Most businesses don’t need more tools.They need one system.
The enquiry arrives in one place, the work happens in another, the record lives in a third, and the part that holds it together is somebody remembering. We build the system that carries all of it — in your repository, in your cloud, yours at the end.
If you already have software other companies are asking to use, start here instead.
Where this usually starts
Nothing is broken. It just doesn’t connect.
Every tool in the stack works. The CRM works, the inbox works, the spreadsheet works. What does not work is the space between them, and that space is being held together by people who remember what happens next.
The same information, typed three times
Once into the messaging app, once into the sheet, once into whatever produces the invoice. Every copy is a chance to be wrong.
Nobody can answer “what happened”
The history of a customer is spread across an inbox, a chat thread and one person’s memory. Reconstructing it takes an afternoon.
Follow-up depends on who is in
The things that get chased are the things somebody remembered to chase. Everything else quietly ages.
None of this shows up as a line item. It shows up as capacity: the business cannot take on more work without taking on more people.
What gets built
One place the work actually lives.
Not a dashboard on top of the tools you already have. The system the operation runs inside, with the tools connected to it.
Where to begin
You don’t commission the whole thing.
Pick the one that is costing you most right now. It ships on its own, it is useful on its own, and it becomes part of the system rather than another tool to connect later.
The calls you miss
Every enquiry answered and qualified, at 9pm and on Sunday, landing in one place with the context attached instead of a voicemail nobody plays.
Usually first when the phone is the front door.
The quote that never got sent
Intake to quote to follow-up, chased on a schedule rather than by whoever remembered. Most lost work is not lost to a competitor. It ages out.
Usually first when work is won by proposal.
The job nobody can find
One record per customer that holds every message, document and decision — so answering “what happened here” takes a search, not an afternoon.
Usually first when the team has grown past what one head holds.
What only lives in one person
The pricing rules, the exceptions, the way it is actually done. Written into the system so it survives that person taking a holiday, or leaving.
Usually first when the owner is the bottleneck.
The line that matters
AI explains and drafts. It doesn’t decide.
The useful question is never whether to use AI. It is which parts of the work are allowed to be probabilistic, and that line gets drawn per business, in writing, before anything is built.
Where a person still reviews the output
- Answering questions from your own documents, with a source attached
- Drafting replies in your business’s own voice
- Summarising a long thread into what actually matters
- Triaging what is urgent and what can wait
- Tagging and routing, with a human able to override
Where it has to be deterministic
- Anything that affects what someone gets paid
- Prices, quotes and contract terms
- Permissions and who can see what
- Reminders and billing, which must simply be correct
- Any number a person will be asked to explain months later
On one system we operate, the moment commission entered the picture that path had to become deterministic — not because the model was inaccurate, but because a number affecting someone’s pay has to be reproducible by a human, months later.
What you would actually get
A real one, read off the repository.
One operating system built for a property agency, running that company’s entire client-facing operation. These are counts, not estimates.
4,700 listings, every enquiry from three channels in one inbox, in two languages. In its first four months activity went from 828 actions to 9,689 while the team went from 5 people to 20 — and actions per person stayed flat. What that data taught us.
How it runs
Sit with the owner. Sit on the floor. Then build it.
Four steps, and the first two happen before any code is written. Half of what people tell you about their own process is the version they wish were true.
Sit with the owner
What the business actually sells, and what it is afraid of. Different questions, both of them load-bearing.
Sit on the floor
Watch how the work happens, at the real desk, at the speed it really runs.
Build it narrow
One workflow, end to end, live. Not a platform. The thing that has to work before anything else is worth building.
Operate it
Owning it in production is where the learning starts, and where you find out which clever ideas were actually clever.
What stays yours
Your repository. Your cloud. Your data.
Built where it will live
Your accounts from the first commit. Nothing gets rebuilt later in somewhere we control, because it was never anywhere else.
No core licensed back
What we build for you is yours. We do not keep the engine and rent it to you afterwards.
Conventional on purpose
Ordinary architecture another engineer can pick up. The clever framework nobody has heard of is a liability, not an asset.
Documented decisions and a team that can carry it. If you would rather we kept operating it, that is a choice — not something the architecture forces.
Before you get in touch
Questions operators ask
Do we own what you build?
Yes. We work inside your repository and your cloud from the first commit, so there is no handover of ownership at the end because it was never ours. The code, the data and the accounts are yours throughout.
How is this different from hiring a development agency?
Most of the work is not the code. It is sitting with the people doing the job and finding out how the operation actually runs, which is different from how anyone describes it. An agency builds to a specification. This starts before there is one.
What if we already have software that half works?
Then this is probably the wrong page. If a system already runs your business and other companies have asked to use it, the question is customer two, not the first build.
How long does it take?
The first working version is usually weeks rather than months, because it is deliberately narrow. What takes longer is operating it against real traffic and fixing what that reveals, which is the part that decides whether it survives.
Where does AI actually go?
Retrieval, drafting, summarising and triage - work where a person reviews the output before it counts. Anything touching money, pricing or a contract stays deterministic and auditable. That line is drawn per business, in writing, before anything is built.
Can our own team maintain it afterwards?
That is the intended end state. Conventional architecture, documented decisions, and no framework nobody else has heard of. If you would rather we kept operating it, that is an option too, but it should be a choice rather than a dependency.
Being honest about it
When you shouldn’t build anything.
If an off-the-shelf product does 90% of what you need, buy it. Custom software is worth it when the way you work is the business — when the sequence, the exceptions and the judgement are the reason customers choose you, and a generic tool would make you work like everybody else. If that is not true, the honest answer is a subscription and a good afternoon of configuration.
The same applies to the AI part. If the work is deterministic and already well handled by software, adding a model makes it slower, more expensive and harder to explain.
Tell us how the work actually happens.
A conversation about your operation, not a demo. What comes in, who touches it, where it gets stuck, and what only works because someone remembers. If building something is the wrong answer, we will say so.
