AI adoption in Welsh SMEs will not be solved by a handful of short pilots, workshops, or generic training sessions. Those things may help introduce the language of AI, but they rarely change how a business operates. If Wales wants AI Growth Zones to raise productivity across the real economy, the focus needs to shift from pilot activity to actual transformation inside firms.
That distinction matters because the productivity challenge is not abstract. Wales has the lowest productivity of any UK nation or region, and SMEs account for around 62.3% of Welsh employment. At the same time, AI adoption among Welsh SMEs remains low, with estimates around 7 to 10%. If AI does not reach the firms where most people work, it will not materially change Wales’s economic trajectory.
Short pilots are attractive because they are easy to announce. They are cheap, quick, and administratively convenient. A programme can recruit a few businesses, run some discovery sessions, test a tool, produce a case study, and claim progress. But that is not the same as embedding AI into the daily operations of an SME. In many cases, pilots stop before the hard work begins.
The hard work is not choosing a chatbot or attending a bootcamp. It is identifying a real business process, understanding the data, redesigning the workflow, training staff, integrating tools, measuring outcomes, and sticking with the change long enough for it to affect performance. That cannot usually be done in a few days or through a generic course that treats every sector as if it has the same needs.
A construction SME does not need the same AI support as a food producer, a tourism business, a logistics operator, a care provider, or a precision engineering firm. Their data is different. Their margins are different. Their operational risks are different. Their staff capacity is different. Their regulatory and customer pressures are different. If AI support is too generic, it becomes interesting but not useful.
From demonstration to implementation
That is why the framework argues for real SME projects rather than short-term pilots. Growth Zone adoption funding should support 6 to 12 month transformation projects in live businesses, organised around sector-based cohorts and measurable outcomes. The point is to move from demonstration to implementation. A serious programme should help firms scope a use case, deploy it, redesign the process around it, and measure whether it improved productivity, quality, cost, resilience, or customer experience.
The difference is practical. A pilot might show that a manufacturer could use AI for quality control. A transformation project helps that manufacturer decide which production line to start with, what data is needed, how staff will use the system, what happens when the model is wrong, how results will be measured, and whether the change improves throughput or reduces defects. One produces a talking point. The other changes a business.
The same applies across other sectors. In tourism, AI might improve demand forecasting, marketing, translation, itinerary planning, or customer service. In food and drink, it might support stock control, compliance documentation, production planning, or sales forecasting. In logistics, it might improve scheduling, routing, maintenance planning, or customer communications. In professional services, it might reduce time spent drafting, researching, summarising, or preparing reports. These are not abstract AI use cases. They are business problems that need implementation support.
Why generic training is not enough
Generic bootcamps also fall short because they often focus on individual skills rather than organisational change. Training is important, but training alone does not create adoption. A member of staff can leave a course more confident with AI tools and still return to a business with poor data, unclear processes, no time to experiment, no budget for integration, and no senior commitment to redesigning work. Skills matter, but they need to be connected to real projects.
That is especially true for smaller firms. SMEs rarely have spare transformation teams sitting around waiting to implement new technology. The owner, finance lead, operations manager, and sales team are often the same small group of people. If an AI programme creates extra work without reducing operational pressure, it will fail. Support needs to be designed around the reality of SME capacity.
This is where sector cohorts can be powerful. Instead of treating each business as an isolated participant, a cohort model brings similar firms through a structured process together. They can learn from each other, compare use cases, share trusted providers, and reduce the risk of adoption. A cohort also gives government and delivery partners a better view of what is actually working in a sector, rather than relying on scattered examples.
But cohorts need to be more than networking groups. They should be built around delivery milestones. Each firm should leave with a defined use case, an implementation plan, support from technical providers, and a way to measure impact. The best programmes will not simply ask, Did the business use AI? They will ask, What changed in the business because AI was used?
What should be measured
That is the productivity question. Did process time fall? Did output per employee rise? Did quality improve? Did errors reduce? Did margins increase? Did staff move into higher-value work? Did the business create a new product or service? Did the firm become more resilient, more competitive, or more able to grow? Those are the measures that matter.
This also explains why voucher-style funding could work well if designed carefully. Vouchers can let SMEs buy practical help from accredited Welsh providers, rather than forcing every firm through a centralised programme. Done well, this would support adoption while also growing the Welsh AI services ecosystem. But the vouchers should be tied to outcomes, not just spend. The test should be whether the support helped a firm make a measurable change.
There also needs to be a rural and Welsh-language lens. If AI adoption funding flows only to already-connected firms in urban clusters, then it will reinforce existing gaps. Rural SMEs often face barriers around digital connectivity, energy costs, skills, and access to support. Welsh-language tools and interfaces also matter if AI adoption is to reach the whole economy, not just the easiest parts of it.
For North Wales, West Wales, the Valleys, and other communities outside the main commercial centres, the question is not whether AI can be demonstrated somewhere. It is whether support reaches firms with real constraints and helps them make practical progress. That means delivery needs to be local enough to understand context, but structured enough to produce measurable outcomes.
Activity or transformation
The danger with pilots is that they make policymakers feel like action is happening while businesses feel little has changed. A programme can generate activity without generating adoption. It can generate attendance without generating productivity. It can generate case studies without changing the underlying economy.
AI Growth Zones should be judged against a higher standard. If each zone receives adoption and skills funding, that money should not be diluted into light-touch activity. It should be concentrated on real projects with real firms. A credible goal would be to double SME AI adoption over the next Senedd term while also increasing the number of Welsh specialist suppliers and AI-enabled firms serving priority sectors.
That will require fewer generic announcements and more disciplined delivery. It means choosing priority sectors, recruiting firms with real problems to solve, pairing them with trusted providers, supporting process redesign, measuring results, and scaling what works. It also means being honest when a project does not deliver, because learning from failure is part of building a serious adoption model.
The right question is not, How many AI pilots did we fund? It is, How many Welsh SMEs changed how they work, improved productivity, and built new capability because of this support?
That is the difference between activity and transformation.
If Wales wants AI Growth Zones to matter beyond the data centres, then short-term pilots will not be enough. Welsh SMEs need practical, sustained, outcome-based support that helps them adopt AI in the real conditions they face. Anything less risks creating a lot of noise around AI without changing the businesses that matter most to the Welsh economy.