The proposal lands in 48 hours, with a fixed price
The diagnostic comes first; then two to eight weeks of building, and the handover. This is what happens in each phase and what it ends with.
48 h from the call to the written proposal · 3 actionable findings or the diagnostic is free · 100% of the code and accounts end up in your name
Built on
An engineering team that has already been on the other side.
Ninety seconds: who we are, how we work and what you get at the end.
From the first call to a system in production
01
1 · Diagnostic
Fifteen minutes, free. You leave with one of three answers: it gets fixed, it gets built, or it is not worth doing. If you already have something answering badly, we measure the system in a week and hand back the failures ordered by impact. Without three actionable findings, there is no invoice.
02
2 · Proposal within 48 hours
Fixed scope, price and date, in writing. It says what is out of scope, which matters as much as what is in, and the numbers that will tell you whether it worked. If the project does not fit, the proposal says so.
03
3 · Build
Two to eight weeks. We ship in parts you can watch running, not in percentage updates. If something changes the scope, we raise it when it surfaces and you decide before it is spent.
04
4 · Handover
The system stays running and the whole annex ends up in your name. No mandatory monthly fee. Support is separate and decided afterwards, with the system already in production.
01
The annex
What ends up in your name
This list goes in the proposal, not in the conversation. It is the difference between buying a system and renting one.
-
Repository
All the code with its history, under your name
If you change providers, the work does not leave with us
-
Infrastructure and credentials
Cloud accounts, keys and access, in your name from day one
No account of ours left in the middle
-
Operating documentation
How it deploys, what to monitor and what it costs per month
Another team picks it up without doing archaeology
-
Evaluation set and baseline
Real questions with their correct answers, measured before going to production
You keep the yardstick to measure anyone later, ourselves included
-
Store publication
App Store and Google Play under your company
Mobile only, and where a client most often ends up tied in
The evaluation set is included in RAG, agents and rescue: without it there is no performance guarantee to sign. In chatbots and automation it is optional. In software, mobile and integrations it does not apply.
What we do not do
No three-week discovery invoiced before knowing whether there is a project, no innovation workshop, no deck with your logo on the first slide.
No monthly subscription to keep running something you already paid for. If you want support, it is a separate contract signed afterwards, with the system in production and with real data on how much actually needs maintaining.
Key takeaways
Four things before the call
- 01
Four phases: diagnostic, proposal within 48 hours, build and handover. It starts with a free fifteen-minute call.
- 02
The proposal carries fixed scope, price and date. It is not an estimate with ranges: it is the figure that gets invoiced.
- 03
If the system already exists and answers badly, the first phase is a one-week measured diagnostic, with a guarantee: if it does not reach three actionable findings, there is no invoice.
- 04
At handover you keep the repository, the infrastructure, the credentials, the operating documentation and the evaluation set with its measured baseline — the yardstick you can hold any later provider to, ourselves included.
A RAG system in production: three years, two million users
We built and operated the assistant for a loyalty platform serving more than two million active users across ten countries.
We built the full pipeline: document ingestion and normalization, chunking, embedding generation, vector store on Azure AI Search and Pinecone, and retrieval with grounded generation on LangChain.
We held it above 99% availability for three years.
RAG systems →Frequently asked questions
01 What happens if the scope changes mid-project?
It gets raised when it surfaces, not at the end. It is quoted separately, approved or not, and the original scope keeps its date. What we do not do is absorb it silently and deliver late: a scope change discussed in time is a decision, one discovered at handover is an argument.
02 Do you work with our team or separately?
Either way, and it is worth deciding in the proposal. When there is an internal team, the usual split is that we build the AI side and they build the integrations into their own systems, which is where their knowledge is worth more than ours. When there is no team, we hand over with documentation written so someone can pick it up.
03 What if the diagnostic says it is not worth doing?
It gets delivered anyway, with the reasoning and a recommendation for what to do instead — which is sometimes a sixty-dollar-a-month subscription platform. It is the third of the three possible answers from the call and it comes up more often than you would expect. A diagnostic that can only end in "yes, hire us" is not a diagnostic.
04 Is there a mandatory subscription or monthly fee?
No. The system runs on your infrastructure with your credentials, and the recurring cost that remains is what it consumes from providers —the model, the cloud— billed directly to your account and not through us. Support is optional and separate.
05 How is payment agreed?
It is settled in the proposal, alongside scope and date, and it depends on the size of the project. What does not change is that the figure in the proposal is the figure invoiced: if additional work appears it is quoted separately and approved before it is done.
Fifteen minutes. A concrete answer.
It gets fixed, it gets built, or it is not worth it. And if the diagnostic does not reach three actionable findings, it is not charged.