AI training for companies built on the tasks your team does every day

Office staff write, sort and look things up all day. This is where AI saves the most manual work, and also where it most often makes a mistake nobody checks. So every exercise ends with a check against the original: for meeting minutes you compare the action points with the notes, and when comparing two versions of a contract or price list you have the differences listed and verify each one. Replies to long emails are built on your own instruction in the company's style, so they do not sound like a robot wrote them.

• Meeting minutes with actions and deadlines → a prompt template to reuse
• A reply to a long email in your tone → a custom instruction with the company's style
• Differences between two versions of a contract or price list → a method for checking output against the original

Sales need to turn call notes into a follow-up email and a CRM entry quickly, marketing need text variants to test. AI handles both, as long as it knows what it must not invent. We practise prompts that stay on brand and work only with the facts you supply. For competitor comparisons from public sources we require a traceable link for every figure and show how to check the numbers before they end up in a proposal to a client. Three manual steps after a meeting become one repeatable routine.

• Follow-up email and CRM summary from call notes → one routine instead of three manual steps
• Competitor comparison from public sources → research with links that can be traced
• Ad or post variants for testing → a prompt that stays on brand and does not invent facts

Leadership is not asking how to write a better email, but where to use AI in the company and under what rules. The half-day programme has at most 8 people, so you can talk openly about specific processes and company data. Together we draft a one-page internal policy: what employees may and may not put into AI tools. We pick three processes for pilots with a measurable goal, so the decision rests on results rather than impressions. And we compare ChatGPT, Copilot and Gemini against what your company runs on and which licences you already pay for.

• Rules on what may go into AI → a one-page internal policy draft
• Choosing where to start → three pilots with a measurable goal
• Choosing between ChatGPT, Copilot and Gemini → a comparison of licences and risks for your situation

HR handles candidates' personal data, so what must stay out of AI matters most here. In the training you write a job advert that matches the real position and save it as a template for future hiring. From the job description you prepare interview questions, including what to look for in the answers. We also cover separately where AI in recruitment runs into GDPR and the AI Act, and which steps therefore stay manual. The aim is for HR to know where AI helps and where a person has to make the decision.

• A job advert that matches the real position → a template for future hiring
• Interview questions from the job description → a question set with what to look for in answers
• Where AI in recruitment meets the rules and the AI Act → an overview of what stays manual and why

Support answers the same questions over and over. In the training we turn those repeated questions into a shared set of answers the team can use and AI can work with: short, self-contained answers instead of one long document. We practise a prompt for an angry customer with firm limits on what the reply must not contain, so the AI does not promise a discount or a date the company will not give. And we show how to turn a week of complaints into a summary for management without assembling it by hand.

• Repeated questions → a shared knowledge base in a form AI can work with
• A reply to an angry customer → a prompt with limits on what the reply must not contain
• A week of complaints → a management summary you do not have to compile by hand

Before the date, the team completes a questionnaire: what they do every day, what slows them down and which tools the company runs on. The exercises are built from the answers. In the workshop, people then work with your emails, spreadsheets and documents, not with internet examples that transfer badly to your work. If the group mixes daily AI users with people who have never opened it, we split the exercises by level, so nobody is bored and nobody gets lost.

• questionnaire completed before the date
• exercises using your emails, spreadsheets and documents
• exercises split by level in mixed groups

The most expensive AI mistake is not a badly written prompt, but an output nobody checked that went out to a client. So every topic ends with a check: comparison with the original, tracing the source, recalculating the figure. We teach people to recognise the typical situations where a model invents things, such as exact figures, quotes, legal references or links. Everyone leaves knowing when to trust an output and when not to, and with a quick way of checking it.

• every exercise ends with an output check
• typical errors: figures, quotes, legal references, links
• a check that takes a minute, not an hour

Employees often paste contracts, personal data or price lists into free AI services because nobody told them not to. In the training we show which data belongs in which service, how to switch off training on your data and where the GDPR line sits. With free versions, data may be used to train the model, so for company data we recommend a business licence. Leadership also gets a draft internal policy, so the rules apply to the whole company and not only to those who attended.

• which data belongs in which AI service
• switching off training on your data
• free vs. business licences from a data point of view
• draft policy in the leadership programme

We do not pick tools by trend. For companies on Microsoft 365 we focus on Copilot, because people already have it in Outlook, Word and Teams. For Google Workspace, Gemini. Otherwise ChatGPT and Claude. What decides it is which licences you already pay for and where your data is allowed to sit. Free versions are enough for the training, but we show what a business licence adds and whether it is worth it for you. That makes it ChatGPT and Copilot training for teams, not a sales pitch for one product.

• Microsoft 365 → Copilot
• Google Workspace → Gemini
• otherwise → ChatGPT and Claude
• chosen by licences and data rules

Workshops and the leadership programme are led by Miroslav Douda, managing director of Webforte. He used to teach, has trained over 9,000 teachers with the PECKA team and now develops software with AI. The company runs its own AI products in production, such as the Dotazio chatbot, which answers only from supplied material and cites its source. When we explain why a model makes things up and how to keep it tied to your documents, that comes from running our own applications, not from an online tutorial.

• over 9,000 teachers trained with the PECKA team
• own AI products in live operation
• small team, you talk directly to the trainer