Small and medium-sized businesses adopt AI faster than they build 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 be 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 exactly has 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 significantly lag behind in the use of advanced technologies in business, including AI. The key barrier is not technology, but the immaturity of processes and a lack of competencies, especially visible in the SME sector.
And this explains why, in practice, AI in SMEs often operates "alongside" the company, rather than at its core. Tools are used, but they are not part of processes. Automation accelerates individual tasks, but it doesn't change the way the organization operates. AI doesn't fail – the moment the company tries to implement it before tidying up the basics fails.
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 in 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 problem is that in this case, no one stopped to ask what exactly was supposed to change in the company's operations.
In practice, it looks quite repetitive. One solution goes to the marketing team, another to sales, and 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 overseeing the whole. After a few months, a 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 a lack of a coherent approach to digital changes – especially in organizations that are growing rapidly but operationally still rely on intuition and human experience. In such an environment, AI doesn't organize work. It only accelerates 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 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've 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 operation, not individual tests). Polish SMEs do not deviate from this trend. This is the difference between "we have AI" and "AI truly works 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, it works." And then it becomes an "implementation," even though no one has 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 "transformation of work culture."
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 lacks at least minimal
And the third reason—the most human one—is the lack of a "owner" for the topic. In smaller organizations, someone has to connect technology with operations (and not just "test tools"). Without this, AI will grow in a piecemeal fashion: customer service here, documents there, sales somewhere else... and in the end, we still return to the classic: "It's cool, 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, rather than 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 for the improvement of individual activities without interfering with the entire process. The problem is that AI then only accelerates the execution of tasks that were previously poorly designed. The effect is superficial: shorter operational time, but no real improvement in quality or predictability. Automation begins to operate at the "activity" level, not the "process" level—and this is where most of the value disappears.
Where to start AI implementation in a small or medium-sized business
Start with one problem that is currently costing the company real money or time. Not "we want AI," but: "the customer waits too long for a response," "people copy data between systems," "we have errors in documents." If this problem cannot be described in one sentence and calculated simply (minutes, zlotys, number of errors), then AI has nothing to improve—it will only look nice in a demo. The simplest selection rule: choose a problem that is frequent, repeatable, and measurable (then "thanks to AI," the effect can be proven, not just felt). 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 work rhythm.
Three steps to prepare before implementing AI in an SME
- Does the problem recur daily/weekly? (If it's rare, process automation makes no sense).
- Are the data available and reasonably consistent? (AI won't work wonders with "data in people's heads").
- Will someone be the owner of 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 like a lever: they amplify what is already organized, instead of covering up mess.
Polish companies that have implemented AI properly – specific examples
Ceramika Paradyż: from chaos in product descriptions to EUR 480,000 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 "somehow dealt with it"—but instead, it paused and asked the right question: what specific problem do we want to solve?
The answer was simple: automate 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, for a catalog of 200,000 SKUs, translates to approximately EUR 480,000 in annual savings. The key to success was preparing the
Bempresa: how AI in food production stopped being sci-fi
Bempresa, a company 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 to integrate processes and IT systems. This wasn't "we bought AI and we'll see what happens"—it was a strategic digitalization with AI as an integral part of the whole.
The CosmoEye AI system analyzed data from various systems as well as industrial CCTV and thermal imaging cameras, contributing to automated surveillance, optimized production processes, and reduced failures. The company achieved full process integration and increased operational efficiency. Why did it work? Because before AI, they streamlined processes, integrated systems (ERP, MES), and only then applied the layer of intelligence.
What connects these two examples? Both companies didn't start with AI. They started with a problem, organized the
2026: either you streamline processes, or AI will only be a cost
AI in Poland is ceasing to be a conference topic and is becoming an everyday reality – especially in the SME sector, where cost pressure forces the search for specific savings. According to reports from 2025, 82% of companies have started implementing AI-based 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 they invest time in organizing data and processes, or AI will remain an expensive toy without ROI. Ceramika Paradyż and Bempresa have shown this: artificial intelligence solutions work where there are
