Training + automation · HU / SK / EN
Practical AI for your company.
We do not just teach it, we build it.
Only a fraction of companies use AI. Even fewer use it well. They will be the winners of the next decade. Be one of them: your first working automation is ready in 2 to 4 weeks. Fixed prices, in English, Hungarian and Slovak, often with state funding.
Not sure where you stand? Take the 2-minute AI test →
- Fixed, published prices
- Training in Hungarian, Slovak and English
- For companies of 10–250 people
- HU / SK / ENthe languages we train in
- 5+ hourssaved per person each week
- 2–4 weeksto your first working automation
- Fixed pricepublished price list, no haggling
What we do
Three steps, one supplier
Training companies do not implement, automation agencies do not train. We hand over both: the knowledge and the working solution.
Corporate AI training
Hands-on training on your own processes: ChatGPT, Claude and Copilot in everyday work. From a half-day workshop to the 6-week AI Champion programme. It also covers your AI literacy obligation under the EU AI Act.
AI automation
Quote request processing, e-mail triage, invoice digitisation, a chatbot trained on your own data. We deliver the first working automation in 2–4 weeks, including training for your team.
AI audit and consulting
We map your processes and show you where AI saves the most time and money, with a prioritised plan and an overview of the training subsidies available to you.
We build onn8nMakeOpenAIAnthropic ClaudeGoogle GeminiMicrosoft 365 Copilot
Use-case catalogue
Where AI delivers the most in a company
This is our internal catalogue: nine areas that really work in a company of 10–250 people. For every process we state the complexity and the typical saving, so you can decide before you order anything.
Customer service
The best area to start with: every company has it, everyone hates it, and the results are measurable straight away.
| Process | What it does | Complexity | Typical saving |
|---|---|---|---|
| Email triage | Categorisation (enquiry / complaint / invoice / spam), prioritisation, routing to the right colleague, summarising long threads. | 2 | 5–10 hours/month/person |
| Reply drafts | The AI writes the draft from earlier correspondence and your knowledge base; your colleague only checks it and sends it. | 2–3 | 40–60% faster |
| Website FAQ chatbot | An assistant built on your company documents that handles 60–80% of the typical questions. It escalates to a person when it cannot answer. | 3 | 30–50% of enquiries |
| Internal knowledge assistant | The same thing facing inwards: “where is the holiday request form”, “what are the delivery terms for customer X”. | 3–4 | 2–4 hours/month/person |
| Call and meeting transcripts | Recording, transcript, extraction of decisions and tasks, automatic entry into the CRM or the task tracker. | 1–2 | 15–30 minutes/meeting |
| Multilingual customer handling | A message in a foreign language → a reply in the customer’s language, using your company terminology. | 2 | SK/HU/DE/EN markets, no limits |
Sales and CRM
The best pilot area: the result shows up on the revenue side, not only in cost savings.
| Process | What it does | Complexity | Typical saving |
|---|---|---|---|
| Enquiry processing | Reading the incoming request, structuring the data, writing it into the CRM, drafting a quote from a pricing template. | 3 | 10–20 minutes/quote |
| Lead research and qualification | Company data from public sources, scoring against your ICP, a “call these five today” list. | 2–3 | 3–5 hours/week/sales rep |
| Personalised outreach | Cold emails and LinkedIn messages that are really about that company, not a template. | 2 | 3–5× more touchpoints |
| No more CRM data entry | Automatic logging of emails, calls and meetings, filling in missing fields, flagging stale deals. | 2–3 | 4–6 hours/week/sales rep |
| Follow-up automation | No reply for X days → a reminder built on the context, sent after your approval. | 2 | 10–20% of lost leads |
| Win-loss analysis | Why you won or lost: patterns across the last 100 quotes. | 3 | pricing decisions |
Finance and admin
In finance, AI prepares the work, it does not decide. Every item goes through human approval. Say this up front, it builds trust.
| Process | What it does | Complexity | Typical saving |
|---|---|---|---|
| Invoice OCR | Reads incoming invoices (PDF, photo, attachment), checks the data and uploads it to your accounting software. | 3 | 15–25 hours/month at 300 invoices |
| Expense and receipt handling | Processes photographed receipts, sorts them into categories and flags anything missing. | 2 | the month-end scramble disappears |
| Receivables management | Tracks overdue invoices, sends reminders tuned to each customer’s payment history, with an escalation ladder. | 2–3 | payment times 5–15 days shorter |
| Matching bank transactions | Matches statements against invoices and lists the differences. | 3 | 4–8 hours/month |
| Month-end close in plain language | Turns the numbers into a clear explanation: what changed, why, and what to watch. | 2–3 | visibility for management |
| Cash flow forecast | A 30/60/90 day outlook based on receivables, payables and seasonality. | 3–4 | no liquidity surprises |
Documents and knowledge
This category makes the strongest impression in training: you watch it work on your own document.
