CI/CD Statistics (2026): Pipelines, Delivery Performance, Testing and AI

CI/CD is now standard practice across software delivery, and ERP programs are starting to use it too. The numbers show something sign-off owners should watch: pipeline volume is growing much faster than pipeline reliability. This page gathers the most recent independent and vendor data on CI/CD adoption, delivery performance, pipeline health, testing and AI. Each number has its source linked next to it.

Key takeaways

CI/CD adoption and tool choice

Most developers now do DevOps work, but measured use of continuous integration is much lower than that. In large organizations, GitHub Actions, GitLab CI and Jenkins usually run side by side rather than one replacing the others.

• 83% of developers reported involvement in DevOps-related activities in Q1 2024. The CD Foundation State of CI/CD Report 2024 draws on SlashData's Developer Nation surveys, which reached more than 150,000 respondents across eight surveys over three and a half years.
• In the same research, only 29% of developers used continuous integration to automatically build and test code changes in Q1 2024, down from 33% in Q3 2023 (historical) (InfoWorld, April 2024). The gap between 83% and 29% comes from the question asked. "DevOps-related activities" includes monitoring and security testing, not just CI.
• In JetBrains' State of CI/CD 2025 survey of 805 developers and DevOps professionals, 32% of organizations used more than one CI/CD tool at the same time. GitHub Actions led for personal projects at 62% (daily.dev summary of JetBrains, October 2025). JetBrains says enterprises still rely heavily on Jenkins and GitLab (JetBrains, October 2025).
• Stack Overflow does not ask about CI/CD tools. It does track the container and infrastructure-as-code tools that pipelines rely on. Docker use rose from 35.0% in 2020 (historical) to 71.1% in 2025, Kubernetes reached 28.5% and Terraform 17.8% (Stack Overflow, September 2026).

Source: JetBrains State of CI/CD 2026 (598 respondents; multi-select share of organizations using each tool), as reported by Safeguard, October 2026

Note on conflicting tool shares: these 2026 figures come from a small CI/CD-specific sample and are reported second-hand. Surveys with broader developer populations give lower shares for each tool. Treat the ranking as more reliable than the exact percentages.

Pipeline volume: how much CI is running

GitHub's yearly Octoverse data is the largest public measure of CI activity. It is vendor data, and it covers one platform only.

Source: GitHub Actions minutes in billions, public projects plus self-hosted usage. GitHub Octoverse 2024 (2023 and 2024 values) and GitHub Octoverse 2025 (2025 value, same method), vendor data

• GitHub changed its method in 2025. Counting only free minutes in public projects, developers used 11.5 billion Actions minutes, up 35% from 8.5 billion. Using the old method, the 2025 figure is 13.5 billion, up 30% (GitHub, 2025). That is why the 2024 value appears as both 8.5 and 10.54 billion in different reports.
• In the 2026 Octoverse year (October 1, 2025 to September 30, 2026), GitHub Actions ran a job 1,440 times every second. That is 124 million jobs a day, more than three times the previous year (GitHub, October 2026). GitHub now counts Actions in jobs, not minutes, so these figures can't be compared directly with the minutes trend above.
• In the same period, developers pushed code 17.9 billion times, made 34.3 billion commits and merged more than 30 million Dependabot pull requests (GitHub, October 2026).

Source: Monthly activity in September 2026 compared with September 2025 (September 2025 = 1×). GitHub Octoverse (vendor data), October 2026

For reviewers, the important number is merged pull requests. Commits grew to 6× while merged pull requests grew to about 3.9×. More change is entering the pipeline than is coming out of review.

Delivery performance benchmarks (DORA)

DORA's 2025 research is based on survey responses from nearly 5,000 technology professionals plus more than 100 hours of qualitative data. In 2025, DORA replaced its elite/high/medium/low tiers with seven team archetypes, so the 2024 tiers below are the last published.

Source: Deployment frequency, share of DORA 2025 survey respondents (bar length relative to the largest group). DORA 2025 data as reported by re-cinq, 2026

Source: Change failure rate, share of DORA 2025 survey respondents (bar length relative to the largest group). DORA 2025 data as reported by re-cinq, 2026

The most common change failure band in 2025 was 8–16%. Only 8.5% of respondents reported failure rates of 2% or less.

Source: DORA 2024 performance clusters, share of respondents (bar length relative to the largest cluster). Accelerate State of DevOps 2024 as summarized by Libertify, 2026

• In the 2024 report, elite performers had 127× faster change lead time, 182× more deployments per year, an 8× lower change failure rate and 2,293× faster recovery from failed deployments than low performers (Libertify summary of DORA 2024).
• Elite performers in 2024 deployed on demand, had a lead time under one day, a change failure rate around 5% and recovered from failed deployments in under an hour (re-cinq, 2026).

Pipeline health: success rates and recovery time

CircleCI publishes medians from millions of real workflows. It is vendor data from one platform's customers, but it measures what actually happens in pipelines rather than what people report in surveys.

