Inside Atlassian

The Atlassian AI SDLC Index

Insights from DX data on how the AI software development lifecycle (SDLC) is evolving. Latest update 11.18.26

Developer experience

What developers are using AI for

AI use by task

What developers are using AI for
CategoryWork typeShare
Code reviewCode reviews10.6
ExplorationArchitecture design and planning11.9
ExplorationQuestions7.2
ExplorationIdeation1.7
FeaturesFeature enhancements9.9
FeaturesNew features4.2
FixesBug fixes20.2
FixesBuild or test failure fixes3.9
FixesVulnerability remediation & hardening2
FixesData or migration fixes1.7
MaintenanceDocumentation10.5
MaintenanceRefactors and rewrites10.3
MaintenanceTest updates8.3
MaintenanceDependency or config updates6.4
MaintenancePerformance improvements2.2
OtherOther4.4
OtherData generation1.1
August–October 2026 (n = 180 companies). A single session can involve more than one type of work, so shares add up to more than 100%.

What’s changing for developers

Change in share of organizations rating aspects of their work favorably

What’s changing for developers
CategoryDriverPercent favorable (2025)Percent favorable (2026)ChangeDirection
Dev foundations and toolingDocumentation47.760.412.7Improved
Dev foundations and toolingDev environment6168.17.1Improved
Dev foundations and toolingBuild and test57.360.83.5Improved
Dev foundations and toolingCode maintainability77.279.62.4Improved
Feedback and operational safetyProduction debugging52.961.68.7Improved
Feedback and operational safetyChange confidence79.778.7-1Declined
Feedback and operational safetyLocal iteration speed77.374.6-2.7Declined
Feedback and operational safetyIncident handling87.384.4-2.9Declined
Direction and coordinationPlanning process64.565.30.8Improved
Direction and coordinationDecision-making65.866.60.8Improved
Direction and coordinationCross-team collaboration56.557.20.7Improved
Direction and coordinationClear direction74.172.1-2Declined
Focus and energyDeep work5963.44.4Improved
Focus and energyEngagement63.164.71.6Improved
Survey responses from August–October 2025 (n = 382 companies) and August–October 2026 (n = 567 companies). Values on the x-axis represent the average company’s share of developers rating each aspect favorably.

Most impactful drivers of developer experience

Elements of a developer’s day-to-day work that drive the highest impact across key outcomes (Speed, Quality, Ease of Delivery, and Ease of Release) and how developers rate them today

Most impactful facets of developer experience
DriverFavorabilityImpactOutcomeQuadrantCallout numberDescription
Production debugging63.60.57SpeedFix first1High impact, low favorability
Decision-making66.70.5SpeedMaintain2High impact, high favorability
Documentation61.50.47SpeedFix first3High impact, low favorability
Incident handling86.40.46SpeedMaintain4High impact, high favorability
Cross-team collaboration57.70.42SpeedFix first5High impact, low favorability
Dev environment68.80.4SpeedMaintain6High impact, high favorability
Change confidence79.20.39SpeedMaintain7High impact, high favorability
Local iteration speed75.40.34SpeedLower impact
Engagement65.10.34SpeedLower impact
Deep work62.80.32SpeedLower impact
Build and test61.60.3SpeedLower impact
Planning process64.50.27SpeedLower impact
Code maintainability800.21SpeedLower impact
Clear direction71.70.21SpeedLower impact
Incident handling86.40.73QualityMaintain1High impact, high favorability
Change confidence79.20.67QualityMaintain2High impact, high favorability
Local iteration speed75.40.6QualityMaintain3High impact, high favorability
Planning process64.50.59QualityFix first4High impact, low favorability
Documentation61.50.56QualityFix first5High impact, low favorability
Clear direction71.70.53QualityMaintain6High impact, high favorability
Code maintainability800.53QualityMaintain7High impact, high favorability
Production debugging63.60.53QualityLower impact
Build and test61.60.45QualityLower impact
Deep work62.80.43QualityLower impact
Dev environment68.70.43QualityLower impact
Engagement65.10.42QualityLower impact
Cross-team collaboration57.70.37QualityLower impact
Decision-making66.70.35QualityLower impact
Production debugging63.60.48Ease of deliveryFix first1High impact, low favorability
Documentation61.50.45Ease of deliveryFix first2High impact, low favorability
Incident handling86.40.44Ease of deliveryMaintain3High impact, high favorability
Change confidence79.20.39Ease of deliveryMaintain4High impact, high favorability
Local iteration speed75.40.38Ease of deliveryMaintain5High impact, high favorability
Cross-team collaboration57.70.38Ease of deliveryFix first6High impact, low favorability
Dev environment68.70.37Ease of deliveryMaintain7High impact, high favorability
Engagement65.10.35Ease of deliveryLower impact
Decision-making66.70.34Ease of deliveryLower impact
Planning process64.50.31Ease of deliveryLower impact
Build and test61.60.31Ease of deliveryLower impact
Code maintainability800.29Ease of deliveryLower impact
Clear direction71.70.26Ease of deliveryLower impact
Deep work62.80.2Ease of deliveryLower impact
Build and test61.60.64Ease of releaseFix first1High impact, low favorability
Incident handling86.40.55Ease of releaseMaintain2High impact, high favorability
Documentation61.50.5Ease of releaseFix first3High impact, low favorability
Production debugging63.60.5Ease of releaseFix first4High impact, low favorability
Change confidence79.20.48Ease of releaseMaintain5High impact, high favorability
Dev environment68.70.45Ease of releaseMaintain6High impact, high favorability
Local iteration speed75.40.42Ease of releaseMaintain7High impact, high favorability
Decision-making66.70.4Ease of releaseLower impact
Planning process64.50.39Ease of releaseLower impact
Cross-team collaboration57.70.36Ease of releaseLower impact
Clear direction71.70.34Ease of releaseLower impact
Deep work62.80.3Ease of releaseLower impact
Engagement65.10.29Ease of releaseLower impact
Code maintainability800.27Ease of releaseLower impact

