In the past year, in every conversation with the owner of a company in Croatia — whether it's a three-person micro-business or an organization with 200 employees — the AI question comes up sooner or later. Usually in the form: „I see everyone talks about AI, but honestly — where could it help us?"
It's a fair question, and the answer is much less glamorous than what you read in tech publications. Most AI content is written for an American or Western European context — startups with 500 employees, corporations with budgets in the millions of euros, or technology enthusiasts who experiment for the sake of it.
Croatian business reality is different. From micro-businesses (founders working alone or in small teams) through small and mid-sized companies to larger corporations — these are all owner-led or tightly-run organizations with concrete operational needs, not research labs. Their needs are different, their pain points are different, their budget is different.
This piece is an attempt to answer honestly: where AI genuinely belongs in Croatian companies of every size, and where it's just an expensive experiment.
Three areas where AI works
1. Sorting and classifying documents
The biggest „boring win" I see in companies is sorting incoming documentation — emails, invoices, orders, support tickets. A three-person micro-business and an organization with 200 employees both receive documents that someone has to sort every day. The volume differs — but the math is the same: manual sorting steals time that should go elsewhere.
AI tools solve this extremely well. The model reads the content, recognizes the document type, classifies by priority, and routes to the responsible person or system. Accuracy is typically above 95% on tasks we've defined well.
Concrete example: a Slavonian distribution company with around 80 employees was receiving 200 to 400 customer email inquiries a day — some about order status, some about complaints, some asking for new quotes. Three people spent half their day just sorting and forwarding. After we put an AI classifier in place, those three people now work on more demanding tasks that require human context.
Why does this work? Because the problem is well defined, repetitive, and has clear structure. AI is better than people at tasks like this — not because it's smarter, but because it doesn't tire, doesn't mind routine, doesn't have to think about other things at the same time.
2. Generating structured content
The second concrete win is generating content that has clear structure — product descriptions, document summaries, reports from data, answers to common questions.
If a company sells 500 products and each needs a description for the online store, human work takes weeks. AI does it much faster, with an editor going through and fine-tuning. The cost difference is significant.
Similar with reports — a financial report that needs a written summary every month, a marketing report every quarter, EU project reports. AI takes the data from the table, writes a readable summary, a person approves or corrects.
Important: this only works if you have good data and clear criteria. If the data doesn't exist or is unreliable, AI will invent. Text generation isn't magic — it's formatting existing information into a readable form.
3. Searching and analyzing your own documents
The third win, and maybe the most underestimated, is AI-assisted searching of your own archive. Most companies have tens or hundreds of GB of internal documents — contracts, reports, emails, financial analyses — in which knowledge is hidden that no one uses.
Typical example: a manufacturing company that, in the past year, had 12 meetings with different equipment suppliers. Each meeting was documented in minutes. When the moment for a new selection comes, no one remembers what each supplier promised, on what terms, what their weak points were.
An AI tool connected to your documents can immediately retrieve: „Show me all the commitments suppliers mentioned around short delivery times." In minutes, not days.
This works especially well in sectors where knowledge „sinks" into documents — legal practice, consulting firms, agencies that work with many clients, EU project teams.
Three areas where AI doesn't (yet) work
1. Making important decisions
AI doesn't make decisions. It proposes, analyzes, sorts options — but the final decision always has to remain with a person, especially for decisions that have big impact.
I see companies deploying AI dashboards with recommendations like „we recommend investing in product X because the data is positive." If leadership makes decisions looking at that dashboard without their own analysis, the consequences can be serious. AI doesn't understand the context you haven't explicitly specified — your long-term goals, market nuances, political realities.
Rule: AI enters the decision process as a source of information, not as a decision-maker.
2. Conversation with clients in complex situations
Chatbots for standard questions — opening hours, order status, where the store is — work fine. But the moment a client has a complicated complaint, frustration, or needs to negotiate terms, AI isn't the answer.
I've seen companies that tried full automation of client service, and the consequences were predictable: loss of key clients who got angry at „robotic" responses in the moments when they needed a conversation with a real person.
Rule: AI for routine inquiries, people for anything that requires empathy or negotiation.
3. Creative strategic work
When you're thinking about a new product, a new market, or a fundamental change in how you work — AI won't give you the answer. It can help with research, gathering information, brainstorming. But the strategic insight that connects a market opportunity with your specific context, strengths, and constraints — that remains human work.
The reason is simple: AI is trained on already-existing information. It doesn't see your specific situation, doesn't understand nuance, doesn't make creative leaps.
How to decide where to start
The biggest mistake I see is a company buying an „AI solution" before it has defined a concrete problem. We work the other way: where does your team spend the most time today on routine, repetitive tasks? That's the first candidate for AI.
Three questions we ask in every first conversation:
- What activities does your team do every day or week, identically or almost identically? If the answer exists, an AI solution that makes sense probably exists.
- Which process has a clear input, clear logic, and clear output? Those processes are easiest to automate.
- Where do people retype, copy, or repeat the same work in different systems? A classic place where AI tools bridge silos.
If answers to these three questions exist, you're in the safe zone — concrete application that saves time and delivers measurable value.
If not — it's best to wait. Better to invest in process, organization, or developing people before buying an expensive AI tool that doesn't solve a clearly defined problem.
Conclusion
AI is a tool, not a strategy. In Croatian companies of every size, the biggest wins are usually „boring" — sorting, generating, searching. Where AI does 80% of the work, people do the 20% that requires context and decision.
The worst scenario is following the hype and investing in something that looks impressive but doesn't solve a real problem. The best scenario is to start small, prove the value, and then expand.
If you're considering where AI makes sense in your business, take a look at our AI & Automation service — through a conversation we assess where it can concretely help, and where it doesn't make sense to invest.