How we work

From a conversation to software you own.

Five steps, and you can see where a project is at any point in them.

Tell our AI scoping assistant what you want

Instead of booking a call and filling in a form, you talk to an AI assistant we built for this. It takes about fifteen minutes. Show it whatever you already have, a rough sketch, an old quote, a document, and it will ask better questions because of it. You come away with a written summary of what you asked for, which you can correct before it goes anywhere.

We come back with a number and a plan

You know the cost before anything starts, along with what it covers and roughly how long it will take. A person reads and approves every quote, so you are never sent something the AI wrote on its own. If we think the project is a bad idea, or smaller than you expected, we will say so here.

You see rough options first, then a working slice

We start by showing you two or three rough versions side by side rather than one polished one. Any stand-in content is labelled as a stand-in, so nothing looks more settled than it is. Once you have picked a direction, we run a thin version of it end to end on the real system, so the parts that usually go wrong are proven before the rest gets built.

We build it, and check it before you see it

AI writes most of the code, the configuration, and the tests, which is why a small team can take on work that used to need a large one. What it does not do is decide whether the result is right. Someone experienced reads it, runs it, and checks it for security problems before it reaches you. If it is wrong, it goes back.

It moves across to you

The project runs on our accounts while we build it, then moves to yours when it is finished and paid for: the code we wrote for you, your data, and the domains and accounts set up for the project. Third-party components keep the licenses they came with. If you want to hand the work to someone else afterwards, you can, and any upkeep you want from us is a separate agreement rather than something you are locked into.

Where AI fits

AI does the building. We decide whether it is any good.

What AI does

Write most of the code, the queries, and the configuration.
Produce first drafts of designs and documents.
Write the tests and run them.
Handle the repetitive work that used to need a much larger team.

What we do

Decide what should be built, and what should not.
Read everything before it reaches you, and check what it claims against your own data. Software that is confidently wrong looks exactly like software that is right.
Run it properly. On a phone as well as a laptop, and on a bad connection.
Approve every quote and sign every agreement. Nothing reaches you unread.
Turn down work we cannot review properly.

Security

The part of AI-built software you should be asking about.

AI writes code that reads correctly and can still leave a door open, and the mistakes it makes are not the ones a reviewer is expecting to find. Security is what we did before we did this, so what the software exposes and who can reach your data get settled while it is being designed, and gone through again before it is handed over. It is part of the build and never quoted separately.

Before anything is built: what an attacker would want here, what they would try, and what that means the design has to do.
What the software exposes to the internet, and closing anything that does not need to be open.
Where your data sits, who can reach it, and what an attacker would get if they did.
That keys and passwords stay on the server and never reach anyone's browser.
Permissions: who can see what, and what a compromised account can do.
Anything touching money, personal data, or who is allowed in gets read line by line rather than skimmed.
Known vulnerabilities in the third-party components we used, checked before handover rather than at the first incident, and anything left unfixed comes to you in writing with the reason.
What happens under the kind of input nobody expects, including deliberate abuse.
Logs that let you reconstruct what happened, without recording things that should never be written down.

Start with a conversation.

No meeting to schedule. Bring whatever you already have and our AI scoping assistant will ask better questions because of it.