AI software development services
An overview of everything we build with AI, what it means and how we compare with other kinds of provider.
How we work
Our AI-assisted software development process, step by step
In our AI-assisted software development process, AI coding tools write most of the first drafts, and engineers decide the architecture, review every change, test it and stay accountable for what goes live. The work runs in six steps, from a discovery call to support after launch, and you see working software every week. This page shows who does what, what you receive at each step and what we need from you.
- Every change reviewed by a person
- Weekly demos
- Full code ownership
AI and people
Who does what: the AI tool and the engineer
"AI-assisted" is a vague phrase, so here is the division of work we actually use. The rule behind it is simple: the AI can draft anything, and a person is responsible for everything.
| Work | What the AI tool does | What the engineer does |
|---|---|---|
| Architecture and data model | Suggests options and trade-offs | Decides, and writes the decision down |
| Screens, forms and routine code | Writes the first version | Reviews it, simplifies it and fits it to the rest of the product |
| Login, payments and access rules | Assists with drafts | Leads the work and reads every line, because mistakes here cost money or data |
| Automated tests | Drafts test cases | Chooses what must be tested and checks that each test can actually fail |
| Bug fixing | Proposes a cause and a fix | Confirms the cause, then checks that the fix does not break something else |
| Deployment and secrets | Not involved with real credentials | Sets up environments, keys, monitoring and backups |
The tools we use for this are Claude and Cursor. The choice of tool matters less than the review that follows it.
Process
The six steps, and what you receive at each
The same six steps apply to an MVP, a larger product or the rescue of an existing app. Only their length changes.
Discovery
A short call about your goals, your users and the features you cannot launch without.
You receive: A written scope, a timeline range and an estimate within 24 hours.
AI-assisted build
Engineers set up the repository, the architecture and the database, then use AI coding tools to build features on that foundation.
You receive: Access to the repository from the first commit, and a running version to click through.
Human code review
A senior engineer reads every change for correctness, security and maintainability before it is merged. Nothing reaches the main branch unreviewed.
You receive: A history of reviewed changes in your repository that any developer can follow.
QA and testing
Automated tests cover the critical paths. People then test by hand on real devices and browsers, looking for what the tests cannot see.
You receive: A test suite that runs on every change, and a list of what was checked by hand.
Launch
We deploy to production, set up monitoring and backups, and handle App Store and Google Play submission for mobile apps.
You receive: The product live on your domain or in the stores.
Support
After launch we stay available for fixes and new features, or we hand the project to your own team.
You receive: Documentation and a setup guide, so the project never depends on us.
What we need from you
A project goes at the speed of its decisions. These are the things only you can provide:
- One person who can decide. Questions answered in a day keep the weekly rhythm; questions that wait for a committee do not.
- Time for the weekly demo. You try the software and tell us what to change or what comes next.
- Access when it is needed: accounts for hosting, payments, the app stores and any system we connect to.
- Real content and examples. Sample data, texts and the awkward cases from your business that we could not guess.
Code ownership and confidentiality
- Your repository from day one. The code lives in your GitHub or GitLab account, not ours.
- Full intellectual property rights, with documentation and deployment access. There is no lock-in.
- NDA before details. We sign yours or send ours before you share anything about the product.
- If we stop, you keep everything: the code, the history of changes, the tests and the documentation up to that day.
Checklist
Eight questions to ask any agency that builds with AI
Every agency now says it uses AI. These questions show whether there is a process behind the claim. Ask them of us too; our answers are on this page.
Get a free project estimate in 24h- Who reviews the code the AI writes?Look for a named senior engineer and a review before every merge, not a spot check.
- Where does the code live during the project?It should be in your repository from the start, not handed over at the end.
- What is covered by automated tests?At least sign-up, payment and the core action of the product.
- How do you handle secrets and customer data?Keys on the server only, and a scan for leaked secrets before release.
- How often will I see working software?Weekly is a good rhythm. "At the end" is a warning sign.
- What happens when I want to change the scope?There should be a clear rule, and an estimate before extra work begins.
- What do I own if we stop halfway?Code, history, tests and documentation, without conditions.
- Who runs it after launch?Either they do, under agreed terms, or your team can, with documentation.
FAQ
Questions about our development process
Which AI coding tools do you use?
Mainly Claude and Cursor. We choose tools that work directly on a normal code repository, so nothing about your project depends on a particular AI product continuing to exist.
Does an AI-assisted process mean lower quality?
Not when the review is real. AI makes the typing faster; quality still comes from the same sources as before: a sound architecture, code review, tests and QA. Those steps are not shortened.
How is this different from a traditional agency process?
The steps are the same. What changes is the proportion of time: less goes into writing routine code, and more into design decisions, review and testing. That is why timelines are measured in weeks.
Can I look at the code while the project is running?
Yes. It is your repository, so you or your own developer can read every change as it is made, and comment on it.
Where to go next
AI MVP development
Launch a focused first version in weeks: scope cut to what matters, built with AI tools and reviewed by engineers.
Fix a vibe coded app
Take a Lovable, Bolt, Replit or Cursor prototype to production: audit, security fixes, tests and a proper deployment.
Want to see the process on your own project?
Tell us what you want to build. Within 24 hours you get a clear scope, a timeline range and a fixed estimate — free, with no obligation.
Get a free project estimate in 24h