TeamCity

The State of CI/CD in 2026

This report brings together data from four JetBrains studies to explore what CI/CD looks like today, offering insights about which tools teams use, where AI is already making an impact, and what is likely to change next.

The findings come from the JetBrains CI/CD Tools Adoption Survey 2026 (completed by 598 respondents), with additional context from the JetBrains Developer Ecosystem Survey 2026 (21,946 respondents) and two AI Pulse studies (13,401 respondents total).

Share:

The landscape in numbers

83%

of respondents run CI/CD pipelines, making it the most widely adopted DevOps practice.

28%

reach their CI/CD tool through MCP or agent skills.

73%

self-host their primary CI/CD tool, on-premises or on self-managed cloud instances.

50%

of organizations use AI somewhere in their CI/CD process.

TL;DR: Six main takeaways

1 – CI/CD is nearly universal.

83.4% of developers run pipelines, ahead of containerization (50%) and infrastructure as code (39%). Developers can no longer be distinguished by whether they have adopted CI/CD practices. The differentiation has moved somewhere else.

Read more

2 – Consolidation is the plan.

The transitional two-tool stack collapsed from 33% of organizations to 25%, while the share of three-or-more grew from 12% to 17%. “Toolchain consolidation” is now a top-three reason teams switch tools.

Read more

3 – Self-hosted approaches are on the rise.

On-premises CI/CD hit a five-year high at 31%, and fully managed cloud CI has lost seven points since its 2022 peak. Nearly three-quarters of developers now self-host their primary CI/CD tool in some form.

Read more

4 – AI plays a role around the pipeline, almost never inside it.

Roughly half of organizations use AI for some CI/CD tasks. But only 8% run an AI step inside the build itself, a figure two independent surveys landed on within 0.1 points of each other.

Read more

5 – Organizations bring their own vendors.

28% of organizations already reach their CI/CD tool through MCP or agent skills. The usage of in-product AI assistants sits at 10%.

Read more

6 – Agents moved the bottleneck downstream.

Among teams using AI in their CI/CD workflows, the top issues are review capacity (37%) and judging whether AI changes are reliable (36%).

Read more

CI/CD overview

CI/CD is still at the center of software delivery

83

82

CI/CD pipelines

50

56

Containerization and orchestration

39

43

Infrastructure as code (IaC)

23

29

GitOps

17

17

DevSecOps

16

17

Platform engineering

14

7

AI-driven operations

13

16

Serverless (FaaS) solutions

7

5

We don't use any DevOps practices

1

0

Other

Developer Ecosystem Survey 2026. Multiselect – the sum of the shares exceeds 100%.

83.4% of respondents who use DevOps practices run CI/CD pipelines

CI/CD is by far the most widely adopted DevOps practice, ahead of containerization and orchestration, infrastructure as code, GitOps, and platform engineering.

The interesting question in 2026 is how CI/CD needs to evolve as the rest of software development changes around it.

Only 7% of respondents report using no DevOps practices at all.

This is the context for everything that follows. CI/CD is no longer a capability teams are deciding whether to adopt; it's infrastructure they already have and are now deciding what to do with. The rest of this report is about that second question.

The most popular CI/CD tools in organizations in 2026

47

41

GitHub Actions

32

34

GitLab CI

28

31

Jenkins

15

14

Azure DevOps Server

10

7

TeamCity

7

0

AWS CodePipeline / AWS CodeBuild

6

8

Other

6

7

Bitbucket Pipelines

5

5

Custom tool

4

4

Google Cloud Build

Data was not provided for Buildkite, Travis CI, Bitrise, CloudBees CI, Harness CI/CD, or Semaphore CI in 2025; these are displayed as 0%. The AWS option changed from “AWS CodePipeline / AWS CodeStar” in 2025 to “AWS CodePipeline / AWS CodeBuild” in 2026; the missing counterpart values are also displayed as 0%.

State of CI/CD Tools Survey 2025/2026. Multi-select – the sum of the shares exceeds 100%.

GitHub is popular both for personal use and in organizations

GitHub Actions leads organizational adoption, with 47% percent of respondents saying their organization uses it, followed by GitLab CI at 32% and Jenkins at 28%.

TeamCity is used in 9% of respondents' organizations. Azure DevOps Server stands at 12%, while AWS CodePipeline, custom tools, and Bitbucket Pipelines comprise the next tier.

Multi-tool reality

The teams that run several CI/CD tools do so for structural reasons, different clients, different tech stacks, or inherited legacy systems.

25%

of organizations use two or more CI/CD tools simultaneously.

17%

of organizations use three or more CI/CD tools.

Anonymous respondent
State of CI/CD Tools 2026 survey responses

“We have different departments working with different customers, each with specific requirements regarding tool usage. Our choice of tools is based on those requirements”

Self-hosting is on the rise

Cloud services dominate the conversation in much of the software industry, but in the CI/CD space, the story is more nuanced.

31

Installed on my own premises

42

Installed on managed cloud instances

24

Fully managed cloud service

3

Other

Developer Ecosystem Survey 2026. Single select – the shares sum to 100%.

