Questions we hear before the first session

Packages have a price set upfront, in Polish zloty and net of VAT: from PLN 500 for an AI Consultation to PLN 15,000 for AI Strategy. The exception is AI Automations & Agents, which start at PLN 20,000 and are quoted after a scoping workshop. We invoice in PLN; EU businesses with a valid VAT-EU number pay the net amount under the reverse charge mechanism. Current rates are on the pricing page. A delivery quote is fixed and does not grow during the build. Tool and infrastructure costs, meaning licences and cloud, stay with your company and are known before the decision. Packages are ordered and settled separately and money paid for one does not carry over to another.

The first effects show up during the first sessions, because from day one the work happens on the company's real tasks. Delivery is usually measured in weeks: the AI Strategy package ends with a 90-day roadmap, and the first automation goes into use before the rest is built. The timeline depends mostly on access and procedures on the company side; the build itself is usually the shortest stage. An exact schedule is part of the quote before the start.

We do both, because only working code shows whether a recommendation was right. We deliver GenAI chatbots and voicebots, AI agents, document intelligence and process automation. Tools are chosen to fit the problem and what the company already runs, and the Tech stack page shows the full range. Strategy and delivery stay in the same hands, so the plan does not drift away from the execution.

The measure of success is agreed before the project starts. Every use case gets an owner and a success condition: what exactly should change, by how much and by when. Most often we count recovered hours, shorter handling times and answer quality measured on live cases. Delivery starts with whatever pays off fastest, so the first numbers appear while the project is still running.

Usually it works in your favour: the licences are already paid for, whether that is Copilot, ChatGPT or Gemini, and the open question is why so few people use them. The problem rarely sits in the tool itself; more often the company lacks use cases tied to specific processes and specific people. We start from what already runs: a review of current licences and how they are used is part of the initial review, and sometimes the outcome is that no new tools need buying at all.

That is more common than perfect data, and it rarely blocks the start. A large share of GenAI use cases works on what already exists: documents, e-mails and files, with no data warehouse and no IT department. The initial review shows which topics can start right away and which need their sources tidied up first. Where IT is genuinely needed, we work with your vendor or the technical person on your side, and the documentation lets the solution run without a full-time engineer.

Every rollout ends with handover documentation: your team knows how the solution works, how to change it and what to watch. Maintenance on our side is an option we agree on openly during project scoping. Small fixes during the stabilisation period are part of the scope. Solutions are built inside the company's environment, so there is no dependency on our infrastructure and no subscription to us.

An NDA is signed on request before the first substantive conversation. Client data never goes into model training, and we use AI tools on terms that rule it out. Where an engagement covers personal data, a data processing agreement under Article 28 GDPR is available before work starts. Where the architecture allows it, documents, data and the models working on them stay inside the company's environment, within one security perimeter; if a use case requires a service outside that perimeter, we say so before the start and record it in the solution architecture.

Hallucinations can be managed with engineering discipline, and we do that from day one. Answers are grounded in the company's knowledge base (RAG), and quality is measured on a golden set: real questions with expected answers, checked before every release. In higher-stakes processes a human approves the decision, and a bot hands the conversation to an agent when it has no reliable answer. After launch, quality monitoring keeps running, so a regression shows up in the numbers before a client sees it.

We design the architecture so that a model or vendor can be swapped without rebuilding the whole solution. We work with OpenAI, Anthropic, Google and open-source models, and the model is chosen per task: cheaper ones handle simple steps, stronger ones come in where answer quality decides the outcome. Code, prompts and delivery documentation stay with your company, with the right to use them for your own business without time limit, so changing a vendor, including us, never starts from zero. The scope of rights to the results and to our reusable components is set out in the Terms of Engagement attached to every quotation.

We treat working with IT and security as a natural part of the project. We deploy inside your ecosystem (M365, Azure, CRM, ERP, API), so we go through your established procedures: security review, permissions and data access rules. IT requirements enter the plan at the very beginning, because they set the delivery date more often than the build itself. Architecture and access are documented so that a review on your side runs smoothly.

An assessment of GDPR and EU AI Act risks and requirements is part of the AI Strategy package: classification of use cases, how data moves, disclosure duties and the points where a human has to approve a decision. Use cases are designed so the obligations can be demonstrated in the documentation from day one. The assessment is advisory and implementation-oriented; it does not replace legal advice, which our terms state plainly.