Skip to main content

AI in software engineering

EveryoneStandardUpdated October 3, 2026
Related

Mission and core values · Our philosophy · Using AI with Exalynt · Using AI to amplify value

This document states Exalynt's position on artificial intelligence (AI) in software engineering. It applies to all Exalynt work, for engineers and clients alike.

AI amplifies software engineering. It does not replace it.

Read next: Using AI with Exalynt for clients · Using AI to amplify value for engineers

Our position​

  1. We embrace AI and encourage everyone to use it. Exalynt uses AI throughout its work and regards it as a positive development for software engineering. Engineers and clients, technical or not, are encouraged to use it for the tedious parts of building software, such as drafting, summarizing, researching, routine code, tests, and documentation.
  2. AI amplifies the practice of software engineering. It does not replace it. Understanding the problem, design, tradeoffs, verification, operation, and maintenance remain the work of the discipline, and AI lets that work be done faster. Like calculators and computer-aided design before it, AI moves engineers to a higher level of abstraction: less time on mechanics, more on what is being built and whether it is sound.
  3. People own the result, and all work meets one standard. Whoever or whatever produced the work, the Exalynt engineer who ships it understands it, has verified it to its stability level, and stands behind it. Work produced with AI is judged like any other: whether it solves the right problem, whether it is understood, and whether it has been verified to the level it claims. Take ownership Pursue excellence
  4. The aim is quality value, not output. Exalynt values the sustainable production of quality value. Volume of code or activity is not a measure of success, whether or not AI produced it.
  5. Client policies, client data, and openness come first. A client's policies on AI tools take precedence for that client's work, and secrets and real client data are not entered into AI tools. We answer plainly when asked how AI was used, and we never describe work as reviewed, tested, or verified unless it was. Be honest and transparent

We measure where AI actually helps, and revise this position as the tools and the evidence change. The date at the top of this page shows when it last changed.

What software engineering is​

The official definition

software engineering noun

The application of a systematic, disciplined, quantifiable approach to the development, operation, and maintenance of software; that is, the application of engineering to software.

ISO/IEC/IEEE 24765, Systems and software engineering: Vocabulary, originally IEEE Std 610.12-1990

The diagram below groups the discipline's 18 knowledge areas, as defined by the IEEE Computer Society's SWEBOK V4.0a, into the phases of a system's life, and shows where AI amplifies each one. In this document, AI means generative AI tools, such as coding and chat assistants, that produce code, text, or other content from instructions.

Software engineering has 18 knowledge areas. Writing the code is one of them.

    • Software Requirements
    • Software Engineering Economics
    1 Understand
    • Software Architecture
    • Software Design
    • Software Engineering Models and Methods
    2 Design
    • Software Construction
    • Software Configuration Management
    3 Build
    • Software Testing
    • Software Quality
    • Software Security
    4 Verify
    • Software Engineering Operations
    5 Run
    • Software Maintenance
    6 Evolve
    • Software Engineering Management
    • Software Engineering Process
    • Software Engineering Professional Practice
    All phases
    • Computing Foundations
    • Mathematical Foundations
    • Engineering Foundations
    Foundations
  1. Understand

    What's needed, and whether software should solve it.

    AI amplifies
    Researching existing tools, and turning notes into clear requirements.
    People decide
    Which problem is worth solving, and whether it needs software.

    Knowledge areas · 2 of 18

    • Software Requirements
    • Software Engineering Economics
  2. Design

    How the pieces fit together.

    AI amplifies
    Laying out options, and building quick prototypes and mockups.
    People decide
    Which tradeoffs fit the budget, users, and risks.

    Knowledge areas · 3 of 18

    • Software Architecture
    • Software Design
    • Software Engineering Models and Methods
  3. BuildAI speeds this up most

    Writing the code.

    AI amplifies
    Routine code, migrations, and boilerplate.
    People decide
    How it fits the system, and what it should not do.

    Knowledge areas · 2 of 18

    • Software Construction (writing the code)
    • Software Configuration Management
  4. Verify

    Knowing it works, not just that it runs.

    AI amplifies
    Generating tests, and spotting possible defects and security issues.
    People decide
    What must be true, and how we know it is.

    Knowledge areas · 3 of 18

    • Software Testing
    • Software Quality
    • Software Security
  5. Run

    Deploying it, watching it, and recovering when something fails.

    AI amplifies
    Drafting runbooks, and summarizing logs and incidents.
    People decide
    When to act, and what users need while we do.

    Knowledge areas · 1 of 18

    • Software Engineering Operations
  6. Evolve

    Changing it safely as needs change.

    AI amplifies
    Explaining unfamiliar code, refactoring, and documentation.
    People decide
    What to change next, and what must stay stable.

    Knowledge areas · 1 of 18

    • Software Maintenance
Across every phasePlanning, managing, and improving how the work gets done.

Knowledge areas · 3 of 18

  • Software Engineering Management
  • Software Engineering Process
  • Software Engineering Professional Practice
FoundationsThe computing, mathematics, and engineering everything else rests on.