For some teams, managed infrastructure is the simplest choice. Others continue to need control over hardware, networks, data, build environments, security requirements, or costs.

The future of CI/CD therefore looks less like an inevitable move from self-hosted to SaaS and more like a mix of deployment models chosen depending on the work at hand.

Self-hosted is here to stay

72.8% of respondents in the Developer Ecosystem Survey run their primary CI/CD system either on their own premises or on self-managed cloud infrastructure.

On-premises installations alone reached 30.5% in 2026, their highest level in the five years covered by this question.

Fully managed CI/CD services account for 24.4%.

AI has arrived in CI/CD, but mostly around the pipeline

48.8%

of organizations use AI for at least one CI/CD-related task.

The use of AI in software delivery is no longer theoretical. But AI is not yet taking over the pipeline itself.

Which areas of the software development cycle does your company use AI agents in, if any?

67

Implementing features or programs

62

Writing tests

60

Code review

53

Refactoring

52

System design / architecture

46

Documentation creation

44

Debugging

43

Requirement analysis / gathering

25

QA / Testing

25

Software maintenance

22

CI / CD

20

Migrating code from one language / stack to another

20

Project management

15

Deployment / Monitoring

11

Security / DevSecOps

10

Incident management

5

I don’t know

4

None

1

Other

State of CI/CD Tools Survey 2026. Multi-select – the sum of the shares exceeds 100%.

22% of respondents say their company uses AI agents in CI/CD.

For now, AI is much more likely to help developers understand, review, or fix what happens around a pipeline than to become another step inside it.

How is AI currently used in your CI/CD pipelines?

43

My company or organization doesn't use AI in its CI/CD

18

Creating or maintaining pipeline configurations

23

Writing or fixing tests that run in the pipeline

34

Reviewing code changes before they're merged

15

AI-powered quality or security checks that can block a build

8

Running AI-powered steps as part of a build or test

24

Diagnosing build or test failures

13

Assisting with deployments, releases, or post-deploy incident response

2

Other

8

I'm not sure

State of CI/CD Tools Survey 2026. Multi-select – the sum of the shares exceeds 100%.

Only 8% use AI inside builds

The most common applications are practical, targeted tasks such as code review, diagnosing failures, and writing or fixing tests. Only 8% report running AI steps directly inside builds.

Anonymous respondent
State of CI/CD Tools 2026 survey responses

“We use AI-powered steps to analyze code quality, identify potential security issues, summarize test failures, and assist with troubleshooting failed builds.

AI also helps review pull requests by suggesting improvements and highlighting potential risks, which speeds up debugging and code reviews while reducing manual effort.”

Developers are bringing their own agents to CI/CD

64

Web UI

37

CLI provided by the CI/CD tool

10

In-product AI assistant provided by the CI/CD tool

18

MCP server provided by the CI/CD tool

18

Agent skills provided by the CI/CD tool (e.g. for Claude or Codex)

20

REST API or custom scripts

7

Other

State of CI/CD Tools Survey 2026, multi-select. The final row is a derived combination of the MCP and agent-skills rows, and was not a separate answer option.

More than a quarter of organizations already interact with their CI/CD tools through agentic interfaces, while only 10% use an AI assistant built directly into their CI/CD tool.

The implication is interesting: Developers are not necessarily waiting for CI/CD vendors to create a separate AI experience. They are increasingly connecting the agents they already use to the tools that are part of their development workflows.

64% of respondents still use a web UI to interact with their CI/CD tool

The traditional web UI is not disappearing, but it is no longer the only front door to CI/CD.

Teams are warming up to AI

Last year, 60% of teams said they did not see a clear use case or value for AI in CI/CD. This year, that figure has fallen to 27%. The share saying that no suitable AI functionality exists also dropped, from 14% to 8%.

The main barriers are now more practical. Among organizations not using AI in their CI/CD workflows, 47% say it is not a priority, 35% cite data privacy and IP concerns, 33% say their current setup works well enough without AI, and 32% point to cost.

Company size makes a difference. While 52% of small companies say their current CI/CD approaches work well enough without AI, only 21% of large companies say the same. For smaller companies, the main reason for holding back may simply be that they are satisfied with their current setup.

47% of respondents

say implementing AI in their CI/CD workflows is not a priority for their organization.

What prevents your company or organization from integrating AI into its CI/CD workflows?

47

0

AI in CI/CD not being a priority for us

36

33

Concerns about data privacy, security, or IP protection

33

0

Concerns about the cost of AI usage

33

0

Our current CI/CD working well enough without AI

29

36

Lack of trust in AI-generated results

27

60

Unclear use cases or uncertain value

16

18

Company or team policies restricting AI usage

15

0

Lack of in-house expertise or skills to implement AI in CI/CD

14

0

Concerns about vendor lock-in or dependency on specific AI providers

14

0

Regulatory or compliance restrictions on AI usage

Note: 0 indicates that the option was not included in that year’s survey.

State of CI/CD Tools Survey 2025/2026. Base: organizations reporting no AI use in CI/CD. Multi-select – the sum of the shares exceeds 100%.