Source: CircleCI Q3 2026 Pulse, more than 22 million workflows (vendor data), October 2026; benchmark per Thoughtworks, 2026

• CircleCI's 2026 State of Software Delivery report put main-branch success at 70.8%, the lowest in five years. Median main-branch throughput fell 7% (Thoughtworks analysis of CircleCI data, 2026).
• In that report, the typical team took 72 minutes to get a failing pipeline back to green, a 13% increase in recovery time (CircleCI, 2026).
• Between Q1 and Q3 2026, median main-branch recovery time improved from about 59 to 54 minutes. Feature-branch recovery got slower, from about 80 to 86 minutes. Median main-branch workflow duration fell from 2 min 25 s in Q2 to 2 min 2 s in Q3 (CircleCI, October 2026).
• Since Q1 2026, the number of feature-branch workflows active on a typical day grew 40%, compared with 17% on main. Teams are starting more work than they finish (CircleCI, October 2026).

Toolchain sprawl and platform engineering

Several independent datasets reach the same conclusion: more tools mean slower releases. Platform engineering is the most common response.

• Delivery performance gets worse as the number of CI/CD tools grows beyond three. With four or five tools, more than 40% of developers reported slow releases (CD Foundation, August 2024).
• Harness's State of Developer Experience 2024 (vendor data) found that 97% of developers lose time to context switching because their tools come from multiple vendors. Two-thirds had code review delays of more than a week (CD Foundation, August 2024).
• GitLab's 2025 survey, run by The Harris Poll, covered 3,266 DevSecOps professionals. Respondents said they lose 7 hours per week to inefficient processes (GitLab, vendor data, November 2025).

Source: GitLab Global DevSecOps survey, 3,266 respondents (vendor data). Washington Technology, November 2025; GitLab, November 2025

Source: DORA, platform engineering capability (2025 research), 2026

Testing and quality engineering in the pipeline

The 2025–26 World Quality Report surveyed more than 2,000 senior executives across 22 countries and 10 sectors (OpenText, November 2025). It shows a large gap between trying AI in testing and running it at scale.

Source: Capgemini, Sogeti and OpenText, World Quality Report 2025, November 2025

• Conflicting figures: The WQR press release puts 52% of organizations in pilot phases (Capgemini, 2025). Capgemini's highlights page says 43% are experimenting with Gen AI in QA (Capgemini, 2025). They appear to measure different survey items. We report both.
• Non-adopters of Gen AI in QE rose to 11% (Capgemini, 2025).
• Organizations reported an average productivity gain of 19%, but one-third saw minimal gains. The top barriers to scaling were data privacy risks (67%), integration complexity (64%) and skill gaps (50%) (OpenText, November 2025).
• 60% of organizations struggle with secure, scalable test data, and 58% have trouble adopting AI-powered tools. Use of synthetic test data rose from 14% in 2024 to 25% in 2025 (Capgemini, 2025).
• Generative AI ranked as the top skill for quality engineers (63%), followed by core quality engineering skills (60%) (Capgemini, 2025).

AI in the delivery pipeline

Developers now use AI almost everywhere, but very few have put it inside CI/CD pipelines. DORA's two most recent reports also describe AI's effect on delivery stability differently.

Source: Google Cloud DORA, September 2025; JetBrains, October 2025

• Stability, then and now: In DORA 2024 (historical baseline), a 25% increase in AI adoption was associated with a 1.5% drop in delivery throughput and a 7.2% drop in delivery stability (OpsLevel summary of DORA 2024). In 2025, DORA found a positive link between AI and throughput, but AI still had a negative relationship with stability (Google Cloud, 2025).
• Respondents to JetBrains' CI/CD survey who don't use AI in pipelines cited cost, unclear value and security concerns. Those experimenting focus on debugging, code quality checks and pipeline optimization (daily.dev summary of JetBrains, 2025).
• Among 49,000+ respondents from 177 countries, 84% were using or planning to use AI tools, up from 76% the year before. 51% of professional developers use them daily. Favorable sentiment fell from more than 70% in 2023 and 2024 to 60% (Stack Overflow Developer Survey, 2025).
• In GitLab's 2025 survey (vendor data), 76% of respondents expect AI to lead to more engineers, not fewer (GitLab, 2025). 83% say AI will significantly change their roles within five years (Washington Technology, 2025).

