Frequently asked questions

We build websites and web applications together with the AI systems behind them: chatbots and autonomous agents trained on your own content. Interface, AI engine and guardrails are designed as one system, so you do not end up with a site and a bolted on chat plugin, but with something where both halves understand your business.

Yes. Many projects start with the site alone: a modern company website, a landing page or a rebuild of an existing site for performance and SEO. The AI layer can come later, because we design the architecture so that adding an agent does not mean rewriting the site from scratch.

Yes, as long as we get access to the code and the hosting. We start with an audit of the stack, performance, security and technical debt, then tell you plainly what is worth keeping and what is cheaper to rebuild. The audit ends with a list of recommendations you can use even without us.

Polish and English as standard, and we build sites as multilingual from day one, with correct hreflang tags and separate URLs per language. We also run German language projects for the DACH markets. The AI agent itself replies in whatever language the visitor writes in.

Mostly service and B2B companies, e-commerce and technology teams, but the industry matters less to us than whether the outcome can be measured. If your problem is repetitive enquiry handling, a slow quoting process or a site that does not sell, we can probably help.

In four steps: discovery (goals, audience, the questions the agent must answer), design (interface and motion prototype), build (implementation on a modern stack and training the agent on real content) and launch (go live, Core Web Vitals measurement, conversion tuning). Each step ends with something you can see and comment on.

A small, well defined company site usually takes a few weeks. A larger site with integrations and an AI agent is more often a few months. The biggest factor is not the code but how fast decisions are made and content arrives on your side, which is why we set the schedule only after a scoping conversation.

Realistically a few hours per stage: the kick off workshop, a design review, supplying content and materials, and sign off. We batch decisions rather than pinging you daily. If you have no content ready, we can draft a first version for your approval.

One contact channel and one person accountable for the project, plus a short progress summary at an agreed rhythm. We work on a staging environment that stays available by link throughout, so you see progress continuously instead of at a final presentation.

There is no off the shelf price list, because scope varies enormously: a single landing page and a site with CRM integration and an enquiry handling agent are different projects. After a short scoping conversation we prepare a quote broken down by stage, so you can see exactly what you are paying for.

The quote and the first consultation are free and commit you to nothing. You are welcome to compare it with other offers. Work starts only once an agreement is signed, and its scope, schedule and payment terms are agreed individually before that.

Yes, and we usually recommend it. We split the work into stages that each produce a result and a sign off, so after every one you have a working part of the system and a real decision point. It lowers risk on both sides and lets you start with the part that pays back fastest.

Usually three: the domain, the hosting and any external services the implementation uses, for example usage based access to a language model. We always list these separately in the quote so nothing surprises you after launch.

A normal chatbot follows a fixed decision tree and breaks on the first unusual question. The agents we build understand the question, use your content and connected tools, can complete a task such as collecting details for a quote, and know when to hand the conversation to a human.

On your material: site content, offers, documentation, knowledge base and the history of your most common enquiries. We do not let outside knowledge into answers about your company, so the agent says what you actually offer rather than what the model finds plausible.

The risk always exists, so we constrain it with rules hardcoded into the system: the agent answers from approved content, has clearly bounded topics, and routes anything binding, such as prices or deadlines, to the team. Conversations can be reviewed, and the agent's behaviour tightened from what they show.

We are provider agnostic. We match the model to the task and the budget, working with OpenAI, Anthropic, Google and open models among others. The architecture is designed so that swapping a model is a configuration change rather than a rewrite of the whole implementation.

Usually yes. We connect to systems that expose an API or webhooks, such as a CRM, calendar, mailbox or spreadsheet. If a system is closed, the audit tells us whether there is a way around it, and we say plainly when an integration does not make economic sense.

Conversation content goes to the model provider only to generate an answer. We do not use it to train our own models. The scope of processing, retention periods and the list of providers are described in our privacy policy, and for company deployments they are additionally fixed in the contract.

We design it to be: minimal data collection, defined legal bases and retention periods, data processing agreements with providers, and documented data flows. Final compliance also depends on your internal processes, so we hand over everything your data protection officer needs to assess it.

You do, on the terms set out in the agreement. We hand over the repository, environment access and documentation. We avoid proprietary lock in: the project runs on widely used technology, so any competent team can carry it forward.

By default we choose providers with infrastructure in the European Union. Some services, language model providers in particular, operate outside the EEA, in which case transfers rely on the mechanisms in Chapter V of the GDPR. We always show that list before we start, so you can accept it knowingly.

We monitor behaviour and performance, fix whatever the first weeks of real use surface, and hand over instructions for editing content yourselves. We can also provide ongoing care: updates, backups, agent improvements and small changes within an agreed monthly scope.

Yes. We run projects for teams in Germany, Austria and Switzerland, North America and the Asia Pacific region. We work remotely, fit meetings to your time zone, and keep documentation and communication in the language we agree on.

Yes, gladly, before we discuss specifics if that is what you need. Send us your template or ask for ours. NDA or not, we never show client material in our portfolio without explicit permission.

Write through the contact form or straight to our email address and describe in a few sentences what you want to achieve. We reply within one business day, arrange a short scoping call, and prepare a stage by stage quote from it.