Skip to content

AI in the software development lifecycle: 7 trends backed by data

Mykhailo Poliarush6 min read

AI is now used at every stage of the software development lifecycle, but unevenly: almost every developer uses it to write code, far fewer let it plan, review or deploy, and trust has not kept up with adoption. In the DORA 2025 report, 90% of technology professionals use AI at work, while 30% report little or no trust in AI-generated code. The seven trends below are drawn only from published surveys and studies, each linked to its primary source.

Is AI now part of every SDLC stage, or only of writing code?

Coding still leads, but the rest of the lifecycle is close behind. Among people who perform each task, DORA found that 71% use AI to write new code, 62% to create test cases, 56% for code review and 56% to create specifications. Security analysis (51%) and requirements analysis (49%) are the least AI-assisted stages.

Share of developers who use AI for each task they perform

Writing new code
71%
Modifying existing code
66%
Writing documentation
64%
Creating test cases
62%
Debugging
59%
Creating specifications
56%
Code review
56%
Maintaining legacy code
55%
Planning
54%
Security analysis
51%
Analyzing requirements
49%

Source: DORA, State of AI-assisted Software Development 2025 (Figure 21, about 5,000 respondents)

Intent tells the same story from the other side. In the Stack Overflow Developer Survey 2025, 75.8% of respondents said they do not plan to use AI for deployment and monitoring, and 69.2% said the same about project planning. Writing code (28.9%) and searching for answers (19.6%) meet the least resistance. The 2026 Stack Overflow results show that only 20% use AI to deploy, operate or troubleshoot production systems.

Developers who do not plan to use AI for each area

Deployment and monitoring
75.8%
Project planning
69.2%
Committing and reviewing code
58.7%
Testing code
44.1%
Documentation
39.6%
Debugging or fixing code
36.4%
Writing code
28.9%
Searching for answers
19.6%

Source: Stack Overflow Developer Survey 2025, AI section (25,349 responses)

What it means: the stages closest to production are where teams keep humans in charge. That is where engineering judgment and verification matter most.

Why do developers use AI more but trust it less?

Adoption keeps rising: 84% of Stack Overflow 2025 respondents use or plan to use AI tools, up from 76% the year before, and 47.1% use them daily. Yet 46% distrust the accuracy of AI output, against 33% who trust it, and only 3% "highly trust" it (33,244 responses). The most common frustration, named by 66%, is answers that are "almost right, but not quite", and 45.2% say debugging AI-generated code takes longer.

How much developers trust AI-generated output

A great deal
4%
A lot
20%
Somewhat
46%
A little
23%
Not at all
7%

Source: DORA, State of AI-assisted Software Development 2025

Trust is conditional. In the 2026 survey, 48% trust AI when they can easily verify the answer, but only 16% trust it for most tasks that are not critical decisions. Verification, not the model, is what earns trust.

Does AI make developers faster, or does it only feel that way?

Self-reported gains are large. In DORA 2025, more than 80% believe AI has increased their productivity. In the JetBrains State of Developer Ecosystem 2025 (24,534 developers), nearly nine in ten save at least an hour a week, and one in five saves eight hours or more.

Measured results are less clear. In a randomized controlled trial by METR, 16 experienced open-source developers working on 246 real tasks took 19% longer with AI tools. Before the study they expected a 24% speed-up, and afterwards they still believed AI had made them 20% faster.

Expected, believed and measured effect of AI on task time (METR, 2025)

Expected, believed and measured effect of AI on task time (METR, 2025)
CriterionResult
Developers' forecast before the study24% faster
Developers' estimate after the study20% faster
Measured in the randomized trial19% slower

Source: METR, randomized controlled trial, 16 developers, 246 tasks (2025-07-10)

METR's February 2026 update points to a smaller slowdown with newer tools, but METR itself calls the new data an unreliable signal: 30% to 50% of developers avoided submitting tasks they did not want to do without AI. The honest summary is that perceived and measured productivity diverge, so teams should measure delivery outcomes rather than rely on impressions.