| Process | What it does | Complexity | Typical saving |
|---|---|---|---|
| Contract review | Risky clauses, payment terms, termination, liability; compared against your company template. | 3 | 1–3 hours/contract |
| Document generation | Fills in contracts, quotes, minutes and certificates from your data and templates. | 2 | 20–40 minutes/document |
| Company knowledge base (RAG) | Makes every internal document searchable and answerable, with sources cited. | 4 | onboarding time cut in half |
| Extracting technical documentation | Supplier datasheets, specifications, catalogues → structured database. | 3–4 | weeks of manual work |
| Minutes and reports | Turns raw notes into a formal document in your company format. | 1–2 | 30–60 minutes each |
| Version comparison | What actually changed between two versions, not character by character. | 2–3 | 70% of the review time |
Marketing and content
A warning: this is where clients most often say “we already do that with ChatGPT”. The value is not text generation, it is the system: a fixed brand voice, a defined process, consistent quality.
| Process | What it does | Complexity | Typical saving |
|---|---|---|---|
| Content production in series | Blog, LinkedIn, newsletter, product copy (in your company voice, from a brief). | 1–2 | 60–70% of the writing time |
| Bulk product descriptions | Descriptions for hundreds or thousands of webshop products, written from attributes, with SEO. | 2–3 | weeks → days |
| Social calendar and posts | Monthly plan, post copy, image generation or template-based design. | 2 | 1–2 days/month |
| Translation and localisation | SK ↔ HU ↔ EN ↔ DE with your own glossary, well above machine-translation quality. | 1–2 | 70–80% of translation costs |
| Campaign reporting | Collects advertising and analytics data, analyses it, suggests next steps, weekly report. | 3 | 3–6 hours/week |
| Competitor monitoring | Monitors competitor prices, ads and content, monthly summary. | 3 | no more blind spots in your market |
HR and recruitment
Legal note: under the EU AI Act, CV screening counts as a high-risk system (Annex III). It is not banned, but it comes with documentation and oversight duties: the date of application has been pushed back to 2 December 2027.
| Process | What it does | Complexity | Typical saving |
|---|---|---|---|
| Job ad and interview plan | Turns a role description into a job ad, a structured interview plan and a scoring framework. | 1–2 | 3–5 hours/position |
| CV pre-screening | Structures applicants, matches them against the requirements, builds summary cards. | 3 | 5–15 hours/position |
| Onboarding assistant | New joiners can ask about internal rules and processes 24/7. | 3–4 | 20–30% of the ramp-up time |
| Internal training material | Turns process descriptions into an e-learning module, a quiz and a quick reference guide. | 2 | weeks of course development |
| Working time and leave | Processes requests, flags clashes, writes them into the calendar. | 2–3 | 3–6 hours/month |
| Employee survey analysis | Turns free-text answers into themes, sentiment and concrete suggestions. | 2–3 | decisions based on data |
Operations and manufacturing
This is where the most money and the most risk sit. Before you promise anything: does the ERP have an API? can it export? who maintains it? Without an API the project is one level harder.
| Process | What it does | Complexity | Typical saving |
|---|---|---|---|
| Order processing | Reads orders that arrive by email or as a PDF, writes them into the ERP, sends the confirmation. | 3–4 | 10–20 minutes/order |
| Stock and purchasing alerts | Forecasts from consumption data, suggests purchase orders automatically. | 3–4 | fewer stockouts and less overstock |
| Comparing supplier quotes | Puts several incoming quotes into one table and shows the differences. | 2–3 | 2–4 hours/purchase |
| Project status report | Builds a weekly status automatically from the task tracker, emails and meetings. | 3 | 2–4 hours/week/project manager |
| Maintenance log | Structures free-text fault reports, identifies recurring causes. | 3 | less downtime |
| Photo-based quality control | Image recognition for defects, automatic documentation. | 4–5 | depends on the case, high |
Management reporting and data
An owner who uses the output personally becomes the sponsor of the project. That is why the management summary is such a good way in.
| Process | What it does | Complexity | Typical saving |
|---|---|---|---|
| Weekly/monthly management summary | Data from several systems on one page: cash, sales, pipeline, risks. | 3 | 4–8 hours/month + better decisions |
| A question instead of a pivot table | “Which product delivered the highest contribution margin in Q2?” (a question, not an Excel sheet). | 3–4 | no more dependence on an analyst |
| Replacing Excel reports | Automates repetitive manual spreadsheet building. | 2–3 | 5–20 hours/month |
| Margin and pricing analysis | Unit economics per product, price-change scenarios. | 3 | direct profit impact |
| Anomaly monitoring | Automatic alerts for unusual spending, falling revenue and lost customers. | 3–4 | early warning |
IT and internal tools
Operation, monitoring and version control are not extras: they are what the monthly fee pays for, and your only protection against a workflow that breaks silently.