Opportunities

High impact, low favorability

  1. Build and test
  2. Documentation
  3. Production debugging
What’s working

High impact, high favorability

  1. Incident handling
  2. Change confidence
  3. Dev environment
  4. Local iteration speed
Impact = how strongly a driver relates to the selected outcome; Favorability = the share of companies where most developers rate it favorably. Speed = speed at which a team delivers software; Quality = technical stability of a product; Ease of delivery = difficulty to do work as a technical contributor; Ease of release = amount of friction to release code to production.

August–October 2026 (n = 385 companies). Scores vary slightly from the previous chart due to varying sample sizes.”

Friction

What’s slowing developers down

Time lost to friction

20%

Developers lose one in five (20%) workdays to friction. That’s equivalent to about 10 work weeks per year.

  • Time lost to friction
  • Other work time
Developer time lost to friction
Segment% of work hoursTotal hours per yearTotal days per year
Time lost to friction20%409 Hrs51 Days
Other work time80%1631 Hrs204 Days
Developer time lost to friction

Hours lost per year, by source

Time lost to friction
Friction areaHours lost per year
CI wait time86.5
Review wait time86.5
Code comprehension71.5
Test wait time58
Information-seeking55
Dev environment toil47.5
August–October 2026 (n = 475 companies). Time losses are self-reported.

State of PR reviews

Minutes to first review, by share of code that is AI-authored

Median time to first review, by share of code that is AI-authored
Share of code that is AI-writtenMedian minutes to first review
20–40%31.2 minutes
40–60%66.6 minutes
60–80%72.6 minutes
80%+97.8 minutes
August–October 2026 (n = 441 companies). Share of AI-authored code is self-reported; median time to first review counts only human (non-bot) reviews and is measured from product data.

Review load per developer over time

Review load per developer over time
MonthPR reviews per person, per month
2026-0112.8
2026-0213.8
2026-0316.3
2026-0415.8
2026-0515
2026-0617.2
2026-0718
2026-0817.5
2026-0919.3
2026-1022.3
January–October 2026 (n = 627 companies). Counts distinct PRs reviewed per active human reviewer.

Impact

Time spent on innovation

Share of developer time spent building new features and capabilities

  • Time spent on innovation
  • Time spent on all other tasks
Time spent on innovation
SegmentShare
Time spent on innovation58.4%
Time spent on all other tasks41.6%
Time spent on innovation

Over time

Time spent on innovation over time
MonthTime spent on innovation
2025-0155.3%
2025-0254.9%
2025-0354.9%
2025-0456.8%
2025-0555.3%
2025-0656.5%
2025-0756.1%
2025-0857%
2025-0956.3%
2025-1055.4%
2025-1155.8%
2025-1257.6%
2026-0156.8%
2026-0256.5%
2026-0356.9%
2026-0456.8%
2026-0556.8%
2026-0656.9%
2026-0757.7%
2026-0858.5%
2026-0958.4%
2026-1057.8%
January 2025–October 2026 (n = 214 companies). Time spent on innovation is self-reported.

How much code AI authors vs. how much survives

Trending percent of AI-authored code and survivorship

How much code AI authors vs. how much survives
MonthAI-authored codeAI-authored code that survives
2026-0449%79.9%
2026-0554.7%79.7%
2026-0656.4%75.1%
2026-0759.4%66%
2026-0865.5%61.8%
2026-0963.5%58.9%
2026-1069.7%68.2%
April–October 2026 (n = 227 companies). Percent AI-authored is self-reported; code survivorship is measured in product as the share of AI-authored code that remains in commits.

AI spend and automation

AI spend per unit of output over time

AI spend per unit of output over time
MonthAI spend in USD
2026-04$8.85
2026-05$17.84
2026-06$23.83
2026-07$32.33
2026-08$38.77
2026-09$39.98
2026-10$29.11
April–October 2026 (n = 488 companies). Output is measured by weighted throughput — merged pull requests, weighted by size and complexity. Total dollar spend isn’t captured, so these values are a floor and likely understate true dollar amounts.

Share of PRs that get a bot review over time

Share of PRs that get a bot review over time
MonthPercent of PRs with bot review
2025-1115.3%
2025-1215.9%
2026-0116.4%
2026-0217.1%
2026-0318%
2026-0419.3%
2026-0519.7%
2026-0620.5%
2026-0721%
2026-0821.6%
2026-0921.6%
2026-1021.2%
November 2025–October 2026 (n = 630 companies). Value represents percent of total PRs with bot reviews only among companies using bot reviewers for >5% of PRs.

See how engineering teams are scaling agentic AI across real work

Methodology

DX is an engineering intelligence platform designed to help engineering organizations measure and improve software health, developer productivity, and AI-human collaboration for a truly AI-first SDLC.

The dataset represents a sample of more than 340,000 engineers at 500+ engineering organizations that use the DX platform. Companies range in size from fewer than 50 to over 10,000 employees, represent a wide geographic spread from Europe to Asia to North America, and encompass many industries, including finance, retail/e-commerce, and healthcare. All self-reported and telemetry data are aggregated at the company level, which means that team-level variation within organizations is not captured.