Knowledge areas · 3 of 18

  • Computing Foundations
  • Mathematical Foundations
  • Engineering Foundations
The knowledge areas are those of IEEE Computer Society, SWEBOK V4.0a, by their SWEBOK names. The six phases are Exalynt's grouping, for readability.

AI can help produce answers. Engineering is the discipline of deciding which questions matter, evaluating the answers, making tradeoffs, and taking responsibility for the outcome.

Supporting evidence​

Statements 1 to 3 rest on published research, summarized below with the limits of what it shows. Statements 4 and 5 rest on Exalynt's core values.

ClaimSupportsKey finding
AI makes software development fasterStatement 126.08% more tasks completed with an AI coding assistant
AI amplifies existing practicesStatement 2AI adoption is associated with higher throughput and lower stability
AI output needs verificationStatement 346% of developers distrust the accuracy of AI output; 33% trust it
AI's benefit must be measuredReviewing this positionExperienced developers expected a speed-up and took longer

AI makes software development faster​

  • Randomized field experiments at Microsoft, Accenture, and another Fortune 100 company found that 4,867 developers given an AI coding assistant completed 26.08% more tasks (Microsoft Research, 2025).
  • In DORA's 2025 survey of nearly 5,000 technology professionals, 90% use AI at work and more than 80% believe it has increased their productivity (DORA, 2025; Google Cloud, 2025).
  • 84% of developers use or plan to use AI tools, and 51% of professional developers use them daily (Stack Overflow, 2025).

Limits: these measure output and adoption, not whether the results are correct or valuable, which is why statement 4 measures success by quality value.

AI amplifies existing practices​

  • DORA describes AI as an amplifier, "magnifying an organization's existing strengths and weaknesses." Its 2025 research associated AI adoption with higher software delivery throughput and lower delivery stability, and found that without strong automated testing, version control practices, and fast feedback loops, more change leads to instability (Google Cloud, 2025).
  • A March 2026 DORA analysis found that time saved in writing code is often reallocated to auditing and verification (DORA, 2026).

Limits and what it means: these are associations, not proof of cause. They are why we pair AI with automated tests, review, fast feedback, and stability levels.

AI output needs verification​

  • 46% of developers distrust the accuracy of AI output versus 33% who trust it, and 3% highly trust it. 66% are frustrated by AI answers that are "almost right, but not quite," and 45% find debugging AI-generated code more time-consuming (Stack Overflow, 2025).
  • Developers are most reluctant to hand AI high-responsibility work: 76% do not plan to use it for deployment and monitoring, and 69% do not plan to use it for project planning.

Limits and what it means: these are developers' reported views, not measurements of accuracy. They show that the people closest to the work check it, as our engineers do, to each piece of work's stability level.

AI's benefit must be measured​

  • 19% longerto complete tasks with AI allowed, for experienced developers who expected to be 24% faster. Early-2025 tools, one setting.METR, 2025 (opens in a new tab)
  • In a randomized controlled trial, 16 experienced open-source developers completed 246 real tasks in mature repositories they knew well, using early-2025 AI tools. They expected AI to make them 24% faster; tasks took 19% longer with AI allowed (METR, 2025).

Limits and what it means: METR states that this is a snapshot of early-2025 tools in one setting, and does not show that AI slows down most developers. Beside the Microsoft study, it shows that results vary by setting, so we measure where AI helps rather than assume it.

note

This document describes how Exalynt approaches engineering. It is not a client agreement, warranty, or guarantee.

Sources​

Frequently asked questions​

Does Exalynt use AI?

Yes. Exalynt uses AI throughout its work and encourages every engineer to use it wherever it helps. The engineer who ships the work remains responsible for it. See Using AI to amplify value.

Are clients encouraged to use AI?

Yes. Clients, technical or not, are encouraged to use AI to describe problems, research options, mock up ideas, and prepare feedback. See Using AI with Exalynt.

Is software engineering the same as writing code?

No. Software engineering is the application of engineering to the development, operation, and maintenance of software. Writing the code is one of its 18 knowledge areas. See What software engineering is.

Will AI replace software engineers?

Nobody can say exactly how the work will change. The U.S. Bureau of Labor Statistics projects employment of software developers to grow 15.8% from 2024 to 2034, adding about 267,700 jobs, with demand for AI-based systems contributing to that growth (BLS, 2026). Projections cannot establish what AI will change, but they do not anticipate a declining need for the discipline. Our position is that AI amplifies software engineering rather than replacing it.

Who is responsible for work produced with AI?

The Exalynt engineer who ships it, in the same way as for any other work. They understand it, have verified it to its stability level, and stand behind it. This describes how we work; it is not a warranty, and a client's agreement with Exalynt governs any commitments.

What if a client does not want AI used on their work?

The client's policy takes precedence for that client's work. We follow it, and tell the client if it changes what a block can deliver.