This article is about small and medium-sized companies in Slovakia, and about the EU rules that apply when they put AI to work. Artificial intelligence is talked about everywhere, and in small firms it still rarely does any work. According to an OECD analysis, AI use in Slovakia is below the EU average, especially in small and medium-sized companies. Larger companies adopt it about three times as often as smaller ones. The OECD also estimates that AI could add 0.2 to 0.8 percentage points a year to labour productivity growth in Slovakia over the next decade.
The good news is that you do not need a large project or a team of data scientists to start. You need to pick the right boring, repeating job.
Why small companies lag
The reasons we hear most often are these:
- There is no time. The owner and the staff are buried in daily operations.
- The payoff is unclear. It is hard to decide on an investment you cannot put a number on.
- Worry about data. What is AI allowed to see, where is the data stored, and what does the GDPR say?
- Messy data. Customers in Excel, prices in email, and stock in one employee's head. Even a strong AI system cannot move that.
- Missing skills. Not everyone can tell a tool from a marketing claim.
None of these reasons is a dead end. The order of the first steps is what matters.
A simple rule: start with what bores you
The best candidates for automation are tasks that repeat, follow rules, and nobody enjoys. Start like this: write down what you spent time on for a week, and pick the three things that came up most often. They almost always fall into the three groups below.
Automation 1: an AI assistant for support and enquiries
What it is. An assistant on the website or in the inbox that answers the questions you hear most (opening hours, prices, order status, availability), collects details from a prospect, and hands the conversation to a person when that is needed.
Who it suits. E-shops, service businesses, accommodation, repair shops, anyone who answers the same questions every day.
What you need. Written questions and answers, current price lists and terms, and a clear rule for when the assistant hands the conversation to a person.
What to watch. The customer has to know they are talking to AI, and they must always be able to reach a person. The assistant must not promise anything that is not in the price list. At the start, have a person check its replies.
How to measure it. Questions handled without a person, response time, and customer satisfaction.
Automation 2: RPA for documents and repeated data entry
What it is. A software robot (RPA, robotic process automation) that does what a person would otherwise do by clicking between programs: it picks invoices out of email, copies the data into a system, checks the amounts, and sends alerts.
Who it suits. Companies that process invoices, orders, or forms every day and retype data from one place to another.
Why now. From 1 January 2027, Slovakia introduces mandatory e-invoicing. Structured invoices can be processed automatically, so preparing for them and automating the handling is sensible to do together. More in Slovakia's mandatory e-invoicing from 1 January 2027.
What you need. A written process (who does what, and where the data is stored) and a few real sample documents.
What to watch. A robot is only as good as the process is clear. Simplify the process first, then automate it. Always set an exception check, where the robot stops and calls in a person.
How to measure it. Hours saved per month, and the number of errors before and after.
Automation 3: workflows in Odoo
What it is. Automatic steps in the business system: after approval, an enquiry becomes a quote, a quote becomes an order, an order becomes an invoice. The system sends reminders for unpaid invoices, flags low stock, and nudges tasks.
Who it suits. Companies that already use Odoo, or that are thinking about bringing Excel, email, and assorted tools into one system.
What you need. Master data that is set up properly (customers, products, prices) and a willingness to align how people work. Odoo does a lot out of the box, and what it does not do can be added with custom modules.
What to watch. Do not carry the old mess into the new system. A smaller scope, expanded later, works better.
How to measure it. Time from enquiry to invoice, forgotten reminders, and the number of manual steps.
How to tell whether it pays off
A simple calculation is enough:
(hours per month this process takes today) × (hourly labour cost, including contributions) = monthly cost of the process
Compare that with the price of the solution and a year of upkeep. If the investment pays back in one to two years, it is worth doing. If not, pick another process. Use real numbers from your company, not estimates from the internet.
What about regulation
The European AI Act is being phased in, and some deadlines have recently moved. For companies that deploy an ordinary chatbot or document automation, the main duty is transparency: tell people they are talking to AI. Check the current rules before you go live, and if you work with personal data, keep the GDPR in mind as well.
A 30-day plan
| Week | What to do |
|---|---|
| 1 | Map repeating work and pick one process |
| 2 | Describe the process, prepare the data, and define how you will measure success |
| 3 | Run a pilot on a small sample |
| 4 | Review, fix, and decide whether to expand |
One process, one month, clear numbers. Only then the next one.
How we can help
We build AI agents and chatbots, deploy RPA, adjust Odoo, and connect business systems. If you do not know which process in the company is the best candidate, we will look at it with you. Get in touch.