AI augmented software development where engineers own every line they merge

In AI augmented software development, our engineers work only with the AI tools you approve, and nothing from your project reaches any other tool.

Abstract 3D illustration of an engineer's code editor with an AI assistant panel and a review checkmark before mergeAbstract 3D illustration of an engineer's code editor with an AI assistant panel and a review checkmark before merge

Definition

What is AI-augmented development?

Engineers use AI tools to draft and check code across the software development lifecycle, and a human engineer keeps control of the design, the logic and every change that merges. The tool proposes. The engineer decides what to keep, rewrites what is wrong and answers for the result in review.

It also goes by AI-assisted development, and some vendors call it AI-native. Here it means AI used to build your software. AI development as a product capability, where AI runs inside the software you ship, is different work with its own page.

Whether a vendor's engineers use AI tools is a yes-or-no question, and the answer tells you little about your code. Ask who decided what those tools may touch, and who read what they wrote before it reached your main branch. Done that way, the tools move an engineer's attention away from typing code and toward the design and the tests.

Abstract 3D illustration of an engineer at a code editor weighing a suggestion drafted by an AI assistantAbstract 3D illustration of an engineer at a code editor weighing a suggestion drafted by an AI assistant

Your code, your rules

Who decides what the AI tools can touch?

Your list decides the tools and the accounts

You tell us which AI tools may be used on your project, and under which accounts. The account matters because the same assistant can come with different data terms depending on whose account it runs under. We have no house tool to defend, so if your security team approved one assistant and nothing else, that one assistant is the whole toolset.

Nothing goes to a tool you haven't allowed

Nothing from your project, whether code, data or a pasted error log, goes into a tool that isn't on your list. If you'd rather have no AI tools on the project at all, there are none.

The list is written into the engagement

The tool list, including a decision to allow none, is agreed at the start and written into the engagement, and changes to it go through us. Code our engineers write with an AI tool is yours on payment, like everything else we deliver.

Why every AI draft gets read

AI coding tools can call an API or import a package that doesn't exist. They can suggest a pattern that used to be common and is now known to be insecure. And they can write tests that pass because the same tool wrote the code under test. Each looks fine at a glance, and a test suite that agrees with its code proves less when one tool wrote both. Spotting them takes a reviewer who knows the library well enough to notice a call that isn't in it.

Where our own systems are hosted

Systems we control are hosted in the EU, and we work under GDPR as an EU company. Where we handle your data, we sign a DPA. Your repositories and your cloud stay wherever you keep them, and an AI tool you approve runs wherever its provider runs it, so the hosting statement covers neither.

Abstract 3D illustration of a code repository behind a gate that lets through only approved tool iconsAbstract 3D illustration of a code repository behind a gate that lets through only approved tool icons

From task to merge

How is AI-written code reviewed?

Whichever AI coding tools helped, the engineer who opens the pull request owns the change. Four steps take it from task to merge. Our testers run testing around AI-written code the way they run it for any other code.

  1. STEP 1

    The engineer sets the task

    The engineer decides what to ask the tool for.

  2. STEP 2

    The tool drafts, the engineer checks

    Before opening a pull request, the engineer reads and corrects the draft and runs the code and its tests.

  3. STEP 3

    Another AI agent reviews

    A tool you approved, run in a different session from the one that wrote the change, reads it fresh and looks for mistakes before a person does.

  4. STEP 4

    Human review, then merge

    A human engineer then reviews the change, and it merges only after both reviews.

AI coding tools

AI coding tools our engineers work with

Our engineers work with these five tools. They differ less in what they can do than in where you work with them: inside the editor, in the terminal, or in a chat window beside the code. Each one drafts code that an engineer then reads, so the review steps stay the same whichever is in use.

Our advice is to approve fewer tools rather than more. One editor assistant that everyone knows well is worth more than five that each engineer half-uses, because a reviewer spots a tool's habits only after seeing them often.

GitHub Copilot in the editor

The least change to a team's day. It suggests code as you type in VS Code or a JetBrains IDE, answers in a chat panel, and in agent mode works through multi-step tasks.

