AI consulting and process automation for business

It depends on how many systems have to be integrated and how well defined the process is to begin with. In the first call we go through how you work and give you an order of magnitude right there, before we prepare the quote.

On top of the build cost sits the model usage cost, which varies with your volumes. We work it out along with the quote, because on high-volume processes it can change the technical choices.

We have worked in fintech, healthcare, logistics, automotive and betting. The sector matters less than the process: data extraction from invoices works the same way in an accounting firm and in a haulage company.

What changes are the constraints. Companies handling health or banking data have rules about where the data ends up and what can pass through an external model. If you are in one of those sectors we cover it in the first call, because it changes the architecture.

Half an hour in which you tell us how the process that costs you the most time works today: who runs it, how many times a week, where it gets stuck. We ask questions.

At the end we tell you whether it is worth automating and give you an order of magnitude for cost and timing. The call is free, and if the answer is that you are better off leaving it as it is, we tell you that too.

Where there is an API we connect directly: CRMs like HubSpot or Salesforce, ERPs, databases, mail systems.

When the system is old and has no API, we work with whatever it does expose: a scheduled export, a folder where it drops files, the underlying database if we can read it. It is less elegant and it works. This is one of the things we look at in the first call, because it moves the quote more than anything else.

In the first weeks the system runs on real data, and that is when the cases you cannot see during analysis show up: the invoice filled in oddly, the request written in a form the system had never met. Post-launch support is included in the price and covers exactly these.

The code and the workflows stay yours. You can keep them in house, or call us when a change is needed. If you then decide to automate a second process, it starts again from a first call.

Databases, automations and hosting for the systems we build are configured in European regions. Language models are the part to watch: if your project needs the data to stay in Europe there too, we settle it during analysis and pick the provider accordingly.

An automation touches more data than the model needs. During analysis we decide which fields reach the model and which stay in your systems, and we put in writing what data enters, where it is kept and for how long. You remain the data controller; we process on your behalf.

The European AI regulation applies from 2 August 2026. Article 50 requires that a person knows when they are talking to an automated system: the systems we build state it explicitly.

Yes. Repositories, workflows and configurations stay yours. You can keep them in-house with your team, or carry on with us.

It is also why the technical choice depends on who will maintain it afterwards: if you have a developer in-house, code; if you do not, a workflow your team can open and change.

Every automated action leaves a log: what came in, what the system decided, what it did. When you need to know why a case was handled a certain way, the answer is there.

On work that changes every couple of months, that happens three times a year, or that carries legal responsibility: building and maintaining it costs more than the time it saves. Automation pays on repetitive processes with stable rules and high volumes — invoices, routing requests, CRM updates, recurring reports.

Then there is the cost per operation: every model call is paid for. At a hundred operations a month you do not notice it; at fifty thousand it becomes a line in the budget. We work it out before the quote, and now and then the answer is that part of the process is better left out of AI.