What this means for your sign-off

• Ask for delivery evidence in DORA terms, not a single green build. In DORA's 2025 data, the most common change failure band was 8–16%. Only 8.5% of teams reported 2% or less. Before you sign, require the release's change failure history and recovery times next to the test results. Our Go/No-Go Readiness Checklist includes a place to record them.
• Treat a red main branch as a defect in your release process. Main-branch success was 70.8% in Q1 2026, against a 90% benchmark, and recovery took about an hour (CircleCI vendor data). For ERP releases, automate regression on business-critical processes first so failures show up early and are fixed quickly. See S/4HANA regression automation: what to automate first.
• Plan for test data as the bottleneck. 60% of organizations struggle with secure, scalable test data, while synthetic data use rose from 14% to 25% in a year (World Quality Report). A pipeline can only run as often as it gets usable data. Our guide to copies, masking and synthetic data for S/4HANA covers the options.
• Put AI-generated changes behind a review gate. 30% of DORA respondents have little or no trust in AI-generated code. Only 37% of GitLab's respondents would let AI handle daily tasks without human review. DORA still finds AI lowering stability. Write the review rules into your lifecycle before AI-generated code and tests reach your transports. See AI-DLC for ERP teams.
• Consolidate tools before adding more. Releases slow down noticeably at four or five CI/CD tools (CD Foundation), and 32% of organizations already run more than one (JetBrains). One pipeline record linked to one test management system gives an auditor a single trail to follow.

Frequently asked questions

What percentage of teams deploy on demand?

In DORA's 2025 survey of nearly 5,000 professionals, 16.2% of respondents deployed on demand. The largest group, 31.5%, deployed between weekly and monthly (DORA 2025 data via re-cinq).

What is a good CI pipeline success rate?

CircleCI recommends a 90% success rate on the main branch. Its customers' median was 70.8% in Q1 2026 and 78.3% in Q3 2026 (CircleCI, vendor data, 2026).

Which CI/CD tool is most widely used?

In JetBrains' 2026 CI/CD survey of 598 respondents, GitHub Actions was used in 47% of organizations, GitLab CI in 32% and Jenkins in 28% (as reported by Safeguard, 2026). Many organizations run more than one tool: 32% in JetBrains' 2025 survey.

How long does it take to fix a broken build?

The typical CircleCI team needed 72 minutes to get back to green in the 2026 State of Software Delivery report (CircleCI, vendor data). By Q3 2026, median main-branch recovery was about 54 minutes (CircleCI, 2026).

Are teams using AI inside their CI/CD pipelines?

Mostly not yet. 73% of JetBrains' CI/CD survey respondents use no AI in CI/CD workflows (JetBrains, 2025), even though 90% of DORA respondents use AI at work (Google Cloud, 2025).

How widespread is platform engineering?

DORA's 2025 research found that 90% of organizations use an internal developer platform and 76% have dedicated platform teams (DORA).

Sources

1. Announcing the 2025 DORA Report: State of AI-Assisted Software Development, Google Cloud, September 2025
2. State of AI-assisted Software Development 2025 (PDF), Google Cloud / DORA, September 2025
3. Platform engineering capability, DORA, accessed 2026
4. DORA Metrics: All 5, With 2025 Benchmarks and AI, re-cinq, October 2026
5. Deployment Frequency in 2026: Benchmarks, Risks, and AI Impact, Axify, 2026
6. TL;DR: Key Takeaways from the 2024 Google Cloud DORA Report, OpsLevel, 2024
7. State of DevOps 2024: DORA Report Findings, Libertify, 2026
8. State of CI/CD Report 2024, CD Foundation, April 2024
9. State of CI/CD Report: New Findings and Trends from CD Foundation, The Linux Foundation, April 16, 2024
10. Decoding the Real State of DevOps in 2024, CD Foundation, August 15, 2024
11. Most developers have adopted devops, survey says, InfoWorld, April 2024
12. Octoverse: AI leads Python to top language as the number of global developers surges, GitHub, October 2024
13. Octoverse: A new developer joins GitHub every second as AI leads TypeScript to #1, GitHub, October 2025
14. Octoverse: Commits, pushes reach new highs in 2026 as developers build with agents, GitHub, October 2026
15. The State of CI/CD in 2025: Key Insights from the Latest JetBrains Survey, JetBrains, October 2025
16. The State of CI/CD in 2025 (summary), daily.dev, October 2025
17. CI Platform Usage Share 2026: Survey Data and Pipeline Security, Safeguard, October 2026
18. 2026 State of Software Delivery Report (PDF), CircleCI, 2026
19. 5 takeaways from the State of Software Delivery Q3 Pulse report, CircleCI, October 2026
20. A Thoughtworks perspective on CircleCI's 2026 State of Software Delivery Report, Thoughtworks, 2026
21. GitLab Survey Reveals the "AI Paradox," Faster Coding Creates New Bottlenecks, GitLab, November 10, 2025
22. AI speeds up coding and creates new bottlenecks, GitLab survey finds, Washington Technology, November 2025
23. World Quality Report 2025: AI adoption surges in Quality Engineering (press release PDF), Capgemini, November 13, 2025
24. World Quality Report 2025-26, Capgemini, 2025
25. World Quality Report 2025 press release, OpenText, November 13, 2025
26. 2025 Stack Overflow Developer Survey, Stack Overflow, July 2025
27. Getting ready for 2026 results: A look back on Developer Survey findings, Stack Overflow, September 30, 2026

Last updated: November 2, 2026. Statistics are refreshed as new research is published.