AI agents have moved the bottleneck downstream

Among organizations using AI in CI/CD, the main challenges now relate to reviewing and verifying AI-generated changes.

37% percent say the volume of AI-generated changes slows down reviews, while 36% find it difficult to assess whether those changes are reliable. Other concerns include unpredictable AI costs (28%), difficulty tracking AI-made changes (22%), and difficulty debugging issues introduced by AI (20%).

Only 19% say AI-generated code failing the build is a problem. This suggests that generating code is no longer the main challenge. The bigger issue is verifying a growing number of changes quickly and reliably.

AI in DevOps: Why Adoption Lags in CI/CD (and What Comes Next)

This article references data from several JetBrains studies in order to analyze how AI is currently used in the areas of DevOps and CI/CD.

Which issues has your company or organization faced when using AI in its CI/CD workflows, if any?

36

The volume of AI-generated changes causes review bottlenecks

35

It’s difficult to assess the reliability of AI-generated changes

29

Cost of AI usage in our pipelines is high or unpredictable

21

It’s difficult to track which changes were made by AI

20

AI-generated code fails during building or testing

20

It’s difficult to debug or trace issues introduced by AI

17

CI/CD run time or cost is increased by AI

16

AI-generated code introduces security vulnerabilities or compliance issues

15

AI-generated changes break CI/CD pipelines or infrastructure

13

AI causes issues during or after application deployment

State of CI/CD Tools Survey 2025/2026. Base: organizations reporting no AI use in CI/CD. Multi-select – the sum of the shares exceeds 100%.

36% of respondents say AI causes review bottlenecks

Meanwhile, 35% say it's hard to assess whether AI-generated changes are reliable, followed by unpredictable AI costs (28%), difficulty tracking which changes AI made (22%), and difficulty debugging AI-introduced issues (20%).

AI-generated code failing the build – the problem we’d have predicted two years ago – ranks below every one of those, at 20%.

Anonymous respondent
State of CI/CD Tools report 2026

“AI in CI/CD has been trending for the past year and a half. I think a comprehensive, end-to-end CI/CD solution built around AI tools and agents could be very promising.

If such a solution emerges, it could significantly impact the CI/CD market.”

AI's role in deployment

38

47

Writing deployment configs & scripts

35

43

Planning the deployment approach

39

44

Creating CI/CD pipelines

AI Pulse Survey, wave 1 (Sep 2025) and Q1 2026 (Jan 2026). Base: respondents using AI for the activity, depending on activity and wave.

Writing deployment configuration shifted the most, increasing by nine points. Creating the pipeline itself moved least, at five.

Developers are handing over the scripts before they hand over the pipeline – the more consequential the artifact, the slower the handover.

13.4% say they want no AI involved in their CI/CD pipelines at all, and that figure did not budge between waves.

Three key takeaways

CI/CD is becoming the verification layer

AI makes writing code faster, but reviewing and trusting it harder. CI/CD already gates changes, making it the natural place for AI checks. This is a hypothesis, as the survey measured the problem, not where developers want it solved.

Agents are gaining ground

28% of organizations use MCP or agent skills to access CI/CD, compared with 10% using built-in assistants. Across the entire software development landscape, AI adoption runs 13–23 points ahead of AI use within the CI/CD product.

Maintenance effort is driving tool switching

62% of developers have no plans to switch CI/CD tools. Those that do cite scaling costs (24%), maintenance effort (23%), tool consolidation (22%), and complex configuration (21%). Teams switch when their current setup becomes too expensive or difficult to run.

How we got these numbers

This report uses four separate datasets. We haven’t combined them because the samples and questions differ.

The 2026 State of CI/CD Tools survey covered 598 people, including 576 who use at least one CI/CD tool. It provides the data on tools, AI use, and barriers.

The Developer Ecosystem Survey ran from May to July 2026 and received 22,082 responses. We use it for insights about DevOps practices, hosting, and pipeline observability.

Two AI Pulse surveys allowed us to track the shift toward agentic workflows. The wave in September 2025 had 4,830 responses, and the one in January 2026 had 8,751.

Respondents were recruited through developer blogs, newsletters, social media, and community lists. The results are weighted, but they reflect an audience of engaged developers rather than the whole industry.

Sample sizes vary by question, from roughly 1,400 to more than 20,000. Each chart lists its own base. Totals may exceed 100% when respondents could choose several answers.

The question about AI adoption in the AI Pulse surveys also changed between 2025 and 2026, so the responses are not directly comparable.

The two surveys conducted in 2026 sometimes yielded different results. For example, the share of developers using Jenkins was 28% in one and 37% in the other, and TeamCity’s share shifted from 9% to 4%. While one survey looks at organizations, the other at individuals, and each recruited respondents differently.

The State of CI/CD Tools in 2025

Compare the findings with those from last year.

Get the next one

We run this study every year. Subscribe to our newsletter, and we'll send the 2027 edition when it lands, plus the underlying data tables for this one.

Subscribe to the newsletter