Are coding agents ready to work without supervision?

Agents are spreading fast, mostly on a leash. GitHub reports in Octoverse 2025 that its Copilot coding agent created more than 1 million pull requests between May and September 2025. In the 2026 Stack Overflow survey, the most used coding agents were Claude Code (66%) and GitHub Copilot (59%).

Autonomy is still limited. A Stack Overflow pulse survey in May 2026 (about 1,100 respondents) found that 63% rarely or never let agents run on autopilot and 60% block agents from making unapproved system changes. In DORA 2025, surveyed in mid-2025, 61% never used AI in agent mode at all. Usage patterns already show the shift: in Anthropic's analysis of 500,000 coding interactions, 79% of Claude Code conversations were automation, where the AI performs the task, compared with 49% on Claude.ai.

Does AI-generated code make software less secure?

It can, unless it is reviewed. In Veracode's 2025 study of more than 100 models on 80 coding tasks, AI-generated code introduced security vulnerabilities in 45% of cases, Java failed more than 70% of the time, and newer models were not more secure. Veracode sells security testing, so read the numbers with that in mind; the direction matches what security teams report.

AI also helps fix security issues: Octoverse 2025 reports that developers accepted AI-generated fixes for Broken Access Control flaws in more than 6,000 repositories a month. The pattern is the same as in testing. AI output is useful when it passes the same checks as human code.

Why does AI speed up delivery and make it less stable at the same time?

DORA 2025 finds that AI adoption now improves software delivery throughput but still increases delivery instability. AI amplifies whatever system it lands in: teams with strong testing, small batches and fast feedback get faster, while weak pipelines ship more changes that fail. In the same report, 59% think AI improved code quality and 10% think it made it worse, which is a perception, not a measurement.

The practical consequence is that automated tests, code review and CI become more important as AI writes more of the code, not less.

What will AI-assisted engineering cost by 2028?

Two forecasts frame the next two years. Gartner predicted in July 2025 that 90% of enterprise software engineers will use AI code assistants by 2028, up from less than 14% in early 2024. In June 2026, Gartner predicted that AI coding costs will exceed the average developer's salary by 2028 as token consumption grows. Both are forecasts, not measurements, but together they say that AI spend will need the same cost control as headcount.

Benchmarks are a weak guide for that decision. In February 2026, OpenAI stopped reporting SWE-bench Verified because of contamination, after an audit found material issues in 59.4% of 138 problems it checked. Teams should evaluate tools on their own codebase and test suite.

How should engineering teams act on these trends?

The evidence points in one direction: adoption is close to universal, trust depends on verification, agents stay supervised, measured gains are smaller than perceived ones, and AI-generated code needs security review. SDCLabs works the same way. AI agents draft specs, code and tests; senior engineers review them and decide what ships; automated test suites, built on the Codeception and CodeceptJS frameworks our founders created, verify every change. See our AI engineering services and QA and test automation services.

Sources were checked on 2026-11-17. Survey figures describe their respondents, not all developers.

Frequently asked questions

How many developers use AI in 2026?

84% of respondents to the Stack Overflow Developer Survey 2025 used or planned to use AI tools, and DORA 2025 found that 90% of technology professionals use AI at work. In the 2026 Stack Overflow survey, 73% of those who use AI coding assistants or agents use them daily.

Which SDLC stages use AI the least?

Deployment, monitoring and planning. In Stack Overflow 2025, 75.8% did not plan to use AI for deployment and monitoring and 69.2% for project planning; in DORA 2025, requirements analysis (49%) and security analysis (51%) had the lowest AI use.

Does AI really make developers more productive?

Developers report gains, but a randomized trial by METR found experienced developers were 19% slower with AI on their own repositories, while believing they were 20% faster. Measure delivery outcomes on your own team before assuming a speed-up.

Related posts

Get an estimate for your product.