ClaudeLab

Implementation

We build the system. We do not install a finished one.

Most providers set up a tool that somebody else built AI into. We develop the routine itself - against your systems, your formats and your rules. That is the difference between a licence and a solution.

from €6,900 · fixed price after the audit

Why off-the-shelf tools stop at the same place

The market is full of tools with AI built in: the CRM that proposes quotations, the accounting package that reads documents, the mailbox that sorts. They are useful, and where one fits you should take it - we say so in the audit even when it argues against us.

They all stop at the same place, though: at the edge of their own data space. The tool knows what is inside it. It does not know the costing in the ERP, the special rate this customer has had for two years, or the drawing on the network drive. But that is exactly where the value arises - and exactly where the work sits.

A system built for you does not have that edge. It reads what has to be read, writes back to where the case actually lives, and follows the rule that applies in your business - not the one a product manager anticipated for the average customer.

What we develop

Agents
Routines set off by an event, pursuing a goal across many steps, coming back when a decision is due.
Document processing
Reading whatever arrives - PDF, scan, drawing, spreadsheet, photo. Including scanned askew and annotated by hand.
Interfaces
Where a system has no way out, we build one. That is the most demanding part and it is our trade.
User interfaces
Only where one is needed. Usually the best interface is the system your people already work in.
Evidence and logging
Every output with its source, every run traceable. Without that a system cannot be verified, and what cannot be verified must not run independently.

Which model is used is decided by the use case - and by the question of where the data may be processed. We are neither a partner nor a reseller of any vendor.

The difference

Custom does not mean from scratch.

Custom-built sounds like twice the effort, and it would be if every project began at zero. It does not. The building blocks - reading documents, connecting to systems, logging, modelling releases - are the same; what differs is your routine and your system landscape.

Which is why the first process is the most expensive and every further one in the same business is cheaper: not as a discount tier, but because the access and the data work are already done.

How a build runs

  1. 01

    Audit

    Record the process, test data access, estimate effort. The result is a fixed price.

  2. 02

    Build

    Connection, routine, checks. You see intermediate states, not just the result.

  3. 03

    Test cases

    Against real past cases from your business, not invented ones.

  4. 04

    Acceptance

    Against criteria agreed and verifiable in advance - after that it runs productively.

Before that

Does your routine carry it at all?

Five questions, two minutes, result immediately: repetition, rule-governance, error cost, volume, reversibility. If one of them fails, you can save yourself the conversation.

To the self-check

Frequently asked

What does custom AI development cost?

One automated process starts at €6,900, each further one in the same project at €4,900. Before that comes the audit at €2,500, credited in full. The firm fixed price follows the audit - before it, any number would be a guess.

Why not simply introduce an off-the-shelf AI tool?

Where one fits, that is exactly what you should do, and we will tell you so in the audit. Off-the-shelf tools stop at the edge of their own data space: they do not know your ERP, your pricing logic or the records on your network drive. As soon as a routine crosses that edge, a subscription runs out of territory.

Do we own what you develop?

Yes. You get the rights of use to what was built for you, and the source code. What we reuse are our own building blocks - not your logic.

Which models do you work with?

That is decided by the use case and by where the data may be processed. We are neither a partner nor a reseller of any model vendor, and where EU processing is required we build on services within the EU or on models we operate ourselves.

How long does development take?

Two to eight weeks per process, depending on the number of systems and formats involved. Building it is usually the shorter half - the time goes into access, coordination and test cases.

Tell us where your time goes.

In the audit we record the process, work out what is worth automating, and you get a plan with fixed prices - credited against the build if you go ahead.