HOW WE WORK
In order, from the first conversation to the system running in your business. The first step is free, the second is a single page, and only the third costs anything - by which point you have seen enough of how we work to judge whether the rest is worth it.
The call
Half an hour, free. Bring three real items from the pile - yesterday's actual quote, claim, order, or email thread. Three real things turn "we're drowning in admin" into work we can both look at, and they tell us more than any description of the problem could.
You might leave that call with an answer that isn't us. If what you need already exists off the shelf, we will say which one. If the trouble is really that you cannot get answers out of your own data, that is a different trade, and we will point you at someone who does it well. Both of those are good outcomes, and they cost you nothing.
The scope note
Within a day, one page: what we would look at, what you get, a fixed fee, a fixed time, and what it will not do. It is written to be forwarded. If it has to go in front of a partner, a finance director, or the person who will actually use the system, it should read the same to them as it does to you.
The evaluation
The first paid step, and the only one before a build. We map how the work actually moves through your business, where the hours go, and what the tools you already have are and are not doing for you. Nothing gets built before this says what is worth building.
What comes back: what to buy off the shelf, what is worth building on top of it, and what to leave alone - plus, if we find it, the work an agent could be doing that nobody has thought to ask for. Each one scoped, with the point where a person signs it off, and what it will not do.
The build
The stack, the integrations, and the point where a person signs off are decided and written down before anything is configured. Then it is built to how your business already runs, rather than asking you to run the way the software expects. That is the part the off-the-shelf tools cannot do, and it is why the evaluation comes first.
Before it goes live
Tested against real cases from your own pile, including the ones where things break. What the system does when something it depends on stops answering is part of the specification: it says so plainly, hands the work back to a person, and leaves a trace you can read afterwards. More on the failure modes that shape this: where agents go wrong.
Handover, then tending
The work lands in your repositories - documented, tested, yours - and your people are shown how to run it and how to read what it reports.
The context behind an agent does need keeping current: policies change, formats change, the business changes. So tending is available and priced honestly, and you choose whether to take it. You should be able to walk away with the thing working, and know that you could.
Throughout, the line is the one we draw everywhere: the agent takes the checking that has to come out the same way every time, and the judgement stays with your team.
Half an hour, free, and you keep whatever comes out of it - book a time or write to vinga@skyforge.sh