Free tools Windows power users keep installed
One-click scans. No signup required.
iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Probably—but not in a uniform or guaranteed way. Current studies show AI assistants helping with parts of implementation, testing, documentation, and operations. They also point to continuing human contributions: defining the work, providing project context, checking generated output, and protecting reliability, security, and human relationships. That is evidence of a changing task mix, not proof that every developer’s job will move upward or that employers will reward those contributions more.
What does “moving up the stack” mean for a developer?
It means spending less of the workday producing routine code or other artifacts and more time deciding what should be built, fitting changes into a particular system, evaluating whether they are correct, and managing the consequences of shipping them. It does not mean that implementation disappears: even when an assistant drafts code, someone still has to determine whether the change solves the right problem and works safely in its actual environment.
The research supports a shift in which some tasks can be assisted or delegated. It does not establish that all developers will move into more strategic work, that the same tasks will be automated in every team, or that this change will improve pay, hiring, or job security.
Which software tasks are AI assistants helping with?
Implementation and testing
In a 2025 analysis of three randomized field experiments at Microsoft, Accenture, and an anonymous Fortune 100 company, researchers reported an estimated 26.08% increase in completed tasks among developers using an AI coding assistant; the estimate had a standard error of 10.3%. The authors also reported higher adoption and greater productivity gains among less experienced developers. This is a combined experimental result, not a promise of the same gain for an individual, another tool, or every kind of task. Microsoft Research’s June 2025 report describes the experiments.
#1 Best Overall
A 2025 JetBrains Research publication, first made public in June 2024, surveyed 481 programmers about tasks including feature implementation, test writing, bug triage, refactoring, and producing natural-language artifacts. Respondents showed interest in delegating some less-enjoyable work, including writing tests and natural-language artifacts. The survey records views and preferences; it is not a controlled measurement of how much time those tasks save. JetBrains Research’s publication page describes the study.
Documentation and operations
A Microsoft Research mixed-methods study of 860 developers, published in October 2025, found strong current use of AI and interest in improving support for coding and testing, as well as substantial demand to reduce toil in documentation and operations. These are reported patterns of use and desired support, not evidence that assistants can independently own operational decisions or documentation quality. The Microsoft Research study also identifies safeguards developers consider important for different kinds of work.
What still calls for human judgment?
Defining the problem and supplying context
Code suggestions depend on what the tool is asked to do and the context it can use. JetBrains Research identified lack of project-size context among reasons respondents did not use coding assistants. A request that omits system constraints, dependencies, intended behavior, or business rules can produce a plausible change that does not fit the project. The developer’s contribution therefore includes making the task legible and deciding what information is relevant—not just accepting or rejecting a block of code.
Checking correctness, reliability, and security
Generated output needs review against the actual requirements, codebase, and failure modes. The Microsoft study of 860 developers identified reliability and security as priorities for systems-facing work, and transparency and steerability as ways to retain control. These findings point to a practical responsibility: treat suggestions as proposed changes, test them, inspect their effects, and keep a clear path to intervene when the tool is wrong.
Maintaining ownership and control
In an enterprise study of IBM’s internal watsonx Code Assistant, researchers surveyed two user cohorts totaling 669 people and conducted unmoderated usability testing with 15 participants. They found that productivity benefits may not be experienced by all users and raised questions about ownership of and responsibility for generated code. Those results make it unwise to equate tool access with value delivered: a team still needs clear review practices and accountability for what it ships. IBM Research’s report, published April 26, 2025, gives the study details.
Working with people
The Microsoft Research task study found clearer limits for identity- and relationship-centered work, including mentoring. It also identified fairness and inclusiveness as priorities for human-facing tasks. An assistant may help prepare materials or suggest options, but those findings do not support treating mentoring, relationship-building, or people-centered decisions as interchangeable with producing code or text.
Why do studies report different productivity results?
There is no single “AI productivity” measure in these studies. The field experiments estimated completed tasks in particular settings; IBM examined user experience and productivity perceptions in an enterprise; JetBrains asked programmers about practices and preferences; and DORA combined a large survey with qualitative data. A measured change in task completion, a respondent’s sense of productivity, and interest in delegating a task are different kinds of evidence.
Results may also depend on the task’s complexity, developer experience, assistant and study period, codebase, organizational conditions, and the review required before a change is safe to ship. A gain on a bounded implementation task should not be assumed to transfer to unfamiliar legacy code, a security-sensitive change, or an organization with weak testing and review.
Why does the organization matter?
DORA’s 2025 report drew on nearly 5,000 technology professionals worldwide and more than 100 hours of qualitative data. Its central framing is that AI acts as an amplifier: it can magnify strengths in high-performing organizations and dysfunctions in struggling ones. That is the report’s finding and interpretation, not proof that adopting AI automatically improves organizational performance. Google Research’s DORA 2025 report page describes the report.
Rank #4
In practice, an assistant is more likely to be useful when the team has clear requirements, context the developer can provide, and ways to catch mistakes. If those foundations are weak, producing more code faster may also produce more work to review, repair, or reconcile. The relevant question is not only how quickly a draft appears, but whether the team can verify and integrate it.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How should developers adapt their work?
The studies do not prescribe a career plan, but their findings support a practical way to work with assistants without confusing output volume with completed engineering:
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Repair Windows errors before they cause bigger problems3Fix the driver behind crashes, sound loss and screen glitches- Choose bounded tasks first. Try assistance on a change whose intended behavior and scope can be stated clearly. Do not assume results from one kind of task apply to all others.
- Provide project context. Include relevant constraints, expected behavior, and surrounding code or documentation. If the assistant lacks context, treat the answer as a draft that needs more investigation.
- Verify the change. Review the code and run the appropriate tests. For systems-facing work, include reliability and security checks rather than relying on plausibility or speed.
- Keep control and responsibility visible. Make it clear who reviews and owns a change, and ensure the workflow allows people to steer or reject the assistant’s output.
- Spend saved effort where judgment matters. Use time freed from repetitive production for understanding the problem, improving the design, checking edge cases, communicating trade-offs, or supporting teammates—where those are genuinely the project’s needs.
This is a reasoned workflow implication of the studies’ findings on context, safeguards, and task preferences, not a guarantee that every team will save time or that every developer can simply exchange one category of work for another.
Best Value
Does the evidence show that developer jobs will disappear or become more valuable?
No. The sources described here illuminate task use, preferences, productivity measures, and organizational conditions. They do not settle long-term effects on employment, hiring, compensation, or occupational demand. It would be just as unsupported to claim that software engineering jobs as a whole will disappear as to claim they will all be upgraded into higher-level roles.
The more defensible conclusion is narrower: AI can take on or assist with parts of software work, while humans remain important in shaping tasks, supplying context, evaluating results, and taking responsibility for quality and people-facing work. Whether that changes a particular developer’s role or market value depends on the work, the team, and how organizations choose to use the tools.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
Recommended Free Tools
