Frequently Asked Questions

Anywhere from nothing to a few tens of thousands of dollars, depending on what you roll out. ChatGPT, Gemini and similar tools have free tiers and many companies start there, while premium subscriptions run from around $20 to a couple of hundred dollars a month per tool. With us the Operations Agent costs about $1,900 in setup plus a care plan from $190 a month, an agent for an internal process starts at $3,000, team training at around $1,800, and a website runs from $2,600 up to about $17,500 on a large project. We don't sell fixed packages - you get a quote after a short conversation about what you want to improve. The full cost breakdown, from free tools to agents with integrations, is in our article on AI costs for business.

From a few days to a few weeks, depending on scope. A simple website chatbot takes 1-3 days, a monitoring agent for SEO or competitor pricing 1-2 weeks, and automating a complex process such as email sorting or CRM reports 2-4 weeks. A brochure site takes 1-2 weeks, an extended project with a blog and integrations 4-6 weeks. What usually moves the date is not the code but how fast materials and decisions arrive on your side. Per-service ranges sit on the AI agents and websites pages.

A chatbot talks, an agent acts. A chatbot answers questions from rules or a knowledge base, while an agent performs tasks: it checks calendar availability, sends a summary by email, sorts tickets, prepares a report from your data. With clients we often start with a chatbot and expand it into actions once we see which tasks eat the most time. Use cases and costs are broken down on the AI agents page.

By default no - a consumer account trains on your data even when you pay for Plus or Pro. Anthropic introduced these rules on 28 August 2025, and anyone who made no choice by 8 October hands their conversations over for training. What is safe: business and Enterprise plans with a signed DPA, plus self-hosted deployments, and that is what we configure before launch. We went through the OpenAI, Anthropic, Microsoft and Google policies one by one in our article on where your company data goes.

Yes, with the right configuration. We use tools with European data centres and a signed DPA, we don't send personal data to AI tools without anonymisation, and we add an information clause when we deploy a chatbot. Worth remembering: GDPR and the EU AI Act are two separate matters - compliance with one does not give you compliance with the other.

Yes, if you use AI in your company - but there are fewer obligations than the noise suggests. Since 2 February 2025, Article 4 applies: you have to build AI competence among the people who use it. From 2 August 2026, Article 50 adds transparency - a chatbot must state that it is AI, and any material a viewer could mistake for authentic needs to be marked. A product description, an email or an article written with AI needs no label, and that is the most common myth on the market. The high-risk category rarely applies to a small business. We check this as part of an AI Trust Layer audit: what is regulated, what is missing, and how to close the gap.

Not in the next few years. AI takes over repetitive, time-consuming pieces of work, not whole jobs. Instead of having someone retype orders from email into a system, an agent does it in the background while the employee handles the customer with a real problem. Same with reports and repetitive questions. The result is not fewer people - it is people who stop losing hours on work nobody enjoys.

Yes, and it is now a discipline of its own. More and more often a client doesn't search Google - they ask ChatGPT, Perplexity or Gemini who does a given job. For AI to point to you, a ready answer has to be extractable from your site: clear copy, data organised in schema, FAQ sections, plain prices and facts. It is called AEO, optimisation for AI answers, and we apply it to every site we build. The one you are reading was built exactly to these rules.

No, you need to be able to write in plain language. AI tools for business have an ordinary graphical interface and require no programming, and the whole difficulty sits in phrasing the instruction, not in the technology. We teach that from scratch on our AI training - after a 4-8 hour workshop the team works on its own.

A 4-8 hour workshop, groups of up to 12 people, online or at your office. We work on a real task from your industry, and participants leave with ready prompts, checklists and two weeks of email support. The programme and sample exercises are on the AI training page.

In Poland, yes. The National Training Fund (KFS) covers up to 90% of the training cost for companies with up to 9 employees, including sole traders, and up to 70% for larger ones. The company files the application, and it has to be filed before the training, not after. How to go through it step by step, and what to watch out for, is in our article on AI training for business.

Yes. Before we build a site you get a design to approve, and for an AI agent a prototype running on your test data. You pay for the full deployment only once you can see how it works in practice.

Small and medium-sized enterprises - from sole traders up to teams of 249 people, which is the full SME range. We work best in services, e-commerce, tourism, real estate and consulting. A simple test: if someone on your team regularly does repetitive digital work, there is something to automate.

After a site goes live: content updates, small changes and performance monitoring for 30 days. After an AI agent: monitoring, rule fixes and two weeks of technical support. Beyond that there is a care plan, from about $150 a month for a site and from $190 for an agent.

You have to assume it will, because models make things up by nature and it cannot be switched off. So we limit an agent to verified sources, keep a human on decisions with financial or legal consequences, and log conversations so you can trace where an answer came from. The mechanism, and how to keep it in check, is in our article on why AI makes things up.

For plenty of tasks it is, and we say so openly. ChatGPT is good when someone sits down and writes a prompt. An agent makes sense when the task has to run by itself, at three in the morning, with nobody at the keyboard, and when it has to reach into your data or systems. The second threshold is security: a consumer account trains by default on whatever you put into it. Which tool for what, we compared in Claude vs ChatGPT for business.

Four steps. A free conversation about what should improve. A quote with the scope written out, so you know what you are paying for. Build, with a checkpoint halfway: a design before we code the site, or an agent prototype on test data. Then launch and care. We work as a team of two, so you talk to the people who actually build the project, not to a ticket queue.