What SMEs Get Wrong When They Start With AI, Themes From Our Cyprus AI Expo Panel

By Konstantinos Karamatzianis · Co-Founder, Dabster Labs

Published 7 July 2026 · Based on my remarks at the Cyprus AI Expo panel, 4 July 2026

I was one of the panelists at the Cyprus AI Expo 2026, “The Future of Intelligence,” held at UCLan Cyprus in Larnaca. A few of the questions stuck with me afterwards, mostly because the honest answers aren’t the ones people expect. So I’ve written them up properly here.

Most conversations about AI for small and medium businesses start in the wrong place. They start with the technology, which model, which tool, which platform. In my experience working with SMEs across Cyprus and Greece, the technology is rarely what decides whether an AI project works. The assumptions people bring to it, before they’ve spent a euro, matter far more.

The biggest misconception: a tool is not a transformation

When I was asked about the single biggest misconception, this was my answer: companies believe that buying an AI tool is the transformation. It isn’t. A tool makes a couple of existing tasks faster, genuinely useful, but a long way from transformation, and the gap between the two is where the disappointment lives.

Bolt a piece of AI software onto the way you already work and you speed up a step or two while leaving the underlying process exactly as inefficient as it always was. The real change comes from rethinking the process itself, what actually happens, in what order, which parts genuinely need a person, and only then bringing automation in to support the better version. The tool is maybe the last tenth of the job. The other nine tenths is understanding how your business really runs, and that part gets skipped because it’s harder and far less fun than buying something.

Why you can’t just lift the American playbook

One point I felt strongly about on the day: European organisations lean heavily on AI practices imported straight from the US. It’s understandable, most of the blog posts, the case studies, the “here’s how we did it” write-ups are American. But those playbooks carry assumptions that quietly break the moment you apply them in Europe.

Two of them matter most. The first is that US playbooks simply don’t have GDPR or the EU AI Act in their DNA, they’re written for a legal environment with far fewer constraints on how you process personal data and govern automated decisions. Follow one to the letter and you can end up non-compliant from the foundations up, not through any carelessness, but because compliance was never in the blueprint you copied.

The second is more technical and, honestly, the one that catches people out: the default American stack is US-hosted. Data goes to US-based services as a matter of course. For a European business handling European customers’ data, that’s a cross-border transfer with real legal weight, the kind of thing that becomes a Schrems II headache, introduced silently, just by following someone else’s default setup.

This is the whole reason we build EU-first at Dabster Labs: data residency, lawful processing, and human oversight designed in from the start rather than bolted on when a client’s lawyer finally asks the awkward question. Retrofit compliance and you’re usually rebuilding, because the foundations assumed the opposite of what you now need.

Where to actually start

The rapid-round question was the best one: if a business could do only one thing, what should it be? My answer was deliberately unglamorous. Find one full pain point that can genuinely be automated. Do it, properly, in practice, not as a pilot that never ships. Capture the value. Then, and only then, move to the next process.

Notice what that isn’t. It isn’t the big-bang AI project that tries to transform everything at once, costs a fortune, and collapses under its own ambition. Pick the one repetitive process that quietly eats hours every week, automate that, prove it, measure what it gave back, and let the wins compound from there. Small and proven beats grand and theoretical every single time.

The through-line

If there was a thread running through all three answers, it’s that starting well with AI has almost nothing to do with picking the right tool. Transformation lives in your processes, not your purchases. European businesses need European foundations, not borrowed American defaults. And the way in is one proven automation at a time. Get those assumptions right and the technology turns out to be the easy part.

If you’re an SME trying to work out where AI genuinely fits, and where it’s just noise, that first painful, automatable process is usually easy to spot once someone knows what to look for. Book a free discovery call and we’ll help you find yours.