If AI is going to change the Welsh economy, it has to change it where most Welsh people actually work.
That is why AI must work for Welsh SMEs.
Small and medium-sized enterprises account for around 62.3% of employment in Wales. That single fact should shape the entire debate. If the majority of employment sits in smaller firms, then the real economic value of AI will not be measured only by what happens inside a small number of large sites, public bodies or global technology companies. It will be measured by whether everyday Welsh businesses become more productive, more resilient and more capable of growth.
The productivity argument starts with SMEs
Wales has a long-running productivity challenge. That means policy cannot stop at attracting infrastructure or celebrating innovation at the frontier. The harder task is diffusion: getting useful tools, better processes and stronger capability into the broad base of firms that make up the economy.
That is why SME adoption matters so much. A small improvement repeated across thousands of businesses can do more for the Welsh economy than a much larger improvement confined to a handful of major operators. Better scheduling, faster quoting, fewer errors, better stock planning, improved marketing workflows, quicker reporting and more efficient customer handling may sound incremental. Across the SME base, they add up.
The problem is that adoption is still low. Depending on the measure, only around 7 to 10% of Welsh SMEs are currently using AI, with Senedd research putting the 2022 figure at 7.42%. That is not a side issue. It is the central delivery problem.
Why smaller firms are not moving faster
The barriers are practical, not abstract. Many SMEs do not lack imagination; they lack bandwidth, trusted support and low-risk ways to begin.
A small manufacturer may know that forecasting and quality control can be improved, but not know which provider to trust. A care business may see the potential in documentation and scheduling tools, but worry about governance, training and cost. A tourism operator may want to use AI for customer communications and marketing, but have no time to run a procurement exercise or compare platforms. Rural firms may face additional constraints around connectivity, specialist access and Welsh-language support.
This is why an SME-first strategy cannot be built around slogans about innovation. It has to reduce friction.
What an SME-first model looks like
First, firms need a simple front door. If support is scattered across too many programmes, institutions and acronyms, smaller businesses will often disengage before they start. The system has to be easier to navigate than the status quo.
Second, support should be built around practical projects. The most useful model is not a short workshop followed by silence. It is a defined piece of work inside a business over several months, with help on scoping, implementation, staff adoption and measurement.
Third, policy should recognise that SME adoption is sectoral. A generic approach is too blunt. Manufacturing, food and drink, tourism, construction, logistics, professional services and care all have different use cases, data realities and workforce needs. Sector-based cohorts make more sense because firms can learn from peers facing similar problems.
Fourth, the provider ecosystem matters. Welsh SMEs need access to trusted, accredited providers who can turn interest into implementation. That support should include rural firms and Welsh-language needs, otherwise adoption will cluster only where support is already easiest to find.
This is not an optional add-on
Sometimes SME-focused policy is treated as if it were a fairness issue alone, something nice to do once the “real” investment has been secured. That gets the logic backwards.
An SME-first model is not a social extra bolted onto AI policy. It is the core economic case.
If Growth Zones and AI strategy do not help the wider SME base adopt useful tools, improve workflows and participate in new supply chains, Wales may still host impressive infrastructure while seeing too little change in the firms that drive everyday employment. The country would have succeeded in attracting assets without fully converting them into broad-based performance.
A test Wales should insist on
The right question is simple: will this AI moment make ordinary Welsh firms better at what they do?
If the answer becomes yes across thousands of SMEs, then AI starts to look like a genuine productivity strategy for Wales. If the answer remains no, then the wider economic promise will stay thinner than the headlines suggest.
For Wales to be an AI nation in any meaningful sense, AI has to work for the businesses that already hold most of the economy together.
That means Welsh SMEs cannot be the audience for AI policy after the fact.
They have to be at the centre of it.