approaches to integrating AI assistance:

The "First Draft" Approach

Some teams use AI suggestions as a starting point for discussion. The product owner creates an epic, requests AI breakdown, and presents the suggestions in refinement. The team then critiques, combines, splits, and revises the AI's proposal.

Teams using this approach report 30-40% reduction in refinement session time, primarily because they're editing rather than creating from scratch.

The "Sanity Check" Approach

Other teams complete their own breakdown and estimation, then compare against AI suggestions afterward. Significant discrepancies trigger discussion: why does the AI think this is bigger/smaller? What context is missing?

This approach preserves team ownership while using AI as a calibration tool.

The "Selective Use" Approach

Many teams use AI for specific story types where historical patterns are reliable (bug fixes, routine features) but skip AI assistance for novel work where historical comparisons are less relevant.

The Estimation Challenge

AI-suggested estimates create an interesting dynamic for planning poker and similar estimation techniques. Research on estimation shows that:

  • Seeing an estimate before voting anchors subsequent estimates toward that number
  • Group estimation value comes partly from diverse perspectives revealing different complexity dimensions
  • Discussion during estimation often surfaces implementation approaches and risks

If teams see AI estimates before voting, they may anchor on those numbers. If AI estimates are consistently accurate, teams may question the value of group estimation at all.

Preserving Estimation Value

Thoughtful teams are addressing this tension by:

  • Hiding AI estimates until after voting: Team estimates first, then compares to AI suggestion
  • Using AI for relative sizing only: AI groups stories by expected complexity; team assigns specific numbers
  • Focusing estimation discussion on risks: Even if point values converge, discussing what could go wrong remains valuable
  • Treating significant discrepancies as signals: A large gap between team and AI estimates triggers investigation

Work Breakdown Quality

AI-generated work breakdowns vary in quality based on the input provided:

What Works Well

  • Epics with detailed descriptions and clear acceptance criteria
  • Work similar to previous items in the project
  • Technical tasks with standard patterns (CRUD operations, API integrations)
  • Breaking large features into obvious functional components

What Works Poorly

  • Novel features without historical precedent
  • Work requiring deep domain knowledge not captured in descriptions
  • Items where technical approach significantly affects breakdown
  • Cross-team dependencies that require coordination

Teams report best results when using AI breakdown suggestions as prompts for discussion rather than accepting them directly.

The Human Element

Scrum ceremonies serve purposes beyond their mechanical outputs:

  • Shared understanding: Discussion during refinement ensures the team interprets requirements consistently
  • Commitment: Participating in estimation creates ownership of the resulting commitments
  • Knowledge transfer: Junior team members learn by hearing how seniors think through complexity
  • Risk identification: Diverse perspectives surface concerns that no single person would identify

AI can accelerate mechanical aspects of planning but can't replace these human dynamics. Teams that over-automate risk losing the collaboration benefits that make agile effective.

Integration with Estimation Tools

Planning poker tools like FreeScrumPoker can complement AI-assisted planning:

  • Independent estimation: Team members vote without seeing AI suggestions, preserving diverse input
  • Structured discussion: Large estimate spreads trigger discussion regardless of AI suggestions
  • Historical calibration: Compare team estimates against both AI suggestions and actual completion times
  • Remote participation: Distributed teams maintain engagement through interactive voting

The combination allows teams to benefit from AI assistance while maintaining the collaborative estimation process.

What's Coming Next

Atlassian's roadmap suggests continued expansion of AI capabilities:

  • Predictive sprint planning: AI suggestions for which items to include based on capacity and priority
  • Automated retrospective insights: AI-identified patterns across sprint history
  • Cross-project learning: Suggestions based on patterns from similar teams (with appropriate privacy controls)
  • Natural language story creation: Describe a feature conversationally; AI generates properly structured stories

Practical Recommendations

For teams evaluating Atlassian Intelligence for sprint planning:

  1. Start with work breakdown: Lower risk than estimation suggestions; provides immediate time savings
  2. Keep estimation human-first: Use AI estimates as calibration, not replacement for team voting
  3. Invest in quality inputs: AI suggestions improve dramatically with detailed descriptions and acceptance criteria
  4. Measure actual outcomes: Track whether AI-assisted estimates prove more or less accurate than team-only estimates
  5. Preserve discussion: Don't let AI efficiency eliminate the valuable conversations that happen during planning

Conclusion

Atlassian Intelligence represents a significant evolution in project management tooling. For sprint planning specifically, the AI features can reduce time spent on mechanical tasks and provide useful calibration for estimates.

However, the value of agile planning comes substantially from human collaboration: shared understanding, diverse perspectives, and team commitment. AI assistance works best when it accelerates without replacing these human elements.

The most successful teams will likely be those that thoughtfully integrate AI capabilities while preserving the collaborative practices that make agile effective. The tool has changed; the principles haven't.

FreeScrumPoker
FreeScrumPoker

Free online planning poker for agile teams. Keep estimation collaborative even as AI assists with other planning tasks.

More articles from FreeScrumPoker Blog →