Cursor, an AI code editor

Worth the switch for engineers who want AI everywhere in the editor. It is a fork of VS Code, so extensions carry over, and its agent applies edits across files and shows its changes in a diff view.

Claude Code in the terminal

Suits engineers who live on the command line. Anthropic's agent runs in the terminal, reads the codebase, runs commands and tests, and has permission modes that can require approval for edits.

ChatGPT / Codex by OpenAI

Two modes from one provider. ChatGPT is for talking a problem through, and Codex, OpenAI's coding agent, takes a scoped task and hands back a change to review.

OpenCode, open-source agent

Useful when your rules name the model provider. This open-source terminal agent works with models from many providers, so you choose which one sees your code.

Bring your AI tool rules to the first call

Tell us which AI tools your team allows and where your code lives, and we will show where our engineers fit inside those rules.

How it is bought

Is AI-augmented engineering a separate service?

No. It is how our engineers work in every engagement, and it has no price of its own.

It comes with every engagement model we run, whether engineers join the team you manage or we deliver a project for you.

Hiring for one stack? Our hire pages list engineers who use AI tools by technology and by role.

To see where this practice sits in the rest of our work, read how our services fit together.

Abstract 3D illustration of an engineer card with an AI tool badge joining a client's team boardAbstract 3D illustration of an engineer card with an AI tool badge joining a client's team board

Terms of work

Engagement models and terms

Project delivery

We deliver an agreed scope end to end: plan, team, milestones, QA and release. You make the product decisions and accept each milestone.

Fixed price per milestone after discovery, or time and materials against an estimate.

  • Scope changes go through change requests.
  • On fixed-price projects, a warranty period agreed in the contract, usually 30–60 days.

Staff augmentation

Our engineers join your team and work in your process and tools. You manage them; we source, vet and contract them.

Hourly rate per engineer by role and seniority, invoiced monthly.

  • If an engineer isn't the right fit, we replace them within 14 days at no extra cost.
  • Knowledge transfer to the replacement engineer isn't billed.
  • There is no free trial period.
  • 30 days' notice to take an engineer off or end the engagement, with no minimum term beyond it.
  • Out-of-hours on-call only if agreed with you; we set its hours and billing together.
  • After two years, moving an engineer onto your own staff is something we can discuss.

Whichever model you choose

  • We reply to an inquiry from the contact form within 24 hours.
  • We sign an NDA before work starts, and IP passes to you on payment.

Walk us through your repository

"Writing code got cheap. Reading it closely didn't, so that is where our engineers spend their time, inside the rules you set."

Dzmitry Kuchynski, Chief Operating Officer, Octalance

FAQ

What buyers ask before AI touches their code

Can AI build our software without an engineering team?

It can draft a feature from a prompt, and sometimes a small app. It can't decide what is worth building, or answer for the code when it breaks in production. Someone has to own those calls, and on our projects that is an engineer.

Who owns the code an AI tool writes?

You do. Everything we deliver is yours on payment, including code our engineers wrote with an AI tool. Each tool also comes with its provider's own terms, so if your legal team wants to read them, the moment to do it is before the tool goes on your list.

Does our code train the AI tools?

No. Training on your code is switched off on the tool accounts we use. If our engineers work under an account you issue, check that setting before they start.

Can your engineers work on our project without AI tools?

Yes, and a rule can be narrower than yes or no, for example AI tools allowed on test code but kept away from a payments module. We exclude the folders you name from the tools, and the engineers follow your rule as well.

How much faster will our project go with AI tools?

We don't quote a speed-up figure. When a vendor does, ask what it was measured against and whether the time spent reviewing the AI's drafts was counted. A tool can make typing faster and leave a release no faster at all if the review queue grows.

Can we change the approved AI tools during the project?

Yes. The list is yours, and our engineers work from the new version once the change reaches us. Removing a tool is the simple direction. Adding one deserves the check you gave the first list, including whose account it will run under.

Which AI coding tool should we approve?

Start with what your security review needs to know about each candidate: where prompts and code are processed, how long the provider keeps them, and whether the tool can run commands on its own. Then try the shortlist on real tickets for a few weeks before you settle on one.