| Process | What it does | Complexity | Typical saving |
|---|---|---|---|
| Rolling out an AI coding assistant | Copilot or Claude Code for your development team, with clear rules. | 2 | 20–40% of development time |
| Internal mini-apps | A calculator, a form or a dashboard without a developer. | 2–3 | hours instead of days |
| Data migration and cleaning | Normalising messy databases and filtering out duplicates. | 3 | weeks of manual work |
| IT helpdesk automation | Triage of typical requests (password, access, faults) with a suggested fix. | 3 | 30–40% of all tickets |
| Generating documentation | Up-to-date documentation from code, processes and systems. | 2–3 | the perennial gap, closed |
ScaleComplexity 1 = switching on a ready-made tool (1–4 hours) · 2 = connecting 2–3 systems (1–3 days) · 3 = branches, exceptions, approvals (1–3 weeks) · 4 = closed system, knowledge base, permissions (3–8 weeks) · 5 = custom development, regulated area (2+ months).
The full catalogue (52 processes) as a PDF
Leave your email and download it right away. Plus one practical article a week. No spam, one-click unsubscribe.
Recommended order
Ten processes worth starting with
The order is deliberate: the processes with the best return-to-complexity ratio at a company of 10–50 people come first. The first five fit into one pilot, the rest belong to the growth package.
- 01Email triage and draft replieseveryone has email, the effect is immediatePilot automation
- 02Processing quote requestsrevenue side, a real pain pointPilot automation
- 03Invoice OCRhard savings you can countPilot automation
- 04Meeting transcripts and task assignmentready in a day, strong impactPilot automation
- 05Document generation from templatesthe most striking demoPilot automation
- 06Putting an end to CRM data entryyour sales people will love itGrowth package
- 07Overdue payment remindersdirect cash flow effectGrowth package
- 08Content production turned into a systemthe way in at marketing-led companiesGrowth package
- 09FAQ chatbot on your websitea highly visible referenceGrowth package
- 10Weekly management summarythe owner uses it personally, so the owner becomes the sponsorGrowth package
The calculation we put into every quote
Work out your return before you contact us
If the investment does not pay for itself within 12 months, we do not take the project on. This is exactly the calculation we go through with you on the consultation. Move the sliders to match your situation.
The calculation deliberately leaves out soft benefits (a faster reply → a higher win rate on quotes). We will mention them, but we do not build the business case on them.
Free 30-minute consultation →Self-assessment · 2 minutes
Where does your company stand with AI?
Ten questions, no email required, instant result: your level and the next steps. Answer for the company as a whole, not for yourself.
01How many people at your company regularly work with AI tools?
02What kind of AI accounts do you use?
03Do you have an internal AI usage policy?
04Do colleagues know which data must never go into AI tools?
05Has the team had any AI training?
06Is any process at your company automated with AI?
07Do you share prompts and templates across the company?
08Do you measure the time or money AI saves?
09Do you use AI in customer communication?
10Does anyone own the AI topic?
Level 1 of 4Starting out
AI use at your company is incidental at best. The good news: this is the phase where first steps pay back fastest, and it takes little to get ahead of most companies around you.
- Half-day AI Fundamentals workshop for the whole team (€990)
- Pick one process from our top 10 for a first pilot
- Free 30-minute consultation on where to start at your company
Level 2 of 4Experimenting
Individuals already use AI, but without a system. This is the riskiest phase: shadow AI without rules, and know-how sitting in two or three heads.
- Full-day AI in Practice training on your own processes (€1,490)
- EU AI Act package: internal policy + AI literacy (€1,290)
- AI audit with a guarantee: no 10 hours of weekly savings found, no invoice
Level 3 of 4Systematic
The basics are in place: accounts, rules, first experience. The next jump comes not from training but from automation with measured payback, extended to more departments.
- AI process audit (€790, with the guarantee) and a prioritised plan
- Pilot automation from €1,900, live in 2–4 weeks
- A shared prompt library and measured savings
Level 4 of 4Front-runner
You are in the small minority of companies that manage AI systematically. Now scaling pays off: connected workflows, internal champions and operations that never fail silently.
- Growth package: 3–5 connected workflows (€5,000 to €12,000)
- AI Champion programme for internal owners (€3,900)
- Operations and development from €190/month
Price list
Fixed prices. No “price on request”.
Prices exclude VAT and apply to closed groups of up to 12 people. The audit fee counts towards the pilot, and the pilot fee towards the growth package. The next step is always clear.
Entry
AI consultation
30 minutes online: where you are, what hurts, what is worth doing. No obligation.
AI process audit
2–3 hours of interviews and a written report: a prioritised list of processes and an overview of available grant contributions.
