Frequently asked questions

Yes, always. All code built for your project is yours from day one. We never use third-party licenses that lock you in.

It depends on scope, but we deliver a first working result in under 4 weeks. Larger projects are split into phases with incremental deliverables.

We offer both. Defined-scope projects get a fixed price. Open or evolving work runs as monthly sprints.

Modern web technology with custom code optimized for top loading speeds and SEO. For AI and automations, we integrate advanced language models and workflows connected to your existing tools. We also operate our own secure infrastructure for hosting. We always pick the best technical solution for each case, not what suits us best.

We host on our own secure infrastructure: hosting, encryption and access under our contract, with GDPR compliance. Tools you already use (CRM, email, etc.) may stay in the cloud; the claim covers what we operate. We issue processor agreements where GDPR applies.

Yes. Every project includes 24/7 monitoring, security updates and technical support. The support level is agreed in the contract.

The order that works is always the same. One: map real processes and find where hours are lost or data is duplicated. Two: prioritize impact versus effort and pick a single use case. Three: prepare the data and integrations that case needs. Four: ship a production version with human oversight and metrics. Five: measure real savings and only then expand. Starting with a tool instead of a process is the most common reason AI projects go nowhere.

You do not need a data lake. You need the material your team already uses: FAQs, catalog, policies, ticket history, quote templates or internal docs. What matters is that it is reachable (Drive, CRM, wiki) and that we know what must not leave toward the model. If data is scattered or stale, the diagnostic starts by sorting it, not by training a model.

With process metrics, not vanity metrics. Hours saved on a concrete task, customer response time, qualified leads, avoided errors or recovered missed calls. We set a baseline before shipping and review it at 4–8 weeks. If there is no number the business already watches, the use case is not ready yet.

Clear candidates combine high volume, relatively stable rules and a lot of manual work: copying data between tools, classifying email and orders, extracting invoice data, repetitive quotes, recurring reports, customer or employee FAQs, and search in internal docs. If someone does the same task more than twenty times a month and can explain the criteria, there is an automation case.

It depends on whether the process is what differentiates you. For standard work —accounting, payroll, email— a market product is usually better: cheaper and maintained by others. Custom software wins when the process is your edge, when off-the-shelf covers 60% and leaves out the 40% that matters, or when accumulated license cost exceeds building the tool once.

Four criteria separate a reliable partner. First: an in-house team, not subcontracted, so you talk to the people who build. Second: ownership of code, data and infrastructure, with no lock-in. Third: milestone work with visible deliveries every two or three weeks. Fourth: verifiable cases or demos in your type of project. Ask who writes the code and who runs it after launch.

A template gives you generic design, third-party plugins and a CMS that becomes a burden. A custom project includes information architecture, original design, code you own, performance and technical SEO from day one, integration with tools you already use, and a panel so your team can edit without depending on us. The upfront cost is higher; years of patching a template is usually worse.

With three decisions before writing code. First, where data lives: European infrastructure by default, and your own cloud or on-premise if the sector requires it. Second, what goes to the model: minimize and anonymize what is not needed, with providers that do not reuse your data to train general models. Third, who can see what: source-system permissions are inherited and every query is logged. We add human oversight on decisions that affect people.

A diagnostic delivers conclusions in days, not months. A chatbot connected to your content is usually live in under 2 weeks. A concrete automation, in 3–6 weeks. A document assistant with RAG, in 3–8 weeks. A custom ops platform has a useful first version in 2–3 months. We work in incremental deliveries so return starts before the project is fully done.

Scope usually gets sharper once you see the first deliveries. That is why we prefer sprints with a visible backlog: new work comes in, less urgent work goes out, and the monthly budget does not explode. On a fixed price, a scope change is documented as an add-on with explicit cost and timeline. We never freeze a project for six months to “surprise” you at the end.