Ape Foundry · AI that grows with your company
The AI handles the routine work.
You get back to what matters.
Writing minutes, summarising documents, preparing drafts, searching old files: for that, we set up your team's own AI workspace, accessible through a browser. It uses your documents and follows your instructions. The AI does the groundwork; a person checks it and signs it off. We run it for you, and we check every new version before your team works with it.
Where the workspace isn't enough, we build software. For the process that needs an application of its own. For the spreadsheet half the company runs on, the one with the note that says “Please don't touch anything.” That's where we start: first we find out what depends on it, then we change something.
How we work
Your process first, then the AI.
a
The problem before the spec
We start with what costs you time today, not with a list of requirements. Sometimes a different process helps more than automating the current one. You already have a spec? Good. We will still ask which problem it is meant to solve, tell you what we see, and you decide.
b
AI where it does real work
A model reads, sorts and pulls out details. Whatever can be calculated or checked against fixed rules, code does. We develop with AI too: some of our code is written by a model. Whether it goes live is a person's decision.
c
We run it, you stay free
We run your AI workspace, on open-source software that anyone may run. You can leave anytime with your working instructions and templates. Software we build for you comes as source code, with the right to use it, change it and have others maintain it. Who runs it is your call.
d
What we don't sell
We don't hire out developers to work under your direction: we build and answer for a result. We take over existing software only after an assessment, and sometimes that assessment says no. And if you don't need new software, or don't need AI, we will tell you.
e
Your data in the AI workspace
Your AI workspace is yours alone: it runs separately from every other company's. Only the people you give access can sign in. You decide which AI models it may use, and we set a budget for their use that you choose.
f
Your data in software we build
Our AI tools never see your real data while we build: our automated tests use made-up or anonymised data. You choose the model and where it runs, if you like an open-weight model on your own servers or in the EU. Before development begins, we tell you whether it can do your task well enough.
Our own products
When we are our own client.
We also build our own products, each under its own brand. For our products, we write tests for every change and record every consequential decision, although nobody but us insists on it. They show you how we approach a problem before you trust us with yours. The mould next to them stays empty until something finished belongs in it.
ImmoChecker.ai
Saturday evening. On the table: the listing, the land register extract, the energy certificate and three years' worth of owners' meeting minutes. The viewing is on Wednesday. The buyer is no expert. The usual routes: hire a surveyor, or read everything yourself.
ImmoChecker is being built to read the listing and the documents itself and say what needs settling before the purchase, without a form and without waiting for Monday.
We trust the model with a lot. Not with arithmetic: it pulls out the facts, and code computes every number.
Contact
Let's talk.
Tell us what's holding up the work. A few lines will do. You'll talk to the people who would actually do the work. Once we understand what matters most, you get a proposal with the next steps we recommend.
Applications, procurement, legal: same address.