The model
The “brain” — the trained intelligence that reads, writes, sorts, answers, or decides. It's the part that does the actual thinking.
Plain answers
Part 1 · The plain version
How these systems actually work
Before the steps, it helps to know what an AI system actually is — because it's simpler than it sounds. It's software we build on top of two things: a model, and something to run it on.
The hardware
What the model runs on. AI models need a particular kind of chip — a GPU — to run at all. A regular office computer isn't built for it; the GPU is what makes a capable model possible.
Those two — a model doing the work, and a GPU for it to run on — are the foundation every AI system stands on. But a model and a chip on their own don't do anything useful for your business. The actual system is the software we build on top: the part that connects that intelligence to your documents, your tools, and the way you actually work.
The foundation comes first, though, because you have two ways to get it — and we work with both.
Option 1
Use a model run by a big AI lab
The major AI labs run enormous models on their own hardware and let you send your work to them and get the answer back. You don't own or run anything — you pay for what you use.
Good when you're starting small, the volume is low, and the information involved isn't especially sensitive.
Trade-off: your data is sent to that outside company to be processed, and you depend on their pricing, their availability, and their rules.
Option 2
We run a private model for you
Instead of sending your work to an outside company, we run a model on hardware dedicated to you, and operate it on your behalf.
Good when your data is sensitive, you're running enough volume that per-use fees add up, or you want the system under your own control rather than a third party's.
Trade-off: a private setup is more to stand up and run, which is reflected in the cost — but that operational work is ours, not yours.
Neither one is “the right answer” for everyone — it depends on your data, your volume, and how much control you want. We'll be straight with you about which fits your situation during discovery.
Either way, you don't run any of this yourself. We set up and maintain the infrastructure — the model, the GPU, and everything around it — so you never have to think about servers or chips. Keeping it running and supported over time is the ongoing maintenance you'll see quoted in your discovery document.
Part 2 · The engagement
What happens if you reach out
01
Discovery phase
We start with a discovery engagement — where we learn your operation, the problem you want solved, and what it would actually take to solve it. How long it takes depends on the size of the automation: it might be a single day, or up to a couple of weeks for something larger. The cost is based on that time, so you only pay for the work the discovery actually requires.
02
Your discovery document
At the end you receive a document laying out the whole picture: scope (exactly what would be built and what it would do), security (how your data and systems are handled), timeframe, a quote for the build, and ongoing maintenance. It's written to be clear enough that you can make a real decision from it — not a sales teaser.
03
The document is yours
Once discovery is paid, that document is yours to keep. If you want to take it for a second opinion, or get competing quotes elsewhere, you're free to. There's no obligation to continue with us. We work this way on purpose — good scoping should stand on its own, and if our plan and our price hold up against the market, that's the point.
04
The build
If you decide to move forward, we build it. The quote holds for the work defined in your discovery document, so the price you agreed to is the price for what was agreed. If you later want to add or change something beyond that scope, that's a straightforward conversation, not a surprise on an invoice. When it's done, your system is running — and the maintenance you saw quoted is what keeps it that way.
Ready to see what it would take?
Tell us about your operation and what you'd want to automate, and we'll take it from there.