Small and medium-sized businesses are adopting AI faster than they are building digital foundations
In many SMEs, AI appears in the company not because someone planned it, but because someone is already using it. Marketing generates content, sales automates messages, administration uploads documents to AI "to make it faster." This is not chaos - it's a natural reaction to readily available tools. The problem starts later, when the company tries to understand what has actually been implemented and whether anything has changed as a result.
According to the KPMG Digital Transformation Monitor 2024 report, Polish companies – despite good digital infrastructure – still clearly lag behind in the use of advanced technologies in business, including AI. The key barrier is not technology, but the immaturity of processes and lack of competence, particularly evident in the SME sector.
And this explains why, in practice, AI in SMEs often operates "alongside" the company, not at its core. Tools are used, but they are not part of processes. Automation accelerates individual tasks, but does not change the way the organization operates. AI doesn't fail – the moment the company tries to implement it before tidying up the basics does.
You bought AI like a coffee machine – and now you don't know what to do with it
The most common problem is not that companies are adopting artificial intelligence. It's how they're doing it. In many organizations, the decision to use AI is made exactly the same way as the decision about a new invoicing or communication tool: someone tested it, someone liked it, so "we're implementing it." The thing is, in this case, no one stopped to ask what exactly was supposed to change in the way the company operates.
In practice, it looks quite repetitive. One solution goes to the marketing team, another to sales, yet another to administration. Each of them works correctly in its own fragment, but is not part of any larger process. There are no common data, no performance metrics, no one looking at the whole picture. After a few months, the familiar conclusion emerges: "AI is interesting, but it's hard to say if it really helped us."
Reports from 2025 show that technology is not the main barrier for companies, but rather the lack of a coherent approach to digital change – especially in organizations that are growing fast but operationally still rely on intuition and human experience. In such an environment, AI does not organize work. It only speeds up what was already disorganized.
And here lies the crux of the error: starting with the tool instead of the business problem. Without a clear answer as to what needs to be improved and how the company can measure the effect of AI implementation, even the best technology remains just another "experiment we once did."
You implemented AI in three departments. Problem: nobody knows if it's worth it
According to the AI Chamber report (2025), over 75% of companies in the CEE region declare that they use AI tools, but only about 25% do so "at scale" (i.e., as a real element of the company's operations, not just individual tests). Polish SMEs do not deviate from this trend. This is the difference between "we have AI" and "AI is truly working in processes" – and it is precisely in this gap that the lack of a plan arises.
The first reason is trivial: AI entered companies through the back door. It starts with one team, one tool, and one "wow, this works." And then it becomes an "implementation," even though no one established: why, where, how we measure the effect, and who is responsible for it. In practice, it looks like buying a coffee machine and announcing that the company has undergone a "work culture transformation."
The second reason is competitive pressure: customers want faster, cheaper, more "here and now." AI seems ideal because it delivers immediate results in small things (e.g., responses, texts, summaries, automation of simple tasks). The problem is that these quick wins rarely come together as a whole if the company doesn't have at least minimal
And the third reason – the most human one – is the lack of an "owner" for the topic. In smaller organizations, someone has to connect technology with operations (and not just "test tools"). Without this, AI will expand haphazardly: here customer service, there documents, elsewhere sales… and in the end, we still return to the classic: "Nice, but we don't know if it really helps us."
Process automation outpaces the understanding of its purpose
Companies most often resort to process automation where something can be "relieved" most quickly, and not where the process actually affects the business outcome. Available AI solutions and the growing number of AI tools for SMEs favor this approach, as they allow individual activities to be improved without interfering with the entire process. The problem is that AI then only speeds up the execution of tasks that were already poorly designed. The effect is illusory: shorter operational time, but no real improvement in quality or predictability. Automation begins to operate at the level of "activities," not "processes" - and this is where most of the value disappears.
How to start AI implementation in a small and medium-sized business
Start with one problem that currently costs the company real money or time. Not "we want AI," but: "the customer waits too long for a response," "people are transcribing data between systems," "we have errors in documents." If this problem cannot be described in one sentence and simply quantified (minutes, zlotys, number of errors), then AI has nothing to improve – it will only look good in a demo. The simplest selection rule: choose a problem that is frequent, repetitive, and measurable (then "thanks to AI" you can prove the effect, not just feel it). And importantly: don't choose the most difficult topic in the company to start – choose one that can be improved in 2-4 weeks to build trust and a work rhythm.
Three preparatory steps before implementing AI in SMEs
- Does the problem recur daily/weekly? (if rare, process automation makes no sense).
- Is the data available and reasonably consistent? (AI won't work miracles with "data in people's heads").
- Will someone own the outcome? (one person, not "a team").
If the answer to any of these questions is "no," then don't buy more AI tools for SMEs – first remove that one bottleneck. Then AI solutions act as a lever: they strengthen what is already organized, instead of covering up the mess.
Polish companies that implemented AI correctly – specific examples
Ceramika Paradyż: from chaos in product descriptions to 480,000 EUR in annual savings
Ceramika Paradyż, a Polish ceramic tile manufacturer, regularly added new products to its online store, which required creating descriptions for collections and translating them into several languages. The process consumed hundreds of man-hours and generated high costs. The company could have bought another tool and "dealt with it somehow" – but instead, it stopped and asked the right question: what specific problem do we want to solve?
The answer was simple: automating the creation of product content in multiple languages. They implemented LemonHub.AI from LemonMind – an advanced AI solution for product data management. The result? 80-90% savings in working time, which, with a catalog of 200,000 SKUs, translates into approximately 480,000 EUR in annual savings. The key to success was preparing the
Bempresa: how AI in food production stopped being sci-fi
Bempresa, a company operating in the food industry, partnered with the Polish startup CosmoEye, specializing in image analysis and AI. ERP, MES, CosmoEye AI, and BI systems were implemented, integrating processes and IT systems. This wasn't "we bought AI and we'll see what happens" – it was strategic digitalization with AI as part of the whole.
The CosmoEye AI system analyzed data from various systems as well as industrial CCTV and thermal cameras, contributing to automated monitoring, optimization of production processes, and reduction of failures. The company achieved full process integration and increased operational efficiency. Why did it work? Because before AI, they organized processes, integrated systems (ERP, MES), and only then applied the intelligence layer.
What connects these two examples? Both companies didn't start with AI. They started with a problem, organized the
2026: either you organize your processes, or AI will only cost you
AI in Poland is no longer just a conference topic but is becoming an everyday reality – especially in the SME sector, where cost pressure forces the search for concrete savings. According to 2025 reports, 82% of companies have started implementing AI solutions to some extent, but most are still experimenting instead of measuring results. The future will not belong to companies that "have AI," but to those that know how AI can help with a specific, measurable business problem. Small and medium-sized entrepreneurs face a choice: either invest time in organizing data and processes, or AI will remain an expensive toy with no ROI. Ceramika Paradyż and Bempresa have shown this: artificial intelligence solutions work where