Training
AI basics workshop
Half a day of hands-on work: ChatGPT, Claude, Copilot, prompting, data security (on your own examples).
AI in practice
A full day on your processes: exercises by role, worked company use cases, an action plan.
Role-based training
Targeted workflows for marketing, HR, finance, customer service or sales.
EU AI Act package
Half-day AI literacy training (Art. 4) plus an internal AI usage policy tailored to you.
AI Champion programme
6 × 3 hours: we train your internal AI leads on a real company project, ending with a presentation to management.
Automation
Pilot automation
One specific process, delivered turnkey (e-mail sorting, enquiry handling or invoice OCR), training included.
Growth package
3–5 connected workflows, integration with your CRM or ERP, documentation.
AI assistant / chatbot
A customer-facing or internal assistant built on your knowledge base, with escalation to a human.
Running and development
Monitoring, fixes, monthly optimisation and one smaller extension per month.
Packages
Starter package
Audit + a full-day training + one pilot automation. The fastest route from zero to a first measured result.
Transformation package
Audit + 2 days of training + AI policy + 3 workflows + 3 months of running support.
In Slovakia, the Zručnosti pre trh práce (Skills for the Labour Market) programme covers up to 1 500 € per employee, and further calls for employers are open alongside it. We check your eligibility in advance free of charge as part of the quote.
Limits
What we do not automate
We bring this list to every meeting. A consultant also tells you what is not worth doing. A salesperson does not.
- ✕Final financial and legal decisions
AI prepares, a human approves. Always.
- ✕Hiring, dismissal, performance reviews
Beyond ethics, this is a legal risk too: the AI Act and Article 22 of the GDPR.
- ✕Sensitive customer conflicts
Escalating a complaint is a human job, not an assistant’s.
- ✕Rare, changeable processes
Twice a month does not pay back three weeks of development.
- ✕Anything the company itself does not understand
It has to be written down first. A bad process, automated, is simply bad faster.
- ✕Decisions critical to health, safety or compliance
Never without human review.
Regulation
EU AI Act: what applies today, and what is coming
Not a bogeyman, a calendar. Some obligations are already running, and one of them affects everyone with a chatbot on their website.
- Article 4: AI literacy
Every employer using AI has to take measures to develop the AI literacy of its staff, and has to be able to prove it. The wording was relaxed in July 2026, but the requirement for documented measures stayed.
- Article 50: transparency
If you run a chatbot or publish AI-generated content, you have to say so. This already applies today, and most companies are unaware of it.
- High-risk systems
Recruitment and HR decision support (Annex III). The Digital Omnibus of July 2026 pushed this back from 2 August 2026: 16 months to prepare.
- Article 22 of the GDPR and the SCHUFA ruling
Producing a score on its own already counts as automated decision-making. This applies today, independently of the AI Act.
Why PraxAI
Three things you will not get elsewhere
Practice, not theory
Every training session runs on your real processes and your own data: you leave with finished workflows, not with notes.
We teach and we build
Training companies do not implement, agencies do not train. We hand over both: the knowledge and the working solution.
Hungarian, Slovak, English
We train in the language your team actually works in, in southern Slovakia and in international teams alike.
What people usually ask
Feel free to push back
“This will take my colleagues‘ jobs.”
We do not replace people, we replace tasks. We start with the work nobody enjoys doing. That is why your team is usually your biggest ally, not the brake.
“AI hallucinates, I cannot trust it.”
We agree, which is why we do not automate decisions, only preparation. Every critical step has human approval, and we measure accuracy on a test set.
“Everything is unique here.”
The process is unique, the pattern is not. In a two-hour audit we show you the three things that are the same at every company, and will be the same at yours.
“Our data cannot leave the building.”
There are several levels: EU data processing, business (non-training) accounts, or a solution running on your own server. First we clarify what the real requirement is.
“It is expensive.”
Let us do the maths together. If it does not pay back within 12 months, we do not do it. We dare to say that out loud: it filters out the bad projects.
“Next year, maybe.”
Training grant calls have deadlines. And your competitor is getting the same offer today.
“We already use ChatGPT.”
Great, then half the work is done. The question is whether there is a system behind it: a policy, shared templates, measurement. Without that, the knowledge stays stuck with 2-3 people.
“What language do you run the training in?”
Hungarian, Slovak and English, depending on which one your team works in. Other languages by separate arrangement.
Next step
Stay current with AI. Let’s start.
30 minutes, online, free. You tell us how you work. We tell you where AI saves you the most. If it does not make sense, we will say that too.
- We reply within 24 hours
- No obligation and no sales pressure
- After the call you get a short written proposal
Record up to 5 minutes (a phone is fine) of how you run the process today and send it to info@praxai.eu. Within 48 hours (working days) you get a video reply with a savings estimate and the build complexity. Free, no strings attached.
Prefer email? info@